Inital commit

src/models/afmoe.cpp

@@ -0,0 +1,285 @@
+#include "models.h"
+
+void llama_model_afmoe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa, false);
+
+    // Set up interleaved sliding window attention (ISWA)
+    // Pattern: 3 sliding - 1 full (global_attn_every_n_layers = 4)
+    if (hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+        uint32_t swa_period = 4;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period);
+
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    } else {
+        hparams.swa_type = LLAMA_SWA_TYPE_NONE;
+    }
+
+    // Default to sigmoid if not set
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+    }
+
+    switch (hparams.n_layer()) {
+        case 56: type = LLM_TYPE_6B; break;
+        case 32: type = LLM_TYPE_26B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_afmoe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    const int64_t n_ff_exp = hparams.n_ff_exp;
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // dual attention normalization
+        layer.attn_norm      = create_tensor(tn(LLM_TENSOR_ATTN_NORM,      "weight", i), {n_embd}, 0);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        // attention projections
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // Q/K normalization
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        // attention gating
+        layer.wqkv_gate = create_tensor(tn(LLM_TENSOR_ATTN_GATE, "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+
+        // dual ffn normalization
+        layer.ffn_norm      = create_tensor(tn(LLM_TENSOR_FFN_NORM,      "weight", i), {n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        if (static_cast<uint32_t>(i) >= hparams.n_layer_dense_lead) {
+            // MoE layers
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0);
+
+            // grouped expert weights
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+
+            // shared expert
+            if (n_expert_shared > 0) {
+                const int64_t n_ff_shexp = n_ff_exp * n_expert_shared;
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, 0);
+            }
+        } else {
+            // Dense layers
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_afmoe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_afmoe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // MuP scaling: embeddings * sqrt(hidden_size)
+    // mup_enabled = true, hidden_size = 1024, scale = 32.0
+    inpL = ggml_scale(ctx0, inpL, sqrtf(float(n_embd)));
+    cb(inpL, "inp_embd_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+    auto * inp_attn = build_attn_inp_kv_iswa();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    const float kq_scale = 1.0f/sqrtf(float(n_embd_head));
+
+    for (int il = 0; il < n_layer; ++il) {
+        const float freq_base_l  = model.get_rope_freq_base (cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        ggml_tensor * inpSA = inpL;
+
+        // This overlaps with SWA layers in current models, so get_rope_freq_base/scale may be superfluous
+        const bool use_rope = hparams.n_no_rope_layer_step > 0 &&
+                              (il + 1) % hparams.n_no_rope_layer_step != 0;
+
+        // dual attention normalization (pre)
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            ggml_tensor * attn_inp = cur;  // save input for gate computation
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // compute gate from input
+            ggml_tensor * gate = build_lora_mm(model.layers[il].wqkv_gate, attn_inp);
+            cb(gate, "attn_gate_proj", il);
+
+            // Q/K normalization
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+            cb(Kcur, "Kcur_normed", il);
+
+            if (use_rope) {
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(Qcur, "Qcur_rope", il);
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(Kcur, "Kcur_rope", il);
+            }
+
+            cur = build_attn(inp_attn,
+                    NULL, NULL, NULL,  // wo will be applied after gating
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+
+            // attention gating: attn_out * sigmoid(gate) BEFORE o_proj
+            gate = ggml_sigmoid(ctx0, gate);
+            cb(gate, "attn_gate_sig", il);
+            cur = ggml_mul(ctx0, cur, gate);
+            cb(cur, "attn_gated", il);
+
+            // now apply output projection
+            cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
+            cb(cur, "attn_o_proj", il);
+        }
+
+        // dual attention normalization (post)
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // dual ffn normalization (pre)
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // MoE or dense FFN
+        if ((uint32_t)il >= hparams.n_layer_dense_lead) {
+            // MoE layer with sigmoid routing, normalization, and scaling
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU,
+                    hparams.expert_weights_norm,           // norm_w (route_norm=True)
+                    hparams.expert_weights_scale,          // w_scale (route_scale=2.826)
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // shared expert
+            if (hparams.n_expert_shared > 0) {
+                ggml_tensor * ffn_shexp = build_ffn(cur,
+                        model.layers[il].ffn_up_shexp,   NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL,
+                        LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            } else {
+                cur = moe_out;
+            }
+        } else {
+            // dense layer
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        // dual ffn normalization (post)
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_post_norm", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/apertus.cpp

@@ -0,0 +1,170 @@
+#include "models.h"
+
+void llama_model_apertus::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    ml.get_key_or_arr(LLM_KV_XIELU_ALPHA_N, hparams.xielu_alpha_n, hparams.n_layer());
+    ml.get_key_or_arr(LLM_KV_XIELU_ALPHA_P, hparams.xielu_alpha_p, hparams.n_layer());
+    ml.get_key_or_arr(LLM_KV_XIELU_BETA,    hparams.xielu_beta,    hparams.n_layer());
+    ml.get_key_or_arr(LLM_KV_XIELU_EPS,     hparams.xielu_eps,     hparams.n_layer());
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_8B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_apertus::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), { n_embd, n_vocab }, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), { n_rot/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), { n_rot/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        } else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), { n_rot/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), { n_embd }, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), { n_embd }, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
+        layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, n_ff }, 0);
+
+        // Q and K layernorms for Apertus
+        layer.attn_q_norm   = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), { n_embd_head_k }, 0);
+        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias",   i), { n_embd_head_k }, TENSOR_NOT_REQUIRED);
+        layer.attn_k_norm   = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), { n_embd_head_k }, 0);
+        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias",   i), { n_embd_head_k }, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_apertus::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_apertus::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    ggml_tensor * inp_pos  = build_inp_pos();
+    auto *        inp_attn = build_attn_inp_kv();
+
+    const float kq_scale =
+        hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur_pos", il);
+            cb(Kcur, "Kcur_pos", il);
+            cb(Vcur, "Vcur_pos", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network with xIELU activation
+        {
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, nullptr, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            // Up projection
+            ggml_tensor * up = build_lora_mm(model.layers[il].ffn_up, cur);
+            cb(up, "ffn_up", il);
+
+            float alpha_n_val = hparams.xielu_alpha_n[il];
+            float alpha_p_val = hparams.xielu_alpha_p[il];
+            float beta_val    = hparams.xielu_beta[il];
+            float eps_val     = hparams.xielu_eps[il];
+
+            // Apply xIELU activation
+            ggml_tensor * activated = ggml_xielu(ctx0, up, alpha_n_val, alpha_p_val, beta_val, eps_val);
+            cb(activated, "ffn_xielu", il);
+
+            // Down projection
+            cur = build_lora_mm(model.layers[il].ffn_down, activated);
+            cb(cur, "ffn_down", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, nullptr, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/arcee.cpp

@@ -0,0 +1,157 @@
+#include "models.h"
+
+void llama_model_arcee::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    // Arcee uses the same structure as Llama
+    switch (hparams.n_layer()) {
+        case 36: type = LLM_TYPE_4B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_arcee::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_arcee::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_arcee::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        // ARCEE uses relu^2 instead of silu
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                NULL,                      NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_RELU_SQR, LLM_FFN_SEQ, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/arctic.cpp

@@ -0,0 +1,180 @@
+#include "models.h"
+
+void llama_model_arctic::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    if (hparams.n_expert == 128) {
+        switch (hparams.n_layer()) {
+            case 35: type = LLM_TYPE_10B_128x3_66B; break;
+            default: type = LLM_TYPE_UNKNOWN;
+        }
+    } else {
+        type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_arctic::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_embd}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_embd, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+        layer.ffn_norm_exps = create_tensor(tn(LLM_TENSOR_FFN_NORM_EXPS, "weight", i), {n_embd}, 0);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, false);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_arctic::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_arctic::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        ggml_tensor * ffn_out = ggml_add(ctx0, cur, ffn_inp);
+        cb(ffn_out, "ffn_out", il);
+
+        // MoE
+        cur = build_norm(inpSA,
+                model.layers[il].ffn_norm_exps, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm_exps", il);
+
+        cur = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(cur, "ffn_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_out);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/arwkv7.cpp

@@ -0,0 +1,202 @@
+#include "models.h"
+
+void llama_model_arwkv7::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,                hparams.f_norm_eps, false);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,            hparams.f_norm_rms_eps, false);
+    ml.get_key(LLM_KV_WKV_HEAD_SIZE,                          hparams.wkv_head_size);
+    ml.get_key(LLM_KV_ATTENTION_DECAY_LORA_RANK,              hparams.n_lora_decay);
+    ml.get_key(LLM_KV_ATTENTION_ICLR_LORA_RANK,               hparams.n_lora_iclr);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_RESIDUAL_MIX_LORA_RANK, hparams.n_lora_value_res_mix);
+    ml.get_key(LLM_KV_ATTENTION_GATE_LORA_RANK,               hparams.n_lora_gate, false);
+    ml.get_key(LLM_KV_TOKEN_SHIFT_COUNT,                      hparams.token_shift_count, false);
+
+    switch (hparams.n_layer()) {
+        case 12:
+            switch (hparams.n_embd) {
+                case 768: type = LLM_TYPE_190M; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 24:
+            switch (hparams.n_embd) {
+                case 1024: type = LLM_TYPE_450M; break;
+                case 2048: type = LLM_TYPE_1_5B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 28:
+            switch (hparams.n_embd) {
+                case 1536: type = LLM_TYPE_1_5B; break;
+                case 3584: type = LLM_TYPE_7B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 32:
+            switch (hparams.n_embd) {
+                case 2560: type = LLM_TYPE_2_9B; break;
+                case 4096: type = LLM_TYPE_7B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 61:
+            switch (hparams.n_embd) {
+                case 4096: type = LLM_TYPE_14B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_arwkv7::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, 0);
+
+    const int n_lora_decay = hparams.n_lora_decay;
+    const int n_lora_iclr = hparams.n_lora_iclr;
+    const int n_lora_value_res_mix = hparams.n_lora_value_res_mix;
+    const int n_lora_gate = hparams.n_lora_gate;
+    const int attn_hidden_size = n_embd;
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.time_mix_w0 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W0, "weight", i), {n_embd}, 0);
+        layer.time_mix_w1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W1, "weight", i), {n_embd, n_lora_decay}, 0);
+        layer.time_mix_w2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W2, "weight", i), {n_lora_decay, n_embd}, 0);
+
+        layer.time_mix_a0 = create_tensor(tn(LLM_TENSOR_TIME_MIX_A0, "weight", i), {n_embd}, 0);
+        layer.time_mix_a1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_A1, "weight", i), {n_embd, n_lora_iclr}, 0);
+        layer.time_mix_a2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_A2, "weight", i), {n_lora_iclr, n_embd}, 0);
+
+        if (i == 0) {
+            // actually not used
+            layer.time_mix_v0 = create_tensor(tn(LLM_TENSOR_TIME_MIX_V0, "weight", i), {n_embd}, 0);
+            layer.time_mix_v1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_V1, "weight", i), {n_embd, n_lora_iclr}, 0);
+            layer.time_mix_v2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_V2, "weight", i), {n_lora_iclr, n_embd}, 0);
+        } else {
+            layer.time_mix_v0 = create_tensor(tn(LLM_TENSOR_TIME_MIX_V0, "weight", i), {n_embd}, 0);
+            layer.time_mix_v1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_V1, "weight", i), {n_embd, n_lora_value_res_mix}, 0);
+            layer.time_mix_v2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_V2, "weight", i), {n_lora_value_res_mix, n_embd}, 0);
+        }
+
+        layer.time_mix_g1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_G1, "weight", i), {n_embd, n_lora_gate}, TENSOR_NOT_REQUIRED);
+        layer.time_mix_g2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_G2, "weight", i), {n_lora_gate, n_embd}, TENSOR_NOT_REQUIRED);
+
+        try {
+            layer.time_mix_lerp_fused = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_FUSED, "weight", i), {n_embd, 1, 1, 6}, 0);
+        } catch(std::runtime_error & e) {
+            // ARWKV models may not have gate tensors
+            layer.time_mix_lerp_fused = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_FUSED, "weight", i), {n_embd, 1, 1, 5}, 0);
+        }
+
+        layer.time_mix_k_k = create_tensor(tn(LLM_TENSOR_TIME_MIX_K_K, "weight", i), {attn_hidden_size}, 0);
+        layer.time_mix_k_a = create_tensor(tn(LLM_TENSOR_TIME_MIX_K_A, "weight", i), {attn_hidden_size}, 0);
+        layer.time_mix_r_k = create_tensor(tn(LLM_TENSOR_TIME_MIX_R_K, "weight", i), {attn_hidden_size}, 0);
+
+        layer.time_mix_key = create_tensor(tn(LLM_TENSOR_TIME_MIX_KEY, "weight", i), {attn_hidden_size, n_embd}, 0);
+        layer.time_mix_value = create_tensor(tn(LLM_TENSOR_TIME_MIX_VALUE, "weight", i), {attn_hidden_size, n_embd}, 0);
+        layer.time_mix_receptance = create_tensor(tn(LLM_TENSOR_TIME_MIX_RECEPTANCE, "weight", i), {attn_hidden_size, n_embd}, 0);
+
+        layer.time_mix_ln = create_tensor(tn(LLM_TENSOR_TIME_MIX_LN, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.time_mix_ln_b = create_tensor(tn(LLM_TENSOR_TIME_MIX_LN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.time_mix_output = create_tensor(tn(LLM_TENSOR_TIME_MIX_OUTPUT, "weight", i), {n_embd, attn_hidden_size}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+
+}
+
+std::unique_ptr<llm_graph_context> llama_model_arwkv7::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_arwkv7::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_build_rwkv7_base(model, params) {
+    GGML_ASSERT(n_embd == hparams.n_embd_r());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+    ggml_tensor * v_first = nullptr;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * rs_inp = build_rs_inp();
+
+    const auto n_embd = hparams.n_embd;
+    const auto n_seq_tokens = ubatch.n_seq_tokens;
+    const auto n_seqs = ubatch.n_seqs;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const llama_layer * layer = &model.layers[il];
+        inpL = ggml_reshape_3d(ctx0, inpL, n_embd, n_seq_tokens, n_seqs);
+
+        ggml_tensor * token_shift = build_rwkv_token_shift_load(rs_inp, ubatch, il);
+
+        ggml_tensor * att_norm = build_norm(inpL, layer->attn_norm, layer->attn_norm_b, LLM_NORM_RMS, il);
+        cb(att_norm, "attn_norm", il);
+
+        ggml_tensor * x_prev = ggml_concat(
+                ctx0,
+                token_shift,
+                ggml_view_3d(ctx0, att_norm, n_embd, n_seq_tokens - 1, n_seqs, att_norm->nb[1], att_norm->nb[2], 0),
+                1
+                );
+
+        cur = build_rwkv7_time_mix(rs_inp, att_norm, x_prev, v_first, ubatch, il);
+
+        token_shift = ggml_view_3d(ctx0, att_norm, n_embd, 1, n_seqs, att_norm->nb[1], att_norm->nb[2], (n_seq_tokens-1)*n_embd*ggml_element_size(att_norm));
+        ggml_build_forward_expand(gf, build_rwkv_token_shift_store(token_shift, ubatch, il));
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur     = ggml_reshape_2d(ctx0, cur,     n_embd, n_tokens);
+        ffn_inp = ggml_reshape_2d(ctx0, ffn_inp, n_embd, n_tokens);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur     = ggml_get_rows(ctx0, cur,     inp_out_ids);
+            ffn_inp = ggml_get_rows(ctx0, ffn_inp, inp_out_ids);
+        }
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+    cur = build_norm(cur, model.output_norm, model.output_norm_b, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/baichuan.cpp

@@ -0,0 +1,155 @@
+#include "models.h"
+
+void llama_model_baichuan::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_7B; break;
+        case 40: type = LLM_TYPE_13B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+
+    if (type == LLM_TYPE_13B) {
+        // TODO: become GGUF KV parameter
+        hparams.f_max_alibi_bias = 8.0f;
+    }
+}
+
+void llama_model_baichuan::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+    {
+        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+        output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_baichuan::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_baichuan::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = model.type == LLM_TYPE_7B ? build_inp_pos() : nullptr;
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            switch (model.type) {
+                case LLM_TYPE_7B:
+                    Qcur = ggml_rope_ext(
+                            ctx0, Qcur, inp_pos, nullptr,
+                            n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow
+                            );
+                    Kcur = ggml_rope_ext(
+                            ctx0, Kcur, inp_pos, nullptr,
+                            n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow
+                            );
+                    break;
+                case LLM_TYPE_13B:
+                case LLM_TYPE_UNKNOWN:
+                    break;
+                default:
+                    GGML_ABORT("fatal error");
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/bailingmoe.cpp

@@ -0,0 +1,180 @@
+#include "models.h"
+
+void llama_model_bailingmoe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+
+    switch (hparams.n_layer()) {
+        case 28: type = LLM_TYPE_16B; break;
+        case 88: type = LLM_TYPE_290B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_bailingmoe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    const int64_t n_ff_exp            = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_head * n_rot, n_head_kv * n_rot, n_head_kv * n_rot, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_rot, n_embd}, 0);
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+        if (n_expert == 0) {
+            throw std::runtime_error("n_expert must be > 0");
+        }
+        if (n_expert_used == 0) {
+            throw std::runtime_error("n_expert_used must be > 0");
+        }
+
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+        layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+        layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+        layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_bailingmoe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_bailingmoe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head_k, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_rot)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        ggml_tensor * moe_out =
+            build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il);
+        cb(moe_out, "ffn_moe_out", il);
+
+        // FFN shared expert
+        {
+            ggml_tensor * ffn_shexp = build_ffn(cur,
+                    model.layers[il].ffn_up_shexp,   NULL, NULL,
+                    model.layers[il].ffn_gate_shexp, NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(ffn_shexp, "ffn_shexp", il);
+
+            cur = ggml_add(ctx0, moe_out, ffn_shexp);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/bailingmoe2.cpp

@@ -0,0 +1,214 @@
+#include "models.h"
+
+void llama_model_bailingmoe2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,         hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,               hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,              hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,               hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,                hparams.expert_gating_func);
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS,              hparams.n_layer_nextn, false);
+
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl");
+
+    switch (hparams.n_layer()) {
+        case 20: type = LLM_TYPE_16B_A1B; break;
+        case 32: type = LLM_TYPE_100B_A6B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_bailingmoe2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    GGML_ASSERT(n_expert > 0 && "n_expert must be > 0 for bailingmoe2");
+    GGML_ASSERT(n_expert_used > 0 && "n_expert_used must be > 0 for bailingmoe2");
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        int flags = 0;
+        if (i >= n_layer) {
+            // skip all tensors in the NextN layers
+            flags |= TENSOR_SKIP;
+        }
+
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, flags);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, flags);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, flags);
+
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, flags);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, flags);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, flags);
+
+        if (static_cast<uint32_t>(i) >= hparams.n_layer_dense_lead) { // MoE layers
+            const int64_t n_ff_shexp = (hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff_exp) * n_expert_shared;
+
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, flags);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED | flags);
+
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, flags);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, flags);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, flags);
+
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, flags);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, flags);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, flags);
+        } else { // Dense layers
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, flags);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, flags);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, flags);
+        }
+
+        // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers
+        if (i >= n_layer) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
+            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED | flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags);
+            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED | flags);
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, TENSOR_NOT_REQUIRED | flags);
+            layer.layer_out_norm         = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, flags);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_bailingmoe2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_bailingmoe2::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * sa_out = ggml_add(ctx0, cur, inpSA);
+        cb(sa_out, "sa_out", il);
+
+        // MoE branch
+        cur = build_norm(sa_out, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if (static_cast<uint32_t>(il) < hparams.n_layer_dense_lead) {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+
+        cur = ggml_add(ctx0, cur, sa_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/bert.cpp

@@ -0,0 +1,233 @@
+#include "models.h"
+
+void llama_model_bert::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 3:
+            type = LLM_TYPE_17M; break; // bge-micro
+        case 6:
+            type = LLM_TYPE_22M; break; // MiniLM-L6
+        case 12:
+            switch (hparams.n_embd) {
+                case 384: type = LLM_TYPE_33M; break; // MiniLM-L12, bge-small
+                case 768: type = LLM_TYPE_109M; break; // bge-base
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 24:
+            type = LLM_TYPE_335M; break; // bge-large
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_bert::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_token_types == 0) {
+        throw std::runtime_error(arch_name() + " model needs to define token type count");
+    }
+    tok_embd     = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0);
+    type_embd    = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, TENSOR_NOT_REQUIRED);
+
+    if (arch == LLM_ARCH_BERT) {
+        pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,    "weight"), {n_embd, n_ctx_train}, 0);
+
+        cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+        cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {n_embd},         TENSOR_NOT_REQUIRED);
+
+        cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+        cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+    }
+
+    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0);
+    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias",   0), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i),   {n_embd}, 0);
+
+        if (hparams.moe_every_n_layers > 0 && i % hparams.moe_every_n_layers == 1) {
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff,   n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff,   n_embd, n_expert}, 0);
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,   "weight", i), {n_embd, n_expert}, 0);
+        } else {
+            layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+            if (arch == LLM_ARCH_NOMIC_BERT) {
+                layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            }
+        }
+
+        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i),   {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_bert::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_bert::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+    ggml_tensor * inp_pos = nullptr;
+
+    if (model.arch != LLM_ARCH_JINA_BERT_V2) {
+        inp_pos = build_inp_pos();
+    }
+
+    // construct input embeddings (token, type, position)
+    inpL = build_inp_embd(model.tok_embd);
+
+    // token types are hardcoded to zero ("Sentence A")
+    if (model.type_embd) {
+        ggml_tensor * type_row0 = ggml_view_1d(ctx0, model.type_embd, n_embd, 0);
+        inpL                    = ggml_add(ctx0, inpL, type_row0);
+    }
+    if (model.arch == LLM_ARCH_BERT) {
+        inpL = ggml_add(ctx0, ggml_get_rows(ctx0, model.pos_embd, inp_pos), inpL);
+    }
+    cb(inpL, "inp_embd", -1);
+
+    // embed layer norm
+    inpL = build_norm(inpL, model.tok_norm, model.tok_norm_b, LLM_NORM, 0);
+    cb(inpL, "inp_norm", 0);
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * cur = inpL;
+
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (model.layers[il].attn_q_norm) {
+                Qcur = ggml_reshape_2d(ctx0, Qcur, n_embd_head * n_head, n_tokens);
+
+                Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, model.layers[il].attn_q_norm_b, LLM_NORM, il);
+
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+            }
+
+            if (model.layers[il].attn_k_norm) {
+                Kcur = ggml_reshape_2d(ctx0, Kcur, n_embd_head * n_head_kv, n_tokens);
+
+                Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, model.layers[il].attn_k_norm_b, LLM_NORM, il);
+
+                Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            }
+
+            // RoPE
+            if (model.arch == LLM_ARCH_NOMIC_BERT || model.arch == LLM_ARCH_NOMIC_BERT_MOE ||
+                model.arch == LLM_ARCH_JINA_BERT_V3) {
+                Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                     ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                     ext_factor, attn_factor, beta_fast, beta_slow);
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+            cb(cur, "kqv_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // re-add the layer input
+        cur = ggml_add(ctx0, cur, inpL);
+
+        // attention layer norm
+        cur = build_norm(cur, model.layers[il].attn_out_norm, model.layers[il].attn_out_norm_b, LLM_NORM, il);
+
+        if (model.layers[il].attn_norm_2 != nullptr) {
+            cur = ggml_add(ctx0, cur, inpL);  // re-add the layer input
+            cur = build_norm(cur, model.layers[il].attn_norm_2, model.layers[il].attn_norm_2_b, LLM_NORM, il);
+        }
+
+        ggml_tensor * ffn_inp = cur;
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        if (hparams.moe_every_n_layers > 0 && il % hparams.moe_every_n_layers == 1) {
+            // MoE branch
+            cur = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    nullptr,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    hparams.n_expert, hparams.n_expert_used,
+                    LLM_FFN_GELU, false,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il);
+            cb(cur, "ffn_moe_out", il);
+        } else if (model.arch == LLM_ARCH_BERT || model.arch == LLM_ARCH_NOMIC_BERT_MOE ||
+                   model.arch == LLM_ARCH_JINA_BERT_V3) {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
+                    NULL, NULL, NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL, NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+        } else if (model.arch == LLM_ARCH_JINA_BERT_V2) {
+            const bool up_contains_gate = !model.layers[il].ffn_gate && model.layers[il].ffn_up->ne[1] != hparams.n_ff();
+            auto type_op = up_contains_gate ? LLM_FFN_GEGLU : LLM_FFN_GELU;
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL, NULL,
+                    type_op, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        // attentions bypass the intermediate layer
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        // output layer norm
+        cur = build_norm(cur, model.layers[il].layer_out_norm, model.layers[il].layer_out_norm_b, LLM_NORM, il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cb(cur, "result_embd", -1);
+    res->t_embd = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/bitnet.cpp

@@ -0,0 +1,170 @@
+#include "models.h"
+
+void llama_model_bitnet::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 26: type = LLM_TYPE_3B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_bitnet::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm     = create_tensor(tn(LLM_TENSOR_ATTN_NORM,     "weight", i), {n_embd}, 0);
+        layer.attn_sub_norm = create_tensor(tn(LLM_TENSOR_ATTN_SUB_NORM, "weight", i), {n_embd}, 0);
+
+        layer.wq       = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+        layer.wq_s     = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+        layer.wk       = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+        layer.wk_s     = create_tensor(tn(LLM_TENSOR_ATTN_K,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+        layer.wv       = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+        layer.wv_s     = create_tensor(tn(LLM_TENSOR_ATTN_V,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+        layer.wo       = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_s     = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm     = create_tensor(tn(LLM_TENSOR_FFN_NORM,     "weight", i), {n_embd}, 0);
+        layer.ffn_sub_norm = create_tensor(tn(LLM_TENSOR_FFN_SUB_NORM, "weight", i), {n_ff}, 0);
+
+        layer.ffn_gate       = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_gate_s = create_tensor(tn(LLM_TENSOR_FFN_GATE, "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down       = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_s = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up         = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up_s   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_bitnet::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_bitnet::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    NULL, NULL, NULL,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+
+            cur = build_norm(cur,
+                    model.layers[il].attn_sub_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "attn_sub_norm", il);
+
+            cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
+            if (model.layers[il].wo_b) {
+                cur = ggml_add(ctx0, cur, model.layers[il].wo_b);
+            }
+            cb(cur, "attn_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward forward
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, model.layers[il].ffn_up_s,
+                model.layers[il].ffn_gate, NULL, model.layers[il].ffn_gate_s,
+                NULL,                      NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_sub_out", il);
+
+        cur = build_norm(cur,
+                model.layers[il].ffn_sub_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_sub_norm", il);
+
+        cur = build_lora_mm(model.layers[il].ffn_down, cur, model.layers[il].ffn_down_s);
+        cb(cur, "ffn_down", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "l_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    // FIXME: do not use model.tok_embd directly, duplicate as model.output
+    cur = build_lora_mm(model.tok_embd, cur);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/bloom.cpp

@@ -0,0 +1,151 @@
+#include "models.h"
+
+void llama_model_bloom::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 24: type = LLM_TYPE_1B; break;
+        case 30:
+            switch (hparams.n_embd) {
+                case 2560: type = LLM_TYPE_3B; break;
+                case 4096: type = LLM_TYPE_7B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+
+    // TODO: become GGUF KV parameter
+    hparams.f_max_alibi_bias = 8.0f;
+}
+
+void llama_model_bloom::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,      "weight"), {n_embd, n_vocab}, 0);
+    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0);
+    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias",   0), {n_embd}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias",   i), {n_embd}, 0);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wqkv_b = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, 0);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias",   i), {n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias",   i), {n_embd}, 0);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, 0);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias",   i), {n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_bloom::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_bloom::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    inpL = build_norm(inpL,
+            model.tok_norm,
+            model.tok_norm_b,
+            LLM_NORM, 0);
+    cb(inpL, "inp_norm", 0);
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // Add the input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FF
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    NULL,                      NULL,                        NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = build_norm(inpL,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/chameleon.cpp

@@ -0,0 +1,204 @@
+#include "models.h"
+#include <float.h>
+
+void llama_model_chameleon::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    hparams.f_norm_eps = 1e-5;  // eps for qk-norm, torch default
+    ml.get_key(LLM_KV_SWIN_NORM, hparams.swin_norm, false);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_7B; break;
+        case 48: type = LLM_TYPE_34B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+   }
+}
+
+void llama_model_chameleon::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}, 0);
+        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i),  {n_embd_head_k, n_head}, TENSOR_NOT_REQUIRED);
+        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias", i),  {n_embd_head_k, n_head_kv}, TENSOR_NOT_REQUIRED);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_chameleon::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_chameleon::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        if (hparams.swin_norm) {
+            cur = inpL;
+        } else {
+            cur = build_norm(inpL,
+                    model.layers[il].attn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+        }
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (model.layers[il].attn_q_norm) {
+                Qcur = build_norm(Qcur,
+                        model.layers[il].attn_q_norm,
+                        model.layers[il].attn_q_norm_b,
+                        LLM_NORM, il);
+                cb(Qcur, "Qcur", il);
+            }
+
+            if (model.layers[il].attn_k_norm) {
+                Kcur = build_norm(Kcur,
+                        model.layers[il].attn_k_norm,
+                        model.layers[il].attn_k_norm_b,
+                        LLM_NORM, il);
+                cb(Kcur, "Kcur", il);
+            }
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, nullptr, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        if (hparams.swin_norm) {
+            cur = build_norm(cur,
+                    model.layers[il].attn_norm, NULL,
+                    LLM_NORM_RMS, il);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        if (!hparams.swin_norm) {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+        }
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        if (hparams.swin_norm) {
+            cur = build_norm(cur,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output_with_img_logits", -1);
+
+    // TODO: this suppresses the output of image tokens, which is required to enable text-only outputs.
+    // Needs to be removed once image outputs are supported.
+    int img_token_end_idx = 8196;
+    int img_token_start_idx = 4;
+    int num_img_tokens = img_token_end_idx - img_token_start_idx;
+    // creates 1d tensor of size num_img_tokens and values -FLT_MAX,
+    // which ensures that text token values are always at least larger than image token values
+    ggml_tensor * img_logits = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, num_img_tokens);
+    img_logits = ggml_clamp(ctx0, img_logits, -FLT_MAX, -FLT_MAX);
+    cb(img_logits, "img_logits", -1);
+
+    cur = ggml_set_1d(ctx0, cur, img_logits, ggml_element_size(cur) * img_token_start_idx);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/chatglm.cpp

@@ -0,0 +1,161 @@
+#include "models.h"
+
+void llama_model_chatglm::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 28: {
+            if (hparams.n_head(0) == 16) {
+                type = LLM_TYPE_1_5B;
+            } else {
+                type = LLM_TYPE_6B;
+            }
+        } break;
+        case 40: {
+            if (hparams.n_head(0) == 24) {
+                type = LLM_TYPE_4B;
+            } else {
+                type = LLM_TYPE_9B;
+            }
+        } break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_chatglm::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,      "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff * 2}, 0);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_chatglm::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_chatglm::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            //printf("freq_base: %f freq_scale: %f ext_factor: %f attn_factor: %f\n", freq_base, freq_scale, ext_factor, attn_factor);
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        // Add the input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FF
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm,
+                    NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    NULL,                      NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SWIGLU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = build_norm(inpL,
+            model.output_norm,
+            NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/codeshell.cpp

@@ -0,0 +1,153 @@
+#include "models.h"
+
+void llama_model_codeshell::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 42: type = LLM_TYPE_7B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_codeshell::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if tok embd is NULL, init from output
+    if (tok_embd == NULL) {
+        tok_embd = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i),   {n_embd, n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i),     {n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_codeshell::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_codeshell::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // add the input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FF
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    NULL,                      NULL,                        NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = build_norm(inpL,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/cogvlm.cpp

@@ -0,0 +1,158 @@
+#include "models.h"
+
+void llama_model_cogvlm::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_13B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_cogvlm::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd_head_k * n_head * 3}, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.visexp_attn_wqkv = create_tensor(tn(LLM_TENSOR_VISEXP_ATTN_QKV, "weight", i), {n_embd, n_embd_head_k * n_head * 3}, 0);
+        layer.visexp_attn_wo = create_tensor(tn(LLM_TENSOR_VISEXP_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+        layer.visexp_ffn_gate = create_tensor(tn(LLM_TENSOR_VISEXP_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.visexp_ffn_down = create_tensor(tn(LLM_TENSOR_VISEXP_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.visexp_ffn_up   = create_tensor(tn(LLM_TENSOR_VISEXP_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_cogvlm::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_cogvlm::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    const float   kq_scale    = 1.0f / sqrtf(float(n_embd_head));
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * inpL;
+    ggml_tensor * cur;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    // check ubatch to see if we have input tokens (text)
+    // or an input embedding vector (image)
+    bool is_text;
+    if (ubatch.token) {
+        is_text = true;
+    } else {
+        is_text = false;
+    }
+
+    for (int il = 0; il < n_layer; ++il) {
+        // get either the text or image weight tensors
+        ggml_tensor *wqkv, *wo, *wo_s;
+        ggml_tensor *ffn_gate, *ffn_down, *ffn_up;
+
+        if (is_text) {
+            wqkv     = model.layers[il].wqkv;
+            wo       = model.layers[il].wo;
+            wo_s     = model.layers[il].wo_s;
+            ffn_gate = model.layers[il].ffn_gate;
+            ffn_down = model.layers[il].ffn_down;
+            ffn_up   = model.layers[il].ffn_up;
+        } else {
+            wqkv     = model.layers[il].visexp_attn_wqkv;
+            wo       = model.layers[il].visexp_attn_wo;
+            wo_s     = nullptr;
+            ffn_gate = model.layers[il].visexp_ffn_gate;
+            ffn_down = model.layers[il].visexp_ffn_down;
+            ffn_up   = model.layers[il].visexp_ffn_up;
+        }
+
+        ggml_tensor * inpSA = inpL;
+        cur = build_norm(inpSA, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+
+        // build self attention
+        {
+            ggml_tensor * qkv = build_lora_mm(wqkv, cur);
+
+            // split qkv into Q, K, V along the first dimension
+            ggml_tensor * Qcur =
+                ggml_view_3d(ctx0, qkv, n_embd_head, n_head, n_tokens, n_embd_head * sizeof(float), qkv->nb[1], 0);
+            ggml_tensor * Kcur = ggml_view_3d(ctx0, qkv, n_embd_head, n_head_kv, n_tokens, n_embd_head * sizeof(float),
+                                              qkv->nb[1], n_embd * ggml_element_size(qkv));
+            ggml_tensor * Vcur = ggml_view_3d(ctx0, qkv, n_embd_head, n_head_kv, n_tokens, n_embd_head * sizeof(float),
+                                              qkv->nb[1], 2 * n_embd * ggml_element_size(qkv));
+
+            Qcur = ggml_rope(ctx0, Qcur, inp_pos, n_embd_head, rope_type);
+            Kcur = ggml_rope(ctx0, Kcur, inp_pos, n_embd_head, rope_type);
+
+            cur = build_attn(inp_attn,
+                wo, nullptr, wo_s,
+                Qcur, Kcur, Vcur,
+                nullptr, nullptr, nullptr,
+                kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                ffn_up, NULL, NULL,
+                ffn_gate, NULL, NULL,
+                ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+    ggml_build_forward_expand(gf, cur);
+}

src/models/cohere2.cpp

@@ -0,0 +1,161 @@
+#include "models.h"
+
+void llama_model_cohere2::load_arch_hparams(llama_model_loader & ml) {
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    uint32_t swa_period = 4;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+    hparams.set_swa_pattern(swa_period);
+
+    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,       hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa);
+    ml.get_key(LLM_KV_LOGIT_SCALE,              hparams.f_logit_scale);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,  hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_8B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_cohere2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+    // init output from the input tok embed
+    output      = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab },
+                                      TENSOR_DUPLICATED);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd, n_embd }, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, n_ff }, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_cohere2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_cohere2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    const float f_logit_scale = hparams.f_logit_scale;
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv_iswa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const bool is_swa = hparams.is_swa(il);
+        // UNUSED:
+        // const float freq_base_l  = model.get_rope_freq_base (cparams, il);
+        // const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+        ggml_tensor * ffn_inp = cur;
+
+        // self-attention
+        {
+            // rope freq factors for 128k context
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (is_swa) {
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, rope_factors,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, rope_factors,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur     = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL    = ggml_get_rows(ctx0, inpL, inp_out_ids);
+            ffn_inp = ggml_get_rows(ctx0, ffn_inp, inp_out_ids);
+        }
+
+        ggml_tensor * attn_out = cur;
+
+        // feed-forward network
+        {
+            cur = build_ffn(ffn_inp,
+                    model.layers[il].ffn_up, NULL, model.layers[il].ffn_up_s,
+                    model.layers[il].ffn_gate, NULL, model.layers[il].ffn_gate_s,
+                    model.layers[il].ffn_down, NULL, model.layers[il].ffn_down_s,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        // add together residual + FFN + self-attention
+        cur = ggml_add(ctx0, cur, inpL);
+        cur = ggml_add(ctx0, cur, attn_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    if (f_logit_scale) {
+        cur = ggml_scale(ctx0, cur, f_logit_scale);
+    }
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/cohere2moe.cpp

@@ -0,0 +1,443 @@
+#include "models.h"
+
+void llama_model_cohere2moe::load_arch_hparams(llama_model_loader & ml) {
+    const bool found_norm     = ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,     hparams.f_norm_eps,     false);
+    const bool found_norm_rms = ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps, false);
+    if (!found_norm && !found_norm_rms) {
+        throw std::runtime_error("missing Cohere2 MoE norm epsilon");
+    }
+    if (!found_norm_rms) {
+        hparams.f_norm_rms_eps = 0.0f;
+    }
+
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa);
+    ml.get_key(LLM_KV_LOGIT_SCALE,                 hparams.f_logit_scale);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS,        hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer");
+
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+    }
+
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    uint32_t swa_period = 4;
+    if (ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false)) {
+        hparams.set_swa_pattern(swa_period, true);
+    } else {
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+    }
+
+    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+
+    switch (hparams.n_layer()) {
+        case 49: type = LLM_TYPE_30B_A3B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_cohere2moe::load_arch_tensors(llama_model_loader & ml) {
+    LLAMA_LOAD_LOCALS;
+
+    const bool mtp_only = (hparams.n_layer_nextn > 0) && (ml.get_weight("blk.0.attn_norm.weight") == nullptr);
+    // Trunk-only: the GGUF declares MTP layers in metadata but the actual MTP
+    // tensors live in a separate file. Mark MTP tensors NOT_REQUIRED so the
+    // trunk loads cleanly.
+    const std::string mtp_probe = "blk." + std::to_string(n_layer) + ".nextn.eh_proj.weight";
+    const bool trunk_only = (hparams.n_layer_nextn > 0) && (ml.get_weight(mtp_probe.c_str()) == nullptr);
+    const int trunk_flags = mtp_only  ? TENSOR_NOT_REQUIRED : 0;
+    const int mtp_flags   = trunk_only ? TENSOR_NOT_REQUIRED : 0;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
+    }
+
+    if (n_expert == 0) {
+        throw std::runtime_error("n_expert must be > 0 for Cohere2Moe");
+    }
+    if (n_expert_used == 0) {
+        throw std::runtime_error("n_expert_used must be > 0 for Cohere2Moe");
+    }
+
+    auto load_block_trunk = [&](int i, int flags) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, flags);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, flags);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, flags);
+
+        if (static_cast<uint32_t>(i) < hparams.n_layer_dense_lead) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, flags);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, flags);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), { n_embd, n_ff }, flags);
+        } else {
+            const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff;
+
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), { n_embd, n_expert }, flags);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff_exp, n_embd, n_expert }, flags);
+            create_tensor_gate_up_exps(layer, i, n_embd, n_ff_exp, n_expert, flags);
+
+            if (hparams.n_expert_shared > 0) {
+                const int64_t n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff_exp * hparams.n_expert_shared;
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), { n_embd, n_ff_shexp }, flags);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_shexp, n_embd }, flags);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), { n_embd, n_ff_shexp }, flags);
+            }
+        }
+    };
+
+    auto load_block_mtp = [&](int i, int flags) {
+        auto & layer = layers[i];
+
+        // MTP block looks like a full-attention Cohere2 MoE decoder block.
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, flags);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, flags);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, flags);
+
+        const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff;
+
+        // Routed experts
+        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), { n_embd, n_expert }, flags);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff_exp, n_embd, n_expert }, flags);
+        create_tensor_gate_up_exps(layer, i, n_embd, n_ff_exp, n_expert, flags);
+
+        if (hparams.n_expert_shared > 0) {
+            const int64_t n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff_exp * hparams.n_expert_shared;
+
+            // Shared experts
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), { n_embd, n_ff_shexp }, flags);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_shexp, n_embd }, flags);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), { n_embd, n_ff_shexp }, flags);
+        }
+
+        // NextN-specific tensors that define the MTP block.
+        layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ,          "weight", i), { 2 * n_embd, n_embd }, flags);
+        layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM,            "weight", i), { n_embd },              flags);
+        layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM,            "weight", i), { n_embd },              flags);
+        layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS,     "weight", i), { n_embd, n_vocab },     TENSOR_NOT_REQUIRED);
+        layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab },     TENSOR_NOT_REQUIRED);
+        layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd },              TENSOR_NOT_REQUIRED);
+    };
+
+    for (int i = 0; i < n_layer; ++i) {
+        load_block_trunk(i, trunk_flags);
+    }
+    // MTP/NextN layers are loaded as extra decoder blocks.
+    for (int i = n_layer; i < n_layer_all; ++i) {
+        load_block_mtp(i, mtp_flags);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_cohere2moe::build_arch_graph(const llm_graph_params & params) const {
+    if (params.gtype == LLM_GRAPH_TYPE_DECODER_MTP) {
+        return std::make_unique<graph_mtp>(*this, params);
+    }
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_cohere2moe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    const llm_norm_type cohere2moe_norm_type = hparams.f_norm_rms_eps == 0.0f ? LLM_NORM : LLM_NORM_RMS;
+    const float f_logit_scale = hparams.f_logit_scale;
+    ggml_tensor * cur;
+    ggml_tensor * inpL = build_inp_embd(model.tok_embd);
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv_iswa();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    // MTP/NextN layers are loaded as extra decoder blocks but not executed in the main pass.
+    for (int il = 0; il < n_layer; ++il) {
+        const bool is_swa = hparams.is_swa(il);
+        // Dense-prefix full-attention layers use RoPE; later layers follow the SWA pattern.
+        const bool force_rope = static_cast<uint32_t>(il) < hparams.n_layer_dense_lead;
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, cohere2moe_norm_type, il);
+        cb(cur, "attn_norm", il);
+
+        ggml_tensor * ffn_inp = cur;
+
+        {
+            const auto & layer = model.layers[il];
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(layer, cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (is_swa || force_rope) {
+                ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, rope_factors,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, rope_factors,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    layer.wo, layer.wo_b, layer.wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr,
+                    1.0f / sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) {
+            cur     = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL    = ggml_get_rows(ctx0, inpL, inp_out_ids);
+            ffn_inp = ggml_get_rows(ctx0, ffn_inp, inp_out_ids);
+        }
+
+        ggml_tensor * attn_out = cur;
+
+        const auto & layer = model.layers[il];
+
+        if (layer.ffn_gate_inp == nullptr) {
+            cur = build_ffn(ffn_inp,
+                    layer.ffn_up,   nullptr, layer.ffn_up_s,
+                    layer.ffn_gate, nullptr, layer.ffn_gate_s,
+                    layer.ffn_down, nullptr, layer.ffn_down_s,
+                    nullptr, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            cur = build_moe_ffn(ffn_inp,
+                    layer.ffn_gate_inp,
+                    layer.ffn_up_exps,
+                    layer.ffn_gate_exps,
+                    layer.ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il,
+                    nullptr, layer.ffn_gate_up_exps,
+                    layer.ffn_up_exps_s,
+                    layer.ffn_gate_exps_s,
+                    layer.ffn_down_exps_s);
+            cb(cur, "ffn_moe_out", il);
+
+            if (layer.ffn_up_shexp) {
+                ggml_tensor * ffn_shexp = build_ffn(ffn_inp,
+                        layer.ffn_up_shexp,   nullptr, layer.ffn_up_shexp_s,
+                        layer.ffn_gate_shexp, nullptr, layer.ffn_gate_shexp_s,
+                        layer.ffn_down_shexp, nullptr, layer.ffn_down_shexp_s,
+                        nullptr, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, cur, ffn_shexp);
+                cur = ggml_scale(ctx0, cur, 0.5f);
+                cb(cur, "ffn_out", il);
+            }
+        }
+
+        cur = ggml_add(ctx0, cur, inpL);
+        cur = ggml_add(ctx0, cur, attn_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        inpL = cur;
+    }
+
+    cur = inpL;
+    cur = build_norm(cur, model.output_norm, nullptr, cohere2moe_norm_type, -1);
+
+    cb(cur, "h_nextn", -1);
+    res->t_h_nextn = cur;
+
+    if (!cparams.embeddings_nextn_masked && inp_out_ids) {
+        cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+    }
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur);
+
+    if (f_logit_scale) {
+        cur = ggml_scale(ctx0, cur, f_logit_scale);
+    }
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+llama_model_cohere2moe::graph_mtp::graph_mtp(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    GGML_ASSERT(hparams.n_layer_nextn > 0 && "COHERE2MOE MTP requires n_layer_nextn > 0");
+    GGML_ASSERT(hparams.n_layer_nextn == 1 && "COHERE2MOE MTP currently only supports a single MTP block");
+
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    const int il = hparams.n_layer();
+    const auto & layer = model.layers[il];
+    GGML_ASSERT(layer.nextn.eh_proj && "MTP block missing nextn.eh_proj");
+    GGML_ASSERT(layer.nextn.enorm   && "MTP block missing nextn.enorm");
+    GGML_ASSERT(layer.nextn.hnorm   && "MTP block missing nextn.hnorm");
+    GGML_ASSERT(layer.ffn_gate_inp  && "MTP block missing ffn_gate_inp");
+
+    const llm_norm_type cohere2moe_norm_type = hparams.f_norm_rms_eps == 0.0f ? LLM_NORM : LLM_NORM_RMS;
+
+    // TODO: extract in a common llm_graph_context::build_inp_embd_h()
+    auto inp = std::make_unique<llm_graph_input_embd_h>(hparams.n_embd);
+
+    inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
+    ggml_set_input(inp->tokens);
+
+    inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd_inp(), n_tokens);
+    ggml_set_input(inp->embd);
+
+    // TODO: make static using `ggml_build_forward_select()`
+    //       see llm_graph_context::build_inp_embd() for reference
+    ggml_tensor * tok_embd;
+    if (ubatch.token) {
+        ggml_tensor * tok_embd_w = layer.nextn.embed_tokens ? layer.nextn.embed_tokens : model.tok_embd;
+        tok_embd = ggml_get_rows(ctx0, tok_embd_w, inp->tokens);
+    } else {
+        tok_embd = inp->embd;
+    }
+    cb(tok_embd, "mtp_tok_embd", il);
+
+    inp->h = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd, n_tokens);
+    ggml_set_input(inp->h);
+    ggml_set_name(inp->h, "mtp_h_input");
+
+    ggml_tensor * h_embd = inp->h;
+
+    res->add_input(std::move(inp));
+
+    ggml_tensor * inp_pos     = build_inp_pos();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+    auto * inp_attn = build_attn_inp_kv_iswa();
+
+    ggml_tensor * h_norm = build_norm(h_embd, layer.nextn.hnorm, nullptr, cohere2moe_norm_type, il);
+    cb(h_norm, "mtp_hnorm", il);
+
+    ggml_tensor * e_norm = build_norm(tok_embd, layer.nextn.enorm, nullptr, cohere2moe_norm_type, il);
+    cb(e_norm, "mtp_enorm", il);
+
+    ggml_tensor * concat = ggml_concat(ctx0, e_norm, h_norm, /*dim=*/ 0);
+    cb(concat, "mtp_concat", il);
+
+    ggml_tensor * cur = build_lora_mm(layer.nextn.eh_proj, concat, layer.nextn.eh_proj_s);
+    cb(cur, "mtp_eh_proj", il);
+
+    ggml_tensor * inpL = cur;
+
+    cur = build_norm(cur, layer.attn_norm, nullptr, cohere2moe_norm_type, il);
+    cb(cur, "mtp_attn_norm", il);
+    ggml_tensor * ffn_inp = cur;
+
+    auto [Qcur, Kcur, Vcur] = build_qkv(layer, cur, n_embd_head, n_head, n_head_kv, il);
+    ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+    Qcur = ggml_rope_ext(
+            ctx0, Qcur, inp_pos, rope_factors,
+            n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+            ext_factor, attn_factor, beta_fast, beta_slow);
+    Kcur = ggml_rope_ext(
+            ctx0, Kcur, inp_pos, rope_factors,
+            n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+            ext_factor, attn_factor, beta_fast, beta_slow);
+
+    cb(Qcur, "mtp_Qcur", il);
+    cb(Kcur, "mtp_Kcur", il);
+    cb(Vcur, "mtp_Vcur", il);
+
+    cur = build_attn(inp_attn,
+            layer.wo, layer.wo_b, layer.wo_s,
+            Qcur, Kcur, Vcur, nullptr, nullptr, nullptr,
+            1.0f / sqrtf(float(n_embd_head)), il);
+    cb(cur, "mtp_attn_out", il);
+
+    ggml_tensor * attn_out = cur;
+
+    cur = build_moe_ffn(ffn_inp,
+            layer.ffn_gate_inp,
+            layer.ffn_up_exps,
+            layer.ffn_gate_exps,
+            layer.ffn_down_exps,
+            nullptr,
+            n_expert, n_expert_used,
+            LLM_FFN_SILU, hparams.expert_weights_norm,
+            hparams.expert_weights_scale,
+            (llama_expert_gating_func_type) hparams.expert_gating_func,
+            il,
+            nullptr, layer.ffn_gate_up_exps,
+            layer.ffn_up_exps_s,
+            layer.ffn_gate_exps_s,
+            layer.ffn_down_exps_s);
+    cb(cur, "mtp_ffn_moe_out", il);
+
+    if (layer.ffn_up_shexp) {
+        ggml_tensor * ffn_shexp = build_ffn(ffn_inp,
+                layer.ffn_up_shexp,   nullptr, layer.ffn_up_shexp_s,
+                layer.ffn_gate_shexp, nullptr, layer.ffn_gate_shexp_s,
+                layer.ffn_down_shexp, nullptr, layer.ffn_down_shexp_s,
+                nullptr, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(ffn_shexp, "mtp_ffn_shexp", il);
+
+        cur = ggml_add(ctx0, cur, ffn_shexp);
+        cur = ggml_scale(ctx0, cur, 0.5f);
+        cb(cur, "mtp_ffn_out", il);
+    }
+
+    cur = ggml_add(ctx0, cur, inpL);
+    cur = ggml_add(ctx0, cur, attn_out);
+    cb(cur, "mtp_post_ffn", il);
+
+    ggml_tensor * head_norm_w = layer.nextn.shared_head_norm
+            ? layer.nextn.shared_head_norm
+            : model.output_norm;
+    GGML_ASSERT(head_norm_w && "COHERE2MOE MTP: missing both nextn.shared_head_norm and output_norm");
+    cur = build_norm(cur, head_norm_w, nullptr, cohere2moe_norm_type, -1);
+
+    cb(cur, "h_nextn", -1);
+    res->t_h_nextn = cur;
+
+    cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+    cb(cur, "mtp_shared_head_norm", -1);
+
+    ggml_tensor * head_w = layer.nextn.shared_head_head ? layer.nextn.shared_head_head : model.output;
+    GGML_ASSERT(head_w && "COHERE2MOE MTP: missing LM head (nextn.shared_head_head or model.output)");
+    cur = build_lora_mm(head_w, cur, layer.nextn.shared_head_head ? layer.nextn.shared_head_head_s : nullptr);
+
+    if (hparams.f_logit_scale) {
+        cur = ggml_scale(ctx0, cur, hparams.f_logit_scale);
+    }
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/command-r.cpp

@@ -0,0 +1,144 @@
+#include "models.h"
+
+void llama_model_command_r::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_LOGIT_SCALE,             hparams.f_logit_scale, false);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 40: type = LLM_TYPE_35B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_command_r::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    // init output from the input tok embed
+    output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (n_layer >= 64){
+            layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head}, 0);
+            layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}, 0);
+        }
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_command_r::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_command_r::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    const float f_logit_scale = hparams.f_logit_scale;
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        ggml_tensor * ffn_inp = cur;
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (model.layers[il].attn_q_norm) {
+                Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM, il);
+                cb(Qcur, "Qcur", il);
+            }
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            if (model.layers[il].attn_k_norm) {
+                Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM, il);
+                cb(Kcur, "Kcur", il);
+            }
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur     = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL    = ggml_get_rows(ctx0, inpL, inp_out_ids);
+            ffn_inp = ggml_get_rows(ctx0, ffn_inp, inp_out_ids);
+        }
+        ggml_tensor * attn_out = cur;
+
+        // feed-forward network
+        {
+            cur = build_ffn(ffn_inp,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        // add together residual + FFN + self-attention
+        cur = ggml_add(ctx0, cur, inpL);
+        cur = ggml_add(ctx0, cur, attn_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    if (f_logit_scale) {
+        cur = ggml_scale(ctx0, cur, f_logit_scale);
+    }
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/dbrx.cpp

@@ -0,0 +1,154 @@
+#include "models.h"
+
+void llama_model_dbrx::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+    ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV,     hparams.f_clamp_kqv);
+
+    switch (hparams.n_layer()) {
+        case 40: type = LLM_TYPE_16x12B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_dbrx::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_expert == 0) {
+        throw std::runtime_error("DBRX model cannot have zero experts");
+    }
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.attn_out_norm = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff,   n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff,   n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_dbrx::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_dbrx::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        // MoE branch
+        cur = build_norm(ffn_inp,
+                model.layers[il].attn_out_norm, NULL,
+                LLM_NORM, il);
+        cb(cur, "attn_out_norm", il);
+
+        cur = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(cur, "ffn_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/deci.cpp

@@ -0,0 +1,191 @@
+#include "models.h"
+
+void llama_model_deci::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_7B; break;
+        case 80: type = LLM_TYPE_70B; break;
+        case 162: type = LLM_TYPE_405B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_deci::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+        const int64_t n_embd_k_gqa  = hparams.n_embd_k_gqa(i);
+        const int64_t n_embd_v_gqa  = hparams.n_embd_v_gqa(i);
+        const int64_t n_ff          = hparams.n_ff(i);
+        const int64_t n_head        = hparams.n_head(i);
+        const int64_t n_head_kv     = hparams.n_head_kv(i);
+
+        if (n_head_kv == 0 && n_head > 0) {
+            // linear attention for DeciLMCausalModel
+            layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        }
+        else if (n_head_kv > 0) {
+            layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+            create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+        }
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        if (n_ff > 0) {
+            layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        }
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+        else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        if (n_ff > 0) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        }
+
+        // optional MLP bias
+        layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_deci::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_deci::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale =
+        hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA     = inpL;
+        const int64_t n_head_kv = hparams.n_head_kv(il);
+        const int64_t n_head    = hparams.n_head(il);
+        const int64_t n_ff      = hparams.n_ff(il);
+
+        if (n_head == 0) {
+            // attention-free layer of Llama-3_1-Nemotron-51B
+            cur = inpL;
+        } else {
+            // norm
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+        }
+        if (n_head > 0 && n_head_kv == 0) {
+            // "linear attention" of Llama-3_1-Nemotron-51B
+            cur = build_lora_mm(model.layers[il].wo, cur);
+            cb(cur, "wo", il);
+        } else if (n_head > 0) {
+            // self-attention
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        // FFN-free layer of Llama-3_1-Nemotron-Ultra-253B
+        if (n_ff == 0) {
+            continue;
+        }
+        // modified to support attention-free layer of Llama-3_1-Nemotron-51B
+        ggml_tensor * ffn_inp = cur;
+        if (n_head > 0) {
+            ffn_inp = ggml_add(ctx0, cur, inpSA);
+            cb(ffn_inp, "ffn_inp", il);
+        }
+        // feed-forward network
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
+                model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/deepseek.cpp

@@ -0,0 +1,194 @@
+#include "models.h"
+
+void llama_model_deepseek::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+
+    switch (hparams.n_ff_exp) {
+        case 1408: type = LLM_TYPE_16B; break;
+        case 1792: type = LLM_TYPE_20B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_deepseek::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    // try to load output.weight, if not found, use token_embd (tied embeddings)
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (i < (int) hparams.n_layer_dense_lead) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        } else {
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            // MoE branch
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+            // Shared expert branch
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_deepseek::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_deepseek::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale =
+        hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, false,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // FFN shared expert
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/deepseek2.cpp

@@ -0,0 +1,438 @@
+#include "models.h"
+
+void llama_model_deepseek2::load_arch_hparams(llama_model_loader & ml) {
+    uint32_t n_vocab = 0;
+    ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false);
+
+    // lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B, Kanana-2-30B-A3B
+    const bool is_lite = (hparams.n_layer() == 27 || hparams.n_layer() == 26 || (hparams.n_layer() == 48 && n_vocab == 128256));
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    if (!is_lite) {
+        ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK, hparams.n_lora_q);
+    }
+    ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK,     hparams.n_lora_kv);
+    ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH_MLA,   hparams.n_embd_head_k_mla_impl, false);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_MLA, hparams.n_embd_head_v_mla_impl, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,        hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,       hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,        hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,         hparams.expert_gating_func, false);
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        // for compatibility with existing DeepSeek V2 and V2.5 GGUFs
+        // that have no expert_gating_func model parameter set
+        if ((hparams.n_layer() == 47 || hparams.n_layer() == 48) && n_vocab == 154880) {
+            // GLM 4.7 Lite
+            hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+        } else {
+            hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX;
+        }
+    }
+
+    if (ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, false)) {
+        // [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
+        // cancel the factor from the convert script
+        hparams.rope_yarn_log_mul /= 0.1f;
+    }
+
+    // (optional) temperature tuning - used by mistral-large
+    ml.get_key(LLM_KV_ATTENTION_TEMPERATURE_SCALE,  hparams.f_attn_temp_scale,       false);
+    ml.get_key(LLM_KV_ATTENTION_TEMPERATURE_LENGTH, hparams.n_attn_temp_floor_scale, false); // FIXME why not use temperature_length?
+
+    hparams.f_attn_temp_offset = 0.0f;
+
+    switch (hparams.n_layer()) {
+        case 27: type = LLM_TYPE_16B; break;
+        case 47: type = LLM_TYPE_30B_A3B; break;
+        case 60: type = LLM_TYPE_236B; break;
+        case 61: type = LLM_TYPE_671B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_deepseek2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    const bool is_mla = hparams.is_mla();
+
+    // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+    const int64_t n_embd_head_k_mla = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v_mla = hparams.n_embd_head_v_mla();
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = n_embd_head_k_mla - n_embd_head_qk_rope;
+    GGML_ASSERT(n_embd_head_qk_nope >= 1);
+
+    const int64_t q_lora_rank  = hparams.n_lora_q;
+    const int64_t kv_lora_rank = hparams.n_lora_kv;
+
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    // try to load output.weight, if not found, use token_embd (tied embeddings)
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        if (q_lora_rank > 0) {
+            layer.attn_q_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, 0);
+        }
+
+        layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, 0);
+
+        if (q_lora_rank > 0) {
+            layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, 0);
+            layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k_mla}, 0);
+        } else {
+            layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_head * n_embd_head_k_mla}, 0);
+        }
+
+        layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + n_embd_head_qk_rope}, 0);
+
+        // note: only old legacy GGUF files will have the unsplit wkv_b tensor in
+        if (is_mla) {
+            layer.wk_b = create_tensor(tn(LLM_TENSOR_ATTN_K_B, "weight", i), {n_embd_head_qk_nope, kv_lora_rank, n_head}, 0);
+            layer.wv_b = create_tensor(tn(LLM_TENSOR_ATTN_V_B, "weight", i), {kv_lora_rank, n_embd_head_v_mla, n_head}, 0);
+        } else {
+            layer.wkv_b = create_tensor(tn(LLM_TENSOR_ATTN_KV_B, "weight", i), {kv_lora_rank, n_head * (n_embd_head_qk_nope + n_embd_head_v_mla)}, 0);
+        }
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_embd_head_v_mla, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (i < (int) hparams.n_layer_dense_lead) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        } else {
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            // MoE branch
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+            create_tensor_gate_up_exps(layer, i, n_embd, n_ff_exp, n_expert, 0);
+
+            // Shared expert branch
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_deepseek2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_deepseek2::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    // lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B
+    bool is_ocr = model.arch == LLM_ARCH_DEEPSEEK2OCR;
+
+    const bool is_mla = hparams.is_mla();
+
+    // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+    const int64_t n_embd_head_k = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v = hparams.n_embd_head_v_mla();
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = n_embd_head_k - n_embd_head_qk_rope;
+
+    const uint32_t kv_lora_rank = hparams.n_lora_kv;
+
+    // We have to pre-scale kq_scale and attn_factor to make the YaRN RoPE work correctly.
+    // See https://github.com/ggml-org/llama.cpp/discussions/7416 for detailed explanation.
+    // And also: https://github.com/ggml-org/llama.cpp/pull/17945 [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
+
+    // first cancel the adjustment from llama_hparams::yarn_attn_factor_adjust to get the original attn_factor
+    GGML_ASSERT(ext_factor >= 0.0f);
+    const float attn_factor_org = attn_factor * (1.0f + 0.1f * logf(1.0f / freq_scale));
+
+    // use the original attn_factor to pre-scale the kq_scale
+    const float mscale   = attn_factor_org * (1.0f + 0.1f * hparams.rope_yarn_log_mul * logf(1.0f / freq_scale));
+    const float kq_scale = 1.0f * mscale * mscale / sqrtf(float(n_embd_head_k));
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    // {n_embd, n_tokens}
+    inpL = build_inp_embd(model.tok_embd);
+
+    // (optional) temperature tuning - used by mistral-large
+    ggml_tensor * inp_attn_scale = nullptr;
+    if (hparams.f_attn_temp_scale != 0.0f) {
+        inp_attn_scale = build_inp_attn_scale();
+    }
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn_kv = !is_mla ? build_attn_inp_kv() : nullptr;
+    auto * inp_attn_k  =  is_mla ? build_attn_inp_k()  : nullptr;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        if (is_ocr) {
+            const int n_embed_head = hparams.n_embd / hparams.n_head();
+            const int ocr_rope_type = GGML_ROPE_TYPE_NEOX;
+            GGML_ASSERT(n_embed_head == n_embd_head_k && n_embed_head == n_embd_head_v);
+
+            ggml_tensor * Qcur = NULL;
+            ggml_tensor * Kcur = NULL;
+            ggml_tensor * Vcur = NULL;
+
+            Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
+            Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
+            Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
+            cb(Qcur, "q", il);
+            cb(Kcur, "k", il);
+            cb(Vcur, "v", il);
+
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embed_head, n_head, n_tokens);
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embed_head, n_head, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embed_head, n_head, n_tokens);
+
+            GGML_ASSERT(fabs(freq_base - 10000.0) < 1e-4);
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_embed_head, ocr_rope_type, 0, freq_base, 1, 0, 1, 0, 0);
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_embed_head, ocr_rope_type, 0, freq_base, 1, 0, 1, 0, 0);
+            cb(Qcur, "q_pe", il);
+            cb(Kcur, "k_pe", il);
+
+            cur = build_attn(inp_attn_kv,
+                        model.layers[il].wo, NULL, model.layers[il].wo_s,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+        else {
+            ggml_tensor * q = NULL;
+
+            const bool is_lite = model.layers[il].wq;
+
+            if (!is_lite) {
+                q = ggml_mul_mat(ctx0, model.layers[il].wq_a, cur);
+                cb(q, "q", il);
+
+                q = build_norm(q, model.layers[il].attn_q_a_norm, nullptr, LLM_NORM_RMS, il);
+                cb(q, "q", il);
+
+                q = ggml_mul_mat(ctx0, model.layers[il].wq_b, q);
+                cb(q, "q", il);
+            } else {
+                q = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
+                cb(q, "q", il);
+            }
+            // split into {n_embd_head_qk_nope, n_head, n_tokens}
+            ggml_tensor * q_nope =
+                ggml_view_3d(ctx0, q, n_embd_head_qk_nope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+                             ggml_row_size(q->type, n_embd_head_k) * n_head, 0);
+            cb(q_nope, "q_nope", il);
+
+            // and {n_embd_head_qk_rope, n_head, n_tokens}
+            ggml_tensor * q_pe = ggml_view_3d(
+                ctx0, q, n_embd_head_qk_rope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+                ggml_row_size(q->type, n_embd_head_k) * n_head, ggml_row_size(q->type, n_embd_head_qk_nope));
+            cb(q_pe, "q_pe", il);
+
+            ggml_tensor * kv_cmpr_pe = ggml_mul_mat(ctx0, model.layers[il].wkv_a_mqa, cur);
+            cb(kv_cmpr_pe, "kv_cmpr_pe", il);
+
+            // split into {kv_lora_rank, n_tokens}
+            ggml_tensor * kv_cmpr =
+                ggml_view_2d(ctx0, kv_cmpr_pe, kv_lora_rank, n_tokens,
+                             ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope), 0);
+            cb(kv_cmpr, "kv_cmpr", il);
+
+            // and {n_embd_head_qk_rope, 1, n_tokens}
+            ggml_tensor * k_pe = ggml_view_3d(ctx0, kv_cmpr_pe, n_embd_head_qk_rope, 1, n_tokens,
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank));
+            cb(k_pe, "k_pe", il);
+
+            q_pe = ggml_rope_ext(ctx0, q_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(q_pe, "q_pe", il);
+
+            k_pe = ggml_rope_ext(ctx0, k_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(k_pe, "k_pe", il);
+
+            kv_cmpr = build_norm(kv_cmpr, model.layers[il].attn_kv_a_norm, nullptr, LLM_NORM_RMS, il);
+            cb(kv_cmpr, "kv_cmpr", il);
+
+            if (is_mla) {
+                // {n_embd_head_qk_nope, n_tokens, n_head}
+                q_nope = ggml_permute(ctx0, q_nope, 0, 2, 1, 3);
+                cb(q_nope, "q_nope_perm", il);
+
+                // {n_embd_head_qk_nope, kv_lora_rank, n_head} x {n_embd_head_qk_nope, n_tokens, n_head}
+                ggml_tensor * q_nope_absorbed = ggml_mul_mat(ctx0, model.layers[il].wk_b, q_nope);
+                cb(q_nope_absorbed, "q_nope_absorbed", il);
+
+                // {kv_lora_rank, n_head, n_tokens}
+                q_nope_absorbed = ggml_permute(ctx0, q_nope_absorbed, 0, 2, 1, 3);
+                cb(q_nope_absorbed, "q_nope_absorbed_perm", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, n_head, n_tokens}
+                // note: rope must go first for in-place context shifting in build_rope_shift()
+                ggml_tensor * Qcur = ggml_concat(ctx0, q_nope_absorbed, q_pe, 0);
+                cb(Qcur, "Qcur", il);
+
+                kv_cmpr = ggml_reshape_3d(ctx0, kv_cmpr, kv_lora_rank, 1, n_tokens);
+                cb(kv_cmpr, "kv_cmpr_reshape", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Kcur = ggml_concat(ctx0, kv_cmpr, k_pe, 0);
+                cb(Kcur, "Kcur", il);
+
+                // {kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Vcur = kv_cmpr;
+                cb(Vcur, "Vcur", il);
+
+                if (inp_attn_scale) {
+                    // apply llama 4 temperature scaling
+                    Qcur = ggml_mul(ctx0, Qcur, inp_attn_scale);
+                    cb(Qcur, "Qcur_attn_temp_scaled", il);
+                }
+
+                // note: MLA with the absorption optimization converts into MQA (ie: GQA with 1 group)
+                cur = build_attn(inp_attn_k,
+                        model.layers[il].wo, NULL, model.layers[il].wo_s,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, model.layers[il].wv_b, kq_scale, il);
+            } else {
+                ggml_tensor * kv = ggml_mul_mat(ctx0, model.layers[il].wkv_b, kv_cmpr);
+                cb(kv, "kv", il);
+
+                // split into {n_embd_head_qk_nope, n_head, n_tokens}
+                ggml_tensor * k_nope =
+                    ggml_view_3d(ctx0, kv, n_embd_head_qk_nope, n_head, n_tokens,
+                                 ggml_row_size(kv->type, n_embd_head_qk_nope + n_embd_head_v),
+                                 ggml_row_size(kv->type, n_embd_head_qk_nope + n_embd_head_v) * n_head, 0);
+                cb(k_nope, "k_nope_view", il);
+
+                // and {n_embd_head_v, n_head, n_tokens}
+                ggml_tensor * Vcur = ggml_view_3d(ctx0, kv, n_embd_head_v, n_head, n_tokens,
+                                                  ggml_row_size(kv->type, n_embd_head_qk_nope + n_embd_head_v),
+                                                  ggml_row_size(kv->type, n_embd_head_qk_nope + n_embd_head_v) * n_head,
+                                                  ggml_row_size(kv->type, n_embd_head_qk_nope));
+                cb(Vcur, "Vcur_view", il);
+
+                Vcur = ggml_cont(ctx0, Vcur);
+                cb(Vcur, "Vcur_cont", il);
+
+                ggml_tensor * Qcur = ggml_concat(ctx0, q_nope, q_pe, 0);
+                cb(Qcur, "Qcur", il);
+
+                ggml_tensor * Kcur = ggml_concat(ctx0, k_nope, ggml_repeat(ctx0, k_pe, q_pe), 0);
+                cb(Kcur, "Kcur", il);
+
+                if (inp_attn_scale) {
+                    // apply llama 4 temperature scaling
+                    Qcur = ggml_mul(ctx0, Qcur, inp_attn_scale);
+                    cb(Qcur, "Qcur_attn_temp_scaled", il);
+                }
+
+                // note: MLA without the absorption optimization converts into MHA (ie: GQA with full n_head groups)
+                cur = build_attn(inp_attn_kv,
+                            model.layers[il].wo, NULL, model.layers[il].wo_s,
+                            Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            }
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il,
+                nullptr,
+                model.layers[il].ffn_gate_up_exps);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // FFN shared expert
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = ggml_mul_mat(ctx0, model.output, cur);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/deepseek2ocr.cpp

@@ -0,0 +1,82 @@
+#include "models.h"
+
+void llama_model_deepseek2ocr::load_arch_hparams(llama_model_loader & ml) {
+    // similar to deepseek2, but without MLA
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX;
+    }
+
+    switch (hparams.n_layer()) {
+        case 12: type = LLM_TYPE_3B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_deepseek2ocr::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    // similar to deepseek2, but without MLA
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    // try to load output.weight, if not found, use token_embd (tied embeddings)
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}, 0);
+        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd}, 0);
+        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        // norm
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (i < (int) hparams.n_layer_dense_lead) {
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        } else {
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            // MoE branch
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+            create_tensor_gate_up_exps(layer, i, n_embd, n_ff_exp, n_expert, 0);
+
+            // Shared expert branch
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_deepseek2ocr::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/deepseek32.cpp

@@ -0,0 +1,506 @@
+#include "models.h"
+
+#include "llama-kv-cache.h"
+#include "llama-kv-cache-dsa.h"
+
+void llama_model_deepseek32::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,     hparams.n_ff_exp);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,    hparams.f_norm_rms_eps);
+    hparams.f_norm_eps = 1e-6;  // eps for layer norm
+    ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false);
+
+    // MoE parameters
+    ml.get_key(LLM_KV_EXPERT_COUNT,                hparams.n_expert);
+    ml.get_key(LLM_KV_EXPERT_USED_COUNT,           hparams.n_expert_used);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+
+    // deepseek MLA parameters
+    ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK,      hparams.n_lora_q);
+    ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK,     hparams.n_lora_kv);
+    ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH_MLA,   hparams.n_embd_head_k_mla_impl, false);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_MLA, hparams.n_embd_head_v_mla_impl, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,        hparams.n_expert_shared);
+
+    // DSA parameters
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_HEAD_COUNT, hparams.indexer_n_head);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_KEY_LENGTH, hparams.indexer_head_size);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_TOP_K,      hparams.indexer_top_k);
+
+    // Expert gating function
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func);
+
+    if (ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, 0.0f)) {
+        // [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
+        // cancel the factor from the convert script
+        hparams.rope_yarn_log_mul /= 0.1f;
+    }
+
+    // NextN/MTP parameters
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer");
+
+    switch (hparams.n_layer()) {
+        case 62: type = LLM_TYPE_685B_A37B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_deepseek32::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const bool is_mla = hparams.is_mla();
+    if (!is_mla) {
+        throw std::runtime_error("DEEPSEEK32 architecture requires MLA");
+    }
+
+    // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+    const int64_t n_embd_head_k_mla = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v_mla = hparams.n_embd_head_v_mla();
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = n_embd_head_k_mla - n_embd_head_qk_rope;
+
+    const int64_t q_lora_rank  = hparams.n_lora_q;
+    const int64_t kv_lora_rank = hparams.n_lora_kv;
+
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    // try to load output.weight, if not found, use token_embd (tied embeddings)
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        int flags = 0;
+        if (i >= n_layer) {
+            // skip all tensors in the NextN layers
+            // TODO @ngxson : TENSOR_NOT_REQUIRED was a hack, need to remove it later
+            flags |= TENSOR_SKIP | TENSOR_NOT_REQUIRED;
+        }
+
+        auto & layer = layers[i];
+
+        layer.attn_norm      = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, flags);
+        layer.attn_q_a_norm  = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, flags);
+        layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, flags);
+
+        layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, flags);
+        layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k_mla}, flags);
+
+        layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + n_embd_head_qk_rope}, flags);
+
+        // note: only old legacy GGUF files will have the unsplit wkv_b tensor in
+        layer.wk_b = create_tensor(tn(LLM_TENSOR_ATTN_K_B, "weight", i), {n_embd_head_qk_nope, kv_lora_rank, n_head}, flags);
+        layer.wv_b = create_tensor(tn(LLM_TENSOR_ATTN_V_B, "weight", i), {kv_lora_rank, n_embd_head_v_mla, n_head}, flags);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_embd_head_v_mla, n_embd}, flags);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, flags);
+
+        // DSA indexer
+        layer.indexer_k_norm   = create_tensor(tn(LLM_TENSOR_INDEXER_K_NORM,   "weight", i), {hparams.indexer_head_size}, flags);
+        layer.indexer_k_norm_b = create_tensor(tn(LLM_TENSOR_INDEXER_K_NORM,   "bias",   i), {hparams.indexer_head_size}, flags);
+        layer.indexer_proj     = create_tensor(tn(LLM_TENSOR_INDEXER_PROJ,     "weight", i), {n_embd, hparams.indexer_n_head}, flags);
+        layer.indexer_attn_k   = create_tensor(tn(LLM_TENSOR_INDEXER_ATTN_K,   "weight", i), {n_embd, hparams.indexer_head_size}, flags);
+        layer.indexer_attn_q_b = create_tensor(tn(LLM_TENSOR_INDEXER_ATTN_Q_B, "weight", i), {q_lora_rank, hparams.indexer_n_head * hparams.indexer_head_size}, flags);
+        if (i < (int) hparams.n_layer_dense_lead) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, flags);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, flags);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, flags);
+        } else {
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, flags);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            // MoE branch
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, flags);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, flags);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, flags);
+
+            // Shared expert branch
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, flags);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, flags);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, flags);
+        }
+
+        // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers
+        if (i >= n_layer) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags);
+
+            // Optional tensors
+            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, flags | TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_deepseek32::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_deepseek32::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const bool is_mla = hparams.is_mla();
+    GGML_ASSERT(is_mla);
+
+    // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+    const int64_t n_embd_head_k = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v = hparams.n_embd_head_v_mla();
+    GGML_UNUSED(n_embd_head_v);
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = n_embd_head_k - n_embd_head_qk_rope;
+
+    const int64_t n_indexer_head = hparams.indexer_n_head;
+    const int64_t n_embd_indexer_head = hparams.indexer_head_size;
+    const int64_t n_embd_indexer_head_rope = hparams.n_rot();
+    const int64_t n_embd_indexer_head_nope = n_embd_indexer_head - n_embd_indexer_head_rope;
+    const uint32_t n_indexer_top_k = hparams.indexer_top_k;
+
+    const uint32_t kv_lora_rank = hparams.n_lora_kv;
+
+    // We have to pre-scale kq_scale and attn_factor to make the YaRN RoPE work correctly.
+    // See https://github.com/ggml-org/llama.cpp/discussions/7416 for detailed explanation.
+    // And also: https://github.com/ggml-org/llama.cpp/pull/17945 [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
+
+    // first cancel the adjustment from llama_hparams::yarn_attn_factor_adjust to get the original attn_factor
+    GGML_ASSERT(ext_factor >= 0.0f);
+    const float attn_factor_org = attn_factor * (1.0f + 0.1f * logf(1.0f / freq_scale));
+
+    // use the original attn_factor to pre-scale the kq_scale
+    const float mscale   = attn_factor_org * (1.0f + 0.1f * hparams.rope_yarn_log_mul * logf(1.0f / freq_scale));
+    const float kq_scale = 1.0f * mscale * mscale / sqrtf(float(n_embd_head_k));
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    // {n_embd, n_tokens}
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    llm_graph_input_attn_k_dsa * inp_attn_dsa = build_attn_inp_k_dsa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            ggml_tensor * qr = ggml_mul_mat(ctx0, model.layers[il].wq_a, cur);
+            cb(qr, "qr", il);
+
+            qr = build_norm(qr, model.layers[il].attn_q_a_norm, nullptr, LLM_NORM_RMS, il);
+            cb(qr, "qr", il);
+
+            ggml_tensor * top_k = nullptr;
+
+            // lightning indexer
+            {
+                ggml_tensor * indexer_q = ggml_mul_mat(ctx0, model.layers[il].indexer_attn_q_b, qr);
+                cb(indexer_q, "indexer_q", il);
+
+                // split into {n_embd_indexer_head_rope, n_indexer_head, n_tokens}
+                ggml_tensor * indexer_q_pe =
+                    ggml_view_3d(ctx0, indexer_q, n_embd_indexer_head_rope, n_indexer_head, n_tokens,
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head) * n_indexer_head, 0);
+                cb(indexer_q_pe, "indexer_q_pe", il);
+
+                // and {n_embd_indexer_head_nope, n_indexer_head, n_tokens}
+                ggml_tensor * indexer_q_nope =
+                    ggml_view_3d(ctx0, indexer_q, n_embd_indexer_head_nope, n_indexer_head, n_tokens,
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head) * n_indexer_head,
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head_nope));
+                cb(indexer_q_nope, "indexer_q_nope", il);
+
+                indexer_q_pe = ggml_rope_ext(ctx0, indexer_q_pe, inp_pos, nullptr, n_rot,
+                                     LLAMA_ROPE_TYPE_NEOX, n_ctx_orig, freq_base, freq_scale,
+                                     ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(indexer_q_pe, "indexer_q_pe", il);
+
+                // {n_embd_indexer_head_rope + n_embd_indexer_head_nope, n_head, n_tokens}
+                indexer_q = ggml_concat(ctx0, indexer_q_pe, indexer_q_nope, 0);
+                cb(indexer_q, "indexer_q", il);
+
+                ggml_tensor * indexer_k = ggml_mul_mat(ctx0, model.layers[il].indexer_attn_k, cur);
+                cb(indexer_k, "indexer_k", il);
+
+                indexer_k = build_norm(indexer_k, model.layers[il].indexer_k_norm, model.layers[il].indexer_k_norm_b, LLM_NORM, il);
+                cb(indexer_k, "indexer_k", il);
+
+                // split into {n_embd_indexer_head_rope, 1, n_tokens}
+                ggml_tensor * indexer_k_pe =
+                    ggml_view_3d(ctx0, indexer_k, n_embd_indexer_head_rope, 1, n_tokens,
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head) * 1, 0);
+                cb(indexer_k_pe, "indexer_k_pe", il);
+
+                // and {n_embd_indexer_head_nope, 1, n_tokens}
+                ggml_tensor * indexer_k_nope =
+                    ggml_view_3d(ctx0, indexer_k, n_embd_indexer_head_nope, 1, n_tokens,
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head) * 1,
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head_nope));
+                cb(indexer_k_nope, "indexer_k_nope", il);
+
+                indexer_k_pe = ggml_rope_ext(ctx0, indexer_k_pe, inp_pos, nullptr, n_rot,
+                                     LLAMA_ROPE_TYPE_NEOX, n_ctx_orig, freq_base, freq_scale,
+                                     ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(indexer_k_pe, "indexer_k_pe", il);
+
+                // {n_embd_indexer_head_rope + n_embd_indexer_head_nope, 1, n_tokens}
+                indexer_k = ggml_concat(ctx0, indexer_k_pe, indexer_k_nope, 0);
+                cb(indexer_k, "indexer_k", il);
+
+                // perform Hadamard transform on indexer q and k
+                indexer_q = ggml_mul_mat(ctx0, inp_attn_dsa->self_k_rot_lid, indexer_q);
+                cb(indexer_q, "indexer_q", il);
+                indexer_k = ggml_mul_mat(ctx0, inp_attn_dsa->self_k_rot_lid, indexer_k);
+                cb(indexer_k, "indexer_k", il);
+
+                // store indexer keys to KV cache
+                const auto * mctx_lid = inp_attn_dsa->mctx->get_lid();
+                const auto & k_idxs_lid = inp_attn_dsa->get_k_idxs_lid();
+                ggml_build_forward_expand(gf, mctx_lid->cpy_k(ctx0, indexer_k, k_idxs_lid, il));
+
+                // prepare indexer weights
+                ggml_tensor * indexer_weights = ggml_mul_mat(ctx0, model.layers[il].indexer_proj, cur);
+                cb(indexer_weights, "indexer_weights", il);
+
+                // get cached indexer keys
+                indexer_k = mctx_lid->get_k(ctx0, il);
+
+                // split the batch into streams if needed
+                const auto n_stream = indexer_k->ne[3];
+                indexer_q = ggml_view_4d(ctx0, indexer_q, indexer_q->ne[0], indexer_q->ne[1], indexer_q->ne[2]/n_stream, n_stream, indexer_q->nb[1], indexer_q->nb[2], indexer_q->nb[3]/n_stream, 0);
+                indexer_weights = ggml_view_4d(ctx0, indexer_weights, indexer_weights->ne[0], indexer_weights->ne[1]/n_stream, indexer_weights->ne[2], n_stream, indexer_weights->nb[1], indexer_weights->nb[2]/n_stream, indexer_weights->nb[3]/n_stream, 0);
+
+                // pre-scale weights to avoid scaling operations on huge indexer_score tensor
+                indexer_weights = ggml_scale(ctx0, indexer_weights, 1.0f / sqrtf(float(n_embd_indexer_head * n_indexer_head)));
+                cb(indexer_weights, "indexer_weights", il);
+
+                ggml_tensor * indexer_score = nullptr;
+                if (cparams.fused_lid) {
+                    indexer_score = ggml_lightning_indexer(ctx0, indexer_q, indexer_k, indexer_weights, inp_attn_dsa->get_kq_mask_lid());
+                    cb(indexer_score, "indexer_score", il);
+                    res->add_fused_node({LLM_FUSED_OP_LIGHTNING_INDEXER, indexer_score, il});
+                } else {
+                    // calculate indexer kq
+                    indexer_q = ggml_permute(ctx0, indexer_q, 0, 2, 1, 3);
+                    cb(indexer_q, "indexer_q", il);
+                    indexer_k = ggml_permute(ctx0, indexer_k, 0, 2, 1, 3);
+                    cb(indexer_k, "indexer_k", il);
+
+                    ggml_tensor * indexer_kq = ggml_mul_mat(ctx0, indexer_k, indexer_q);
+                    cb(indexer_kq, "indexer_kq", il);
+
+                    // ReLU requires contiguous tensors
+                    indexer_kq = ggml_cont(ctx0, ggml_permute(ctx0, indexer_kq, 2, 1, 0, 3));
+                    cb(indexer_kq, "indexer_kq", il);
+
+                    // apply ReLU
+                    indexer_score = ggml_relu(ctx0, indexer_kq);
+                    cb(indexer_score, "indexer_score", il);
+
+                    // multiply scores by indexer weights
+                    indexer_score = ggml_mul(ctx0, indexer_score, indexer_weights);
+                    cb(indexer_score, "indexer_score", il);
+
+                    // sum by q n_indexer_head dimension
+                    indexer_score = ggml_sum_rows(ctx0, indexer_score);
+                    cb(indexer_score, "indexer_score", il);
+
+                    // permute result to match KQ mask
+                    indexer_score = ggml_cont(ctx0, ggml_permute(ctx0, indexer_score, 2, 1, 0, 3));
+                    cb(indexer_score, "indexer_score", il);
+
+                    // mask indexer scores
+                    ggml_tensor * indexer_kq_mask = inp_attn_dsa->get_kq_mask_lid();
+                    indexer_score = ggml_add(ctx0, indexer_score, indexer_kq_mask);
+                    cb(indexer_score, "indexer_score", il);
+                }
+
+                // get indices of top k indexer scores
+                uint32_t n_top_k = indexer_score->ne[0] < n_indexer_top_k ? indexer_score->ne[0] : n_indexer_top_k;
+                top_k = ggml_cont(ctx0, ggml_top_k(ctx0, indexer_score, n_top_k));
+                cb(top_k, "top_k", il);
+            }
+
+            ggml_tensor * q = ggml_mul_mat(ctx0, model.layers[il].wq_b, qr);
+            cb(q, "q", il);
+
+            // split into {n_embd_head_qk_nope, n_head, n_tokens}
+            ggml_tensor * q_nope =
+                ggml_view_3d(ctx0, q, n_embd_head_qk_nope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+                             ggml_row_size(q->type, n_embd_head_k) * n_head, 0);
+            cb(q_nope, "q_nope", il);
+
+            // and {n_embd_head_qk_rope, n_head, n_tokens}
+            ggml_tensor * q_pe = ggml_view_3d(
+                ctx0, q, n_embd_head_qk_rope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+                ggml_row_size(q->type, n_embd_head_k) * n_head, ggml_row_size(q->type, n_embd_head_qk_nope));
+            cb(q_pe, "q_pe", il);
+
+            ggml_tensor * kv_cmpr_pe = ggml_mul_mat(ctx0, model.layers[il].wkv_a_mqa, cur);
+            cb(kv_cmpr_pe, "kv_cmpr_pe", il);
+
+            // split into {kv_lora_rank, n_tokens}
+            ggml_tensor * kv_cmpr =
+                ggml_view_2d(ctx0, kv_cmpr_pe, kv_lora_rank, n_tokens,
+                             ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope), 0);
+            cb(kv_cmpr, "kv_cmpr", il);
+
+            // and {n_embd_head_qk_rope, 1, n_tokens}
+            ggml_tensor * k_pe = ggml_view_3d(ctx0, kv_cmpr_pe, n_embd_head_qk_rope, 1, n_tokens,
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank));
+            cb(k_pe, "k_pe", il);
+
+            q_pe = ggml_rope_ext(ctx0, q_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(q_pe, "q_pe", il);
+
+            k_pe = ggml_rope_ext(ctx0, k_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(k_pe, "k_pe", il);
+
+            kv_cmpr = build_norm(kv_cmpr, model.layers[il].attn_kv_a_norm, nullptr, LLM_NORM_RMS, il);
+            cb(kv_cmpr, "kv_cmpr", il);
+
+            // MLA attention
+            {
+                // {n_embd_head_qk_nope, n_tokens, n_head}
+                q_nope = ggml_permute(ctx0, q_nope, 0, 2, 1, 3);
+                cb(q_nope, "q_nope_perm", il);
+
+                // {n_embd_head_qk_nope, kv_lora_rank, n_head} x {n_embd_head_qk_nope, n_tokens, n_head}
+                ggml_tensor * q_nope_absorbed = ggml_mul_mat(ctx0, model.layers[il].wk_b, q_nope);
+                cb(q_nope_absorbed, "q_nope_absorbed", il);
+
+                // {kv_lora_rank, n_head, n_tokens}
+                q_nope_absorbed = ggml_permute(ctx0, q_nope_absorbed, 0, 2, 1, 3);
+                cb(q_nope_absorbed, "q_nope_absorbed_perm", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, n_head, n_tokens}
+                // note: rope must go first for in-place context shifting in build_rope_shift()
+                ggml_tensor * Qcur = ggml_concat(ctx0, q_nope_absorbed, q_pe, 0);
+                cb(Qcur, "Qcur", il);
+
+                kv_cmpr = ggml_reshape_3d(ctx0, kv_cmpr, kv_lora_rank, 1, n_tokens);
+                cb(kv_cmpr, "kv_cmpr_reshape", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Kcur = ggml_concat(ctx0, kv_cmpr, k_pe, 0);
+                cb(Kcur, "Kcur", il);
+
+                // {kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Vcur = kv_cmpr;
+                cb(Vcur, "Vcur", il);
+
+                // note: MLA with the absorption optimization converts into MQA (ie: GQA with 1 group)
+                cur = build_attn(inp_attn_dsa,
+                        model.layers[il].wo, NULL, model.layers[il].wo_s,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, model.layers[il].wv_b, top_k, kq_scale, il);
+            }
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, model.layers[il].ffn_up_s,
+                model.layers[il].ffn_gate, NULL, model.layers[il].ffn_gate_s,
+                model.layers[il].ffn_down, NULL, model.layers[il].ffn_down_s,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il,
+                nullptr,
+                model.layers[il].ffn_gate_up_exps,
+                model.layers[il].ffn_up_exps_s,
+                model.layers[il].ffn_gate_exps_s,
+                model.layers[il].ffn_down_exps_s);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // FFN shared expert
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, model.layers[il].ffn_up_shexp_s,
+                        model.layers[il].ffn_gate_shexp, NULL, model.layers[il].ffn_gate_shexp_s,
+                        model.layers[il].ffn_down_shexp, NULL, model.layers[il].ffn_down_shexp_s,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = ggml_mul_mat(ctx0, model.output, cur);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/delta-net-base.cpp

@@ -0,0 +1,606 @@
+#include "models.h"
+
+#include "llama-impl.h"
+#include "llama-memory-recurrent.h"
+
+// utility to get one slice from the third dimension
+// input dim:  [x, y, c, b]
+// output dim: [x, y, 1, b]
+static ggml_tensor * get_slice_2d(ggml_context * ctx0, ggml_tensor * t, int64_t c) {
+    return ggml_view_4d(ctx0, t, t->ne[0], t->ne[1], 1, t->ne[3],
+        t->nb[1], t->nb[2], t->nb[3], t->nb[2] * c);
+}
+
+llm_build_delta_net_base::llm_build_delta_net_base(const llm_graph_params & params) : llm_graph_context(params) {}
+
+std::pair<ggml_tensor *, ggml_tensor *> llm_build_delta_net_base::build_delta_net_chunking(
+        ggml_tensor * q,
+        ggml_tensor * k,
+        ggml_tensor * v,
+        ggml_tensor * g,
+        ggml_tensor * b,
+        ggml_tensor * s,
+        int           il) {
+    const int64_t S_k      = q->ne[0];
+    const int64_t H_k      = q->ne[1];
+    const int64_t n_tokens = q->ne[2];
+    const int64_t n_seqs   = q->ne[3];
+
+    const int64_t S_v = v->ne[0];
+    const int64_t H_v = v->ne[1];
+    const bool kda = (g->ne[0] == S_k && g->ne[1] == H_k);
+
+    GGML_ASSERT(S_k == S_v);
+    GGML_ASSERT(H_v % H_k == 0);
+
+    GGML_ASSERT(q->ne[0] == S_k && q->ne[1] == H_k && q->ne[2] == n_tokens && q->ne[3] == n_seqs);
+    GGML_ASSERT(k->ne[0] == S_k && k->ne[1] == H_k && k->ne[2] == n_tokens && k->ne[3] == n_seqs);
+    GGML_ASSERT(v->ne[0] == S_v && v->ne[1] == H_v && v->ne[2] == n_tokens && v->ne[3] == n_seqs);
+
+    GGML_ASSERT(g->ne[0] == 1   || g->ne[0] == S_v);
+    GGML_ASSERT(                   g->ne[1] == H_v && g->ne[2] == n_tokens && g->ne[3] == n_seqs);
+    GGML_ASSERT(b->ne[0] == 1   && b->ne[1] == H_v && b->ne[2] == n_tokens && b->ne[3] == n_seqs);
+    GGML_ASSERT(s->ne[0] == S_v && s->ne[1] == S_v && s->ne[2] == H_v      && s->ne[3] == n_seqs);
+
+    const float scale = 1.0f / sqrtf(S_k);
+
+    q = ggml_scale(ctx0, q, scale);
+
+    cb(q, "q_in", il);
+    cb(k, "k_in", il);
+    cb(v, "v_in", il);
+    cb(b, "b_in", il);
+    cb(g, "g_in", il);
+
+    q = ggml_permute(ctx0, q, 0, 2, 1, 3); // [S_k, n_tokens, H_k, n_seqs]
+    k = ggml_permute(ctx0, k, 0, 2, 1, 3); // [S_k, n_tokens, H_k, n_seqs]
+    v = ggml_permute(ctx0, v, 0, 2, 1, 3); // [S_v, n_tokens, H_v, n_seqs]
+    g = ggml_permute(ctx0, g, 0, 2, 1, 3); // [g_0, n_tokens, H_v, n_seqs]
+    b = ggml_permute(ctx0, b, 0, 2, 1, 3); // [  1, n_tokens, H_v, n_seqs]
+
+    const int CS = kda ? 16 : 64; // chunk size
+
+    const int pad = (CS - n_tokens % CS) % CS;
+    const int n_chunks = (n_tokens + pad) / CS;
+
+    q = ggml_pad(ctx0, q, 0, pad, 0, 0);
+    k = ggml_pad(ctx0, k, 0, pad, 0, 0);
+    v = ggml_pad(ctx0, v, 0, pad, 0, 0);
+    g = ggml_pad(ctx0, g, 0, pad, 0, 0);
+    b = ggml_pad(ctx0, b, 0, pad, 0, 0);
+
+    ggml_tensor * v_b = ggml_mul(ctx0, v, b);
+    ggml_tensor * k_b = ggml_mul(ctx0, k, b);
+
+    cb(v_b, "v_b", il);
+    cb(k_b, "k_b", il);
+
+    q   = ggml_reshape_4d(ctx0, q,   S_k, CS, n_chunks, H_k * n_seqs);
+    k   = ggml_reshape_4d(ctx0, k,   S_k, CS, n_chunks, H_k * n_seqs);
+    k_b = ggml_reshape_4d(ctx0, k_b, S_k, CS, n_chunks, H_v * n_seqs);
+    v   = ggml_reshape_4d(ctx0, v,   S_v, CS, n_chunks, H_v * n_seqs);
+    v_b = ggml_reshape_4d(ctx0, v_b, S_v, CS, n_chunks, H_v * n_seqs);
+
+    g = ggml_reshape_4d(ctx0, g, g->ne[0], CS, n_chunks, H_v * n_seqs);
+    b = ggml_reshape_4d(ctx0, b, 1,        CS, n_chunks, H_v * n_seqs);
+
+    // [CS, g_0, n_chunks, H_v * n_seqs]
+    // TODO: extend ggml_cumsum with axis parameter to avoid transpose
+    ggml_tensor * g_cs = ggml_cumsum(ctx0, ggml_cont(ctx0, ggml_transpose(ctx0, g)));
+    cb(g_cs, "g_cs", il);
+
+    ggml_tensor * kb = nullptr;
+    ggml_tensor * kq = nullptr;
+    if (kda) {
+        const int64_t CHB = n_chunks * H_k * n_seqs;
+
+        ggml_tensor * g_cs_i = ggml_reshape_4d(ctx0, g_cs, CS, 1, S_k, CHB);  // [chunk_size, 1, S_k, CHB]
+        ggml_tensor * g_cs_j = ggml_reshape_4d(ctx0, g_cs, 1, CS, S_k, CHB);  // [1, chunk_size, S_k, CHB]
+
+        g_cs_j = ggml_repeat_4d(ctx0, g_cs_j, CS, CS, S_k, CHB);  // [1, chunk_size, S_k, CHB] -> [chunk_size, chunk_size, S_k, CHB]
+
+        // decay_mask [chunk_size,chunk_size,S_k,CHB]
+        ggml_tensor * decay_mask;
+        decay_mask = ggml_sub(ctx0, g_cs_j, g_cs_i);
+        decay_mask = ggml_tri(ctx0, decay_mask, GGML_TRI_TYPE_LOWER_DIAG);
+        decay_mask = ggml_exp(ctx0, decay_mask);
+        cb(decay_mask, "decay_mask", il);
+
+        // decay_mask [S_k,BT_j,BT_i,CHB] *Note* second and third chunk_sizes are switched
+        decay_mask = ggml_cont_4d(ctx0, ggml_permute(ctx0, decay_mask, 2, 1, 0, 3), S_k, CS, CS, CHB);
+
+        ggml_tensor * k_b_i = ggml_reshape_4d(ctx0, k_b, S_k, CS,  1, CHB);
+        ggml_tensor * k_j   = ggml_reshape_4d(ctx0, k,   S_k,  1, CS, CHB);
+        ggml_tensor * q_i   = ggml_reshape_4d(ctx0, q,   S_k, CS,  1, CHB);
+
+        ggml_tensor * decay_k_b_i = ggml_mul(ctx0, decay_mask, k_b_i);
+        ggml_tensor * decay_q_i   = ggml_mul(ctx0, decay_mask, q_i);
+
+        // decay_k_b_i [S,BT,BT,CHB] @ k_j [S,1,BT,CHB] = Akk [BT,1,BT,CHB]
+        kb = ggml_mul_mat(ctx0, decay_k_b_i, k_j);
+        kq = ggml_mul_mat(ctx0, decay_q_i,   k_j);
+
+        kb = ggml_cont(ctx0, ggml_transpose(ctx0, ggml_reshape_4d(ctx0, kb, CS, CS, n_chunks, H_v * n_seqs)));
+        kq = ggml_cont(ctx0, ggml_transpose(ctx0, ggml_reshape_4d(ctx0, kq, CS, CS, n_chunks, H_v * n_seqs)));
+    } else {
+        ggml_tensor * g_cs_i = g_cs;
+        ggml_tensor * g_cs_j = ggml_reshape_4d(ctx0, g_cs, 1, CS, n_chunks, H_v * n_seqs);
+
+        g_cs_j = ggml_repeat_4d(ctx0, g_cs_j, CS, CS, n_chunks, H_v * n_seqs);
+
+        // [CS, CS, n_chunks, H_v * n_seqs]
+        ggml_tensor * decay_mask;
+        decay_mask = ggml_sub(ctx0, g_cs_j, g_cs_i);
+        decay_mask = ggml_tri(ctx0, decay_mask, GGML_TRI_TYPE_LOWER_DIAG);
+        decay_mask = ggml_exp(ctx0, decay_mask);
+        cb(decay_mask, "decay_mask", il);
+
+        // [CS, CS, n_chunks, H_k * n_seqs]
+        kb = ggml_mul_mat(ctx0, k,  k_b);
+        kb = ggml_mul    (ctx0, kb, decay_mask);
+
+        // [CS, CS, n_chunks, H_k * n_seqs]
+        kq = ggml_mul_mat(ctx0, k, q);
+        kq = ggml_mul(ctx0, kq, decay_mask);
+    }
+
+    kq = ggml_tri(ctx0, kq, GGML_TRI_TYPE_LOWER_DIAG);
+    cb(kq, "kq", il);
+
+    // [CS, CS, n_chunks, H_k * n_seqs]
+    ggml_tensor * attn;
+    attn = ggml_tri(ctx0, kb, GGML_TRI_TYPE_LOWER);
+    cb(attn, "attn", il);
+
+    ggml_tensor * identity;
+    identity = ggml_view_1d(ctx0, attn, CS, 0);
+    identity = ggml_fill   (ctx0, identity, 1.0f);
+    identity = ggml_diag   (ctx0, identity);
+
+    ggml_tensor * lhs = ggml_add(ctx0, attn, identity);
+    cb(lhs, "dnet_add_ch_lhs", il);
+
+    attn = ggml_neg(ctx0, attn);
+    cb(attn, "attn_pre_solve", il);
+
+    ggml_tensor * lin_solve = ggml_solve_tri(ctx0, lhs, attn, true, true, false);
+    attn = ggml_add(ctx0, lin_solve, identity);
+    cb(attn, "dnet_add_ch_attn_solved", il); // [CS, CS, n_chunks, H_k * n_seqs]
+
+    // [S_v, CS, n_chunks, H_v * n_seqs]
+    v = ggml_mul_mat(ctx0, ggml_cont(ctx0, ggml_transpose(ctx0, v_b)), attn);
+
+    // [CS, 1, n_chunks, H_v * n_seqs] KDA: [CS, S_k, n_chunks, H_v * n_seqs]
+    ggml_tensor * g_exp = ggml_exp(ctx0, g_cs);
+
+    k_b = ggml_cont(ctx0, ggml_transpose(ctx0, k_b));
+
+    // [CS, S_k, n_chunks, H_k * n_seqs]
+    ggml_tensor * kbg = ggml_mul(ctx0, k_b, g_exp);
+    cb(kbg, "k_beta_g_exp", il);
+
+    // [S_k, CS, n_chunks, H_k * n_seqs]
+    ggml_tensor * k_cd = ggml_mul_mat(ctx0, kbg, attn);
+    cb(k_cd, "k_cumdecay", il);
+
+    // [1, CS, n_chunks, H_k * n_seqs] KDA: [S_k, CS, n_chunks, H_k * n_seqs]
+    ggml_tensor * g_exp_t = ggml_cont(ctx0, ggml_transpose(ctx0, g_exp));
+    ggml_tensor * q_g_exp = ggml_mul(ctx0, q, g_exp_t);
+
+    // vectorized calculation of key_gdiff
+    // improved from the chunked version:
+    //   g_last = torch.clamp(g_cum[:, :, -1], max=50.0).exp().unsqueeze(-1).unsqueeze(-1)
+    //   g_diff = torch.clamp(g_cum[:, :, -1:] - g_cum, max=50.0).exp()
+    //   key_gdiff = key * g_diff.unsqueeze(-1)
+    //   kgdmulvnew = (key_gdiff).transpose(-1, -2) @ v_new
+    //   last_recurrent_state = last_recurrent_state * g_last + kgdmulvnew
+
+    // get last element in g_cumsum along CS dimension (ne0)
+    // example: [[x, y, z, ..., last], ...] -> [[last], ...]
+    // [1, 1, n_chunks, H_v * n_seqs] KDA: [1, S_k, n_chunks, H_v * n_seqs]
+    ggml_tensor * g_last = ggml_view_4d(ctx0, g_cs, 1, g_cs->ne[1], g_cs->ne[2], g_cs->ne[3],
+            g_cs->nb[1],
+            g_cs->nb[2],
+            g_cs->nb[3],
+            ggml_row_size(g_cs->type, g_cs->ne[0] - 1));
+    cb(g_last, "g_last", il);
+
+    // TODO: remove this cont when CUDA supports non-cont unary ops
+    g_last = ggml_cont(ctx0, g_last);
+
+    // [1, 1, n_chunks, H_v * n_seqs] KDA: [S_k, 1, n_chunks, H_v * n_seqs]
+    ggml_tensor * g_last_exp_t = ggml_transpose(ctx0, ggml_exp(ctx0, g_last));
+    cb(g_last_exp_t, "g_last_exp_t", il);
+
+    // [CS, 1, n_chunks, H_v * n_seqs] KDA: [CS, S_k, n_chunks, H_v * n_seqs]
+    ggml_tensor * g_diff = ggml_neg(ctx0, ggml_sub(ctx0, g_cs, g_last));
+    cb(g_diff, "g_diff", il);
+
+    ggml_tensor * g_diff_exp_t = ggml_cont(ctx0, ggml_transpose(ctx0, ggml_exp(ctx0, g_diff)));
+
+    // [S_k, CS, n_chunks, H_v * n_seqs]
+    ggml_tensor * kg = ggml_mul(ctx0, k, g_diff_exp_t);
+    cb(kg, "key_gdiff", il);
+
+    // [CS, S_k, n_chunks, H_v * n_seqs]
+    ggml_tensor * kg_t = ggml_cont(ctx0, ggml_transpose(ctx0, kg));
+    cb(kg_t, "key_gdiff_t", il);
+
+    s = ggml_reshape_4d(ctx0, s, S_v, S_v, 1, H_v * n_seqs);
+    cb(s, "dnet_add_ch_state", il);
+
+    // [CS, S_v, n_chunks, H_v * n_seqs]
+    ggml_tensor * v_t = ggml_cont(ctx0, ggml_transpose(ctx0, v));
+
+    for (int64_t chunk = 0; chunk < n_chunks; chunk++) {
+        ggml_tensor * ch_k_cd    = get_slice_2d(ctx0, k_cd,    chunk); // [S_k,  CS, 1, H_k * n_seqs]
+        ggml_tensor * ch_v_t     = get_slice_2d(ctx0, v_t,     chunk); // [ CS, S_v, 1, H_v * n_seqs]
+        ggml_tensor * ch_kq      = get_slice_2d(ctx0, kq,      chunk); // [ CS,  CS, 1, H_k * n_seqs]
+        ggml_tensor * ch_q_g_exp = get_slice_2d(ctx0, q_g_exp, chunk); // [S_k,  CS, 1, H_k * n_seqs]
+        ggml_tensor * ch_kg_t    = get_slice_2d(ctx0, kg_t,    chunk); // [ CS, S_k, 1, H_v * n_seqs]
+
+        // [CS, S_v, 1, H_v * n_seqs]
+        ggml_tensor * v_t_p = ggml_mul_mat(ctx0, ch_k_cd, s);
+        cb(v_t_p, "v_prime", il);
+
+        // [CS, S_v, 1, H_v * n_seqs]
+        ggml_tensor * v_t_new = ggml_sub(ctx0, ch_v_t, v_t_p);
+        cb(v_t_new, "v_t_new", il);
+
+        // [S_v, CS, 1, H_v * n_seqs]
+        ggml_tensor * v_attn = ggml_mul_mat(ctx0, v_t_new, ch_kq);
+        cb(v_attn, "v_attn", il);
+
+        // [S_v, CS, 1, H_v * n_seqs]
+        ggml_tensor * attn_inter = ggml_mul_mat(ctx0, s, ch_q_g_exp);
+        cb(attn_inter, "attn_inter", il);
+
+        // [S_v, CS, 1, H_v * n_seqs]
+        ggml_tensor * o_ch = ggml_add(ctx0, attn_inter, v_attn);
+        cb(o_ch, "dnet_add_ch_attn_out", il);
+
+        v = ggml_set_inplace(ctx0, v, o_ch, v->nb[1], v->nb[2], v->nb[3], chunk * v->nb[2]);
+
+        // kgdmulvnew = (key_gdiff).transpose(-1, -2) @ v_new
+        // TODO: head broadcast might not work here - probably will need a transpose
+        ggml_tensor * kgv = ggml_mul_mat(ctx0, ch_kg_t, v_t_new); // [S_k, S_v, 1, H_k * n_seqs]
+
+        // last_recurrent_state = last_recurrent_state * g_last + kgdmulvnew
+        ggml_tensor * ch_g_last_exp_t = get_slice_2d(ctx0, g_last_exp_t, chunk);
+
+        s = ggml_mul(ctx0, s, ch_g_last_exp_t);
+        s = ggml_add(ctx0, s, kgv);
+        cb(s, "dnet_add_ch_state", il);
+    }
+
+    // truncate padded tokens
+    ggml_tensor * o = ggml_view_4d(ctx0, v,
+            S_v, n_tokens, H_v, n_seqs,
+            ggml_row_size(v->type, S_v),
+            ggml_row_size(v->type, S_v * CS * n_chunks),
+            ggml_row_size(v->type, S_v * CS * n_chunks * H_v), 0);
+    o = ggml_permute  (ctx0, o, 0, 2, 1, 3); // [S_v, H_v, n_tokens, n_seqs]
+    s = ggml_reshape_4d(ctx0, s, S_v, S_v, H_v, n_seqs);
+    cb(s, "output_state", il);
+
+    return {o, s};
+}
+
+std::pair<ggml_tensor *, ggml_tensor *> llm_build_delta_net_base::build_delta_net_autoregressive(
+        ggml_tensor * q,
+        ggml_tensor * k,
+        ggml_tensor * v,
+        ggml_tensor * g,
+        ggml_tensor * b, // beta
+        ggml_tensor * s, // state
+        int           il) {
+    const int64_t S_k      = q->ne[0];
+    const int64_t H_k      = q->ne[1];
+    const int64_t n_tokens = q->ne[2];
+    const int64_t n_seqs   = q->ne[3];
+
+    const int64_t S_v = v->ne[0];
+    const int64_t H_v = v->ne[1];
+
+    GGML_ASSERT(n_tokens == 1);
+
+    GGML_ASSERT(S_k == S_v);
+    GGML_ASSERT(H_v % H_k == 0);
+
+    GGML_ASSERT(q->ne[0] == S_k && q->ne[1] == H_k && q->ne[2] == n_tokens && q->ne[3] == n_seqs);
+    GGML_ASSERT(k->ne[0] == S_k && k->ne[1] == H_k && k->ne[2] == n_tokens && k->ne[3] == n_seqs);
+    GGML_ASSERT(v->ne[0] == S_v && v->ne[1] == H_v && v->ne[2] == n_tokens && v->ne[3] == n_seqs);
+
+    GGML_ASSERT(g->ne[0] == 1   || g->ne[0] == S_v);
+    GGML_ASSERT(                   g->ne[1] == H_v && g->ne[2] == n_tokens && g->ne[3] == n_seqs);
+    GGML_ASSERT(b->ne[0] == 1   && b->ne[1] == H_v && b->ne[2] == n_tokens && b->ne[3] == n_seqs);
+    GGML_ASSERT(s->ne[0] == S_v && s->ne[1] == S_v && s->ne[2] == H_v      && s->ne[3] == n_seqs);
+
+    const float scale = 1.0f / sqrtf(S_k);
+
+    q = ggml_scale(ctx0, q, scale);
+
+    q = ggml_permute(ctx0, q, 0, 2, 1, 3); // [S_k, n_tokens, H_k, n_seqs]
+    k = ggml_permute(ctx0, k, 0, 2, 1, 3); // [S_k, n_tokens, H_k, n_seqs]
+    v = ggml_permute(ctx0, v, 0, 2, 1, 3); // [S_v, n_tokens, H_v, n_seqs]
+
+    cb(q, "q_in", il);
+    cb(k, "k_in", il);
+    cb(v, "v_in", il);
+    cb(b, "b_in", il);
+    cb(g, "g_in", il);
+
+    // GDA: [1,  1,  H_v, n_seqs]
+    // KDA: [1, S_k, H_v, n_seqs]
+    g = ggml_reshape_4d(ctx0, g, 1, g->ne[0], H_v, n_seqs);
+    b = ggml_reshape_4d(ctx0, b, 1,        1, H_v, n_seqs);
+
+    // [S_v, S_v, H_v, n_seqs]
+    g = ggml_exp(ctx0, g);
+    s = ggml_mul(ctx0, s, g);
+
+    // [1, S_v, H_v, n_seqs]
+    ggml_tensor * sk;
+    sk = ggml_mul     (ctx0, s, k);
+    sk = ggml_sum_rows(ctx0, sk);
+
+    // [S_v, 1, H_v, n_seqs]
+    ggml_tensor * d;
+    d = ggml_sub(ctx0, v, ggml_transpose(ctx0, sk));
+    d = ggml_mul(ctx0, d, b);
+
+    // [1, S_v, H_v, n_seqs]
+    ggml_tensor * d_t;
+    d_t = ggml_transpose(ctx0, d);
+
+    // [S_v, S_v, H_v, n_seqs]
+    ggml_tensor * kd;
+    k  = ggml_repeat(ctx0, k, s);
+    kd = ggml_mul   (ctx0, k, d_t);
+
+    s = ggml_add(ctx0, s, kd);
+
+    cb(s, "dnet_add_ar_state", il);
+
+    ggml_tensor * s_q = ggml_mul     (ctx0, s, q);
+    ggml_tensor * o   = ggml_sum_rows(ctx0, s_q);
+
+    o = ggml_permute  (ctx0, o, 2, 0, 1, 3); // [S_v, H_v, n_tokens, n_seqs]
+
+    return {o, s};
+}
+
+std::pair<ggml_tensor *, ggml_tensor *> llm_build_delta_net_base::build_delta_net_fused(
+        ggml_tensor * q,
+        ggml_tensor * k,
+        ggml_tensor * v,
+        ggml_tensor * g,
+        ggml_tensor * b,
+        ggml_tensor * s,
+        int           il) {
+    const int64_t S_k      = q->ne[0];
+    const int64_t H_k      = q->ne[1];
+    const int64_t n_tokens = q->ne[2];
+    const int64_t n_seqs   = q->ne[3];
+
+    const int64_t S_v = v->ne[0];
+    const int64_t H_v = v->ne[1];
+
+    GGML_ASSERT(S_k == S_v);
+    GGML_ASSERT(H_v % H_k == 0);
+
+    GGML_ASSERT(q->ne[0] == S_k && q->ne[1] == H_k && q->ne[2] == n_tokens && q->ne[3] == n_seqs);
+    GGML_ASSERT(k->ne[0] == S_k && k->ne[1] == H_k && k->ne[2] == n_tokens && k->ne[3] == n_seqs);
+    GGML_ASSERT(v->ne[0] == S_v && v->ne[1] == H_v && v->ne[2] == n_tokens && v->ne[3] == n_seqs);
+
+    GGML_ASSERT(g->ne[0] == 1   || g->ne[0] == S_v);
+    GGML_ASSERT(                   g->ne[1] == H_v && g->ne[2] == n_tokens && g->ne[3] == n_seqs);
+    GGML_ASSERT(b->ne[0] == 1   && b->ne[1] == H_v && b->ne[2] == n_tokens && b->ne[3] == n_seqs);
+    GGML_ASSERT(s->ne[0] == S_v && s->ne[1] == S_v && s->ne[2] == H_v      && s->ne[3] == n_seqs);
+
+    // K=1: output carries the final state only. state s is 4D [S_v, S_v, H_v, n_seqs].
+    ggml_tensor * result = ggml_gated_delta_net(ctx0, q, k, v, g, b, s, /*K=*/1);
+    if (n_tokens == 1) {
+        res->add_fused_node({LLM_FUSED_OP_GDN_AR, result, il});
+    } else {
+        res->add_fused_node({LLM_FUSED_OP_GDN_CH, result, il});
+    }
+
+    ggml_tensor * output = ggml_view_4d(ctx0, result,
+            S_v, H_v, n_tokens, n_seqs,
+            ggml_row_size(result->type, S_v),
+            ggml_row_size(result->type, S_v * H_v),
+            ggml_row_size(result->type, S_v * H_v * n_tokens), 0);
+
+    ggml_tensor * new_state = ggml_view_4d(ctx0, result,
+            S_v, S_v, H_v, n_seqs,
+            ggml_row_size(result->type, S_v),
+            ggml_row_size(result->type, S_v * S_v),
+            ggml_row_size(result->type, S_v * S_v * H_v),
+            ggml_row_size(result->type, S_v * H_v * n_tokens * n_seqs));
+
+    return {output, new_state};
+}
+
+std::pair<ggml_tensor *, ggml_tensor *> llm_build_delta_net_base::build_delta_net(
+        ggml_tensor * q,
+        ggml_tensor * k,
+        ggml_tensor * v,
+        ggml_tensor * g,
+        ggml_tensor * b,
+        ggml_tensor * s,
+        int           il) {
+    const int64_t n_seq_tokens = q->ne[2];
+
+    if (n_seq_tokens == 1) {
+        if (cparams.fused_gdn_ar) {
+            return build_delta_net_fused(q, k, v, g, b, s, il);
+        }
+        return build_delta_net_autoregressive(q, k, v, g, b, s, il);
+    }
+
+    if (cparams.fused_gdn_ch) {
+        return build_delta_net_fused(q, k, v, g, b, s, il);
+    }
+
+    return build_delta_net_chunking(q, k, v, g, b, s, il);
+}
+
+ggml_tensor * llm_build_delta_net_base::build_conv_state(
+        llm_graph_input_rs * inp,
+        ggml_tensor *        conv_states_all,
+        ggml_tensor *        qkv_mixed,
+        int64_t              conv_kernel_size,
+        int64_t              conv_channels,
+        int                  il) {
+    const auto * mctx_cur = inp->mctx;
+
+    const auto kv_head  = mctx_cur->get_head();
+    const auto mem_size = mctx_cur->get_size();
+
+    const int64_t n_seqs = ubatch.n_seqs;
+
+    ggml_tensor * conv_states = build_rs(inp, conv_states_all, hparams.n_embd_r(), n_seqs);
+    cb(conv_states, "conv_states", il);
+
+    conv_states = ggml_reshape_3d(ctx0, conv_states, conv_kernel_size - 1, conv_channels, n_seqs);
+    cb(conv_states, "conv_states_reshaped", il);
+
+    qkv_mixed = ggml_transpose(ctx0, qkv_mixed);
+    cb(qkv_mixed, "qkv_mixed_transposed", il);
+
+    ggml_tensor * conv_input = ggml_concat(ctx0, conv_states, qkv_mixed, 0);
+    cb(conv_input, "conv_input", il);
+
+    const int64_t row_count = (conv_kernel_size - 1) * conv_channels;
+
+    const size_t row_size  = ggml_row_size(conv_states_all->type, row_count);
+
+    if (cparams.n_rs_seq == 0) {
+        const int64_t s_idx  = conv_input->ne[0] - conv_states->ne[0];
+        const int64_t s_slot = 0;
+
+        ggml_tensor * conv_state_last =
+            ggml_view_3d(ctx0, conv_input,
+                    conv_kernel_size - 1, conv_channels, n_seqs,
+                    conv_input->nb[1], conv_input->nb[2],
+                    ggml_row_size(conv_input->type, s_idx));
+        cb(conv_state_last, "conv_state_last", il);
+
+        ggml_tensor * conv_state_update =
+            ggml_view_2d(ctx0, conv_states_all,
+                    row_count, n_seqs, conv_states_all->nb[1],
+                    (s_slot * mem_size + kv_head) * row_size);
+        cb(conv_state_update, "conv_state_update", il);
+
+        ggml_build_forward_expand(gf, ggml_cpy(ctx0, conv_state_last, conv_state_update));
+    } else {
+        // [TAG_RECURRENT_ROLLBACK_SPLITS]
+        // this logic assumes that the last (n_rs_seq + 1) tokens of a sequence in a batch are inside
+        //   the same ubatch, which `split_equal()` guarantees via its n_keep_tail argument
+
+        const int64_t K = (int64_t) cparams.n_rs_seq + 1;
+
+        for (int64_t t = 1; t <= K; ++t) {
+            const int64_t s_idx  = std::max<int64_t>(0, conv_input->ne[0] - conv_states->ne[0] - K + t);
+            const int64_t s_slot = K - t;
+
+            ggml_tensor * conv_state_last =
+                ggml_view_3d(ctx0, conv_input,
+                        conv_kernel_size - 1, conv_channels, n_seqs,
+                        conv_input->nb[1], conv_input->nb[2],
+                        ggml_row_size(conv_input->type, s_idx));
+
+            ggml_tensor * conv_state_update =
+                ggml_view_2d(ctx0,
+                        conv_states_all, row_count, n_seqs,
+                        conv_states_all->nb[1],
+                        (s_slot * mem_size + kv_head) * row_size);
+
+            ggml_build_forward_expand(gf, ggml_cpy(ctx0, conv_state_last, conv_state_update));
+        }
+    }
+
+    return conv_input;
+}
+
+ggml_tensor * llm_build_delta_net_base::build_recurrent_attn(
+        llm_graph_input_rs * inp,
+        ggml_tensor *        ssm_states_all,
+        ggml_tensor *        q,
+        ggml_tensor *        k,
+        ggml_tensor *        v,
+        ggml_tensor *        g,
+        ggml_tensor *        b,
+        ggml_tensor *        s,
+        int                  il) {
+    const auto * mctx_cur   = inp->mctx;
+    const auto   kv_head    = mctx_cur->get_head();
+    const uint32_t mem_size = mctx_cur->get_size();
+
+    const int64_t S_v          = s->ne[0];
+    const int64_t H_v          = s->ne[2];
+    const int64_t n_seqs       = s->ne[3];
+    const int64_t n_seq_tokens = q->ne[2];
+
+    const bool keep = cparams.n_rs_seq > 0;
+
+    if (!keep) {
+        auto attn_out = build_delta_net(q, k, v, g, b, s, il);
+        ggml_tensor * output    = attn_out.first;
+        ggml_tensor * new_state = attn_out.second;
+        cb(output, "attn_output", il);
+        cb(new_state, "new_state", il);
+
+        ggml_build_forward_expand(gf,
+                ggml_cpy(ctx0, new_state,
+                    ggml_view_2d(ctx0, ssm_states_all, hparams.n_embd_s(), n_seqs, ssm_states_all->nb[1],
+                        kv_head * hparams.n_embd_s() * ggml_element_size(ssm_states_all))));
+
+        return output;
+    }
+
+    const int64_t D = S_v * S_v * H_v;
+    const int64_t K = cparams.n_rs_seq + 1;
+
+    // state s is 4D [S_v, S_v, H_v, n_seqs]; K snapshot slots are written into the output.
+    ggml_tensor * gdn_out = ggml_gated_delta_net(ctx0, q, k, v, g, b, s, K);
+    if (n_seq_tokens > 1) {
+        res->add_fused_node({LLM_FUSED_OP_GDN_CH, gdn_out, il});
+    } else {
+        res->add_fused_node({LLM_FUSED_OP_GDN_AR, gdn_out, il});
+    }
+
+    const int64_t attn_score_elems    = S_v * H_v * n_seq_tokens * n_seqs;
+    const int64_t state_size_per_snap = S_v * S_v * H_v * n_seqs;
+
+    ggml_tensor * output = ggml_view_4d(ctx0, gdn_out,
+        S_v, H_v, n_seq_tokens, n_seqs,
+        ggml_row_size(gdn_out->type, S_v),
+        ggml_row_size(gdn_out->type, S_v * H_v),
+        ggml_row_size(gdn_out->type, S_v * H_v * n_seq_tokens),
+        0);
+    cb(output, "attn_output", il);
+
+    const size_t row_size = hparams.n_embd_s() * ggml_element_size(ssm_states_all);
+
+    // op writes the last min(n_seq_tokens, K) snapshots; trailing slots are left unwritten
+    const int64_t n_written = std::min<int64_t>(n_seq_tokens, K);
+
+    // write the produced snapshots into the recurrent cache (snapshot slot i -> rollback group i)
+    ggml_tensor * src = ggml_view_3d(ctx0, gdn_out,
+        D, n_seqs, n_written,
+        ggml_row_size(gdn_out->type, D),
+        ggml_row_size(gdn_out->type, state_size_per_snap),
+        ggml_row_size(gdn_out->type, attn_score_elems));
+
+    ggml_tensor * dst = ggml_view_3d(ctx0, ssm_states_all,
+        D, n_seqs, n_written,
+        ssm_states_all->nb[1],
+        (size_t) mem_size * row_size,
+        (size_t) kv_head * row_size);
+
+    ggml_build_forward_expand(gf, ggml_cpy(ctx0, src, dst));
+
+    return output;
+}

src/models/dflash.cpp

@@ -0,0 +1,405 @@
+#include "models.h"
+
+#include "llama-impl.h"
+#include "llama-kv-cache.h"
+#include "llama-kv-cache-iswa.h"
+
+void llama_model_dflash::load_arch_hparams(llama_model_loader & ml) {
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    if (!ml.get_arr(LLM_KV_TARGET_LAYERS, target_layer_ids, false)) {
+        throw std::runtime_error("DFlash model requires 'target_layers' in GGUF metadata");
+    }
+
+    hparams.n_embd_inp_enc_impl = (uint32_t) target_layer_ids.size() * hparams.n_embd;
+
+    LLAMA_LOG_INFO("%s: DFlash extract_layers = [", __func__);
+    for (size_t i = 0; i < target_layer_ids.size(); ++i) {
+        LLAMA_LOG_INFO("%d%s", target_layer_ids[i], i + 1 < target_layer_ids.size() ? ", " : "");
+    }
+    LLAMA_LOG_INFO("]\n");
+
+    // optional interleaved sliding-window attention with per-layer pattern array.
+    // DFlash has a single rope, so the SWA rope == main rope.
+    if (ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false) && hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+    }
+
+    type = LLM_TYPE_UNKNOWN;
+}
+
+void llama_model_dflash::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_embd_inp = hparams.n_embd_inp_enc();
+
+    // DSpark = DFlash + a semi-autoregressive Markov head and Confidence head
+    //
+    // TODO: only Qwen3-style backbones are supported for now; other backbones (e.g. Gemma4)
+    //       need their own conversion path and graph tweaks
+    const struct ggml_tensor * markov_meta = ml->get_tensor_meta("markov_w1.weight");
+    if (markov_meta) {
+        const int64_t dspark_markov_rank = markov_meta->ne[0];
+
+        dspark_markov_w1 = create_tensor(tn(LLM_TENSOR_DSPARK_MARKOV_W1, "weight"), { dspark_markov_rank, n_vocab }, 0);
+        dspark_markov_w2 = create_tensor(tn(LLM_TENSOR_DSPARK_MARKOV_W2, "weight"), { dspark_markov_rank, n_vocab }, 0);
+
+        dspark_conf_proj   = create_tensor(tn(LLM_TENSOR_DSPARK_CONF_PROJ, "weight"), { n_embd + dspark_markov_rank, 1 }, 0);
+        dspark_conf_proj_b = create_tensor(tn(LLM_TENSOR_DSPARK_CONF_PROJ, "bias"),   { 1 },             TENSOR_NOT_REQUIRED);
+
+        LLAMA_LOG_INFO("%s: DFlash with DSpark markov head (rank = %lld)\n", __func__, (long long) dspark_markov_rank);
+    }
+
+    fc              = create_tensor(tn(LLM_TENSOR_FC,              "weight"), { n_embd_inp, n_embd }, 0);
+    output_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), { n_embd }, 0); // encoder hidden_norm (after fc)
+    output_norm     = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM,    "weight"), { n_embd }, 0); // decoder final norm
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+
+        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), { n_embd, n_embd_head_k * n_head }, 0);
+        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), { n_embd, n_embd_k_gqa }, 0);
+        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), { n_embd, n_embd_v_gqa }, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
+
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), { n_embd_head_k }, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), { n_embd_head_k }, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), { n_embd }, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), { n_embd, n_ff }, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_dflash::build_arch_graph(const llm_graph_params & params) const {
+    switch (params.gtype) {
+        case LLM_GRAPH_TYPE_ENCODER:
+            return std::make_unique<graph<true>>(*this, params);
+        case LLM_GRAPH_TYPE_DEFAULT:
+        case LLM_GRAPH_TYPE_DECODER:
+            return std::make_unique<graph<false>>(*this, params);
+        default:
+            GGML_ABORT("invalid graph type");
+    };
+}
+
+template <>
+ggml_tensor * llama_model_dflash::graph<true>::build_inp_embd_enc() const {
+    auto inp_target = std::make_unique<llm_graph_input_embd>(hparams.n_embd_inp_enc());
+
+    inp_target->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd_inp_enc(), n_tokens);
+    ggml_set_input(inp_target->embd);
+
+    ggml_tensor * cur = inp_target->embd;
+    cb(cur, "inp_embd", -1);
+
+    res->add_input(std::move(inp_target));
+
+    return cur;
+}
+
+// DFlash Encoder: processes target model features through feature fusion layer
+template <>
+llama_model_dflash::graph<true>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    ggml_tensor * cur = build_inp_embd_enc();
+
+    cur = build_lora_mm(model.fc, cur);
+    cb(cur, "fc_out", -1);
+
+    cur = build_norm(cur, model.output_norm_enc, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "enc_norm_out", -1);
+
+    ggml_set_output(cur);
+    res->t_h_nextn = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// DSpark (DFlash + Markov & Confidence head): Markov bias on the draft logits, chained per block position
+static void build_dspark_markov_head(llm_graph_context & g, const llama_model & model, ggml_tensor * tokens) {
+    ggml_context * ctx0 = g.ctx0;
+    auto         & res  = g.res;
+
+    ggml_tensor * w1 = model.dspark_markov_w1;
+    ggml_tensor * w2 = model.dspark_markov_w2;
+    GGML_ASSERT(w1 && w2 && model.dspark_conf_proj && "DSpark markov/confidence weights not loaded");
+
+    ggml_tensor * base = res->t_logits; // [n_vocab, n_tokens]
+    const int64_t n_vocab = base->ne[0];
+    const int64_t n_tok   = base->ne[1];
+
+    const auto it = model.gguf_kv.find("dflash.block_size");
+    GGML_ASSERT(it != model.gguf_kv.end() && "DSpark draft requires 'dflash.block_size' in GGUF metadata");
+    const int64_t block_size = std::stoi(it->second);
+    GGML_ASSERT(block_size > 0);
+
+    const int64_t n_blocks = g.ubatch.n_seqs_unq;
+    GGML_ASSERT(n_blocks > 0 && n_tok % n_blocks == 0 && "DSpark markov head requires equal-size blocks");
+    // runtime tokens per block in this ubatch (anchor + drafted positions), bounded by training block_size
+    const int64_t block_drafts = n_tok / n_blocks;
+    if (block_drafts > block_size) {
+        return;
+    }
+
+    // anchor (committed last) token of every block: token 0 of each block, i.e. a strided view
+    const size_t token_stride = (size_t) block_drafts * tokens->nb[0];
+    const size_t base_stride = (size_t) block_drafts * base->nb[1];
+
+    ggml_tensor * prev = ggml_view_2d(ctx0, tokens, 1, n_blocks, token_stride, 0);
+    prev = ggml_cont_1d(ctx0, prev, n_blocks);
+
+    // confidence head input: predicts per-position acceptance
+    ggml_tensor * conf_inp = res->t_embd; // [n_embd, n_tok]
+
+    ggml_tensor * cat      = nullptr;
+    ggml_tensor * cat_conf = nullptr;
+
+    // TODO: the in-graph chain is greedy (argmax); sampling params affect only the final
+    //       token pick, not the Markov conditioning path
+    for (int64_t i = 0; i < block_drafts; ++i) {
+        ggml_tensor * w1_prev = ggml_get_rows(ctx0, w1, prev);   // [R, n_blocks]
+        ggml_tensor * bias    = ggml_mul_mat(ctx0, w2, w1_prev); // [n_vocab, n_blocks]
+
+        // position i of every block: strided view [n_vocab, n_blocks]
+        ggml_tensor * base_i = ggml_view_2d(ctx0, base, n_vocab, n_blocks, base_stride, i*base->nb[1]);
+        ggml_tensor * col    = ggml_add(ctx0, base_i, bias);
+
+        cat = cat ? ggml_concat(ctx0, cat, col, 1) : col;
+
+        // conf(i) = sigmoid(conf_proj . [conf_inp(i); markov_w1[prev(i)]] + b)  -- [1, n_blocks]
+        ggml_tensor * conf_inp_i = ggml_view_2d(ctx0, conf_inp, conf_inp->ne[0], n_blocks,
+                                                (size_t) block_drafts * conf_inp->nb[1], i*conf_inp->nb[1]);
+        ggml_tensor * feat = ggml_concat(ctx0, ggml_cont(ctx0, conf_inp_i), w1_prev, 0);
+        ggml_tensor * conf = ggml_mul_mat(ctx0, model.dspark_conf_proj, feat);
+        if (model.dspark_conf_proj_b) {
+            conf = ggml_add(ctx0, conf, model.dspark_conf_proj_b);
+        }
+        conf = ggml_sigmoid(ctx0, conf);
+
+        cat_conf = cat_conf ? ggml_concat(ctx0, cat_conf, conf, 1) : conf;
+
+        if (i + 1 < block_drafts) {
+            prev = ggml_argmax(ctx0, col);
+        }
+    }
+
+    // cat is position-major; restore ubatch block-major order
+    ggml_tensor * out = ggml_reshape_3d(ctx0, cat, n_vocab, n_blocks, block_drafts);
+    out = ggml_cont(ctx0, ggml_permute(ctx0, out, 0, 2, 1, 3)); // [n_vocab, block_drafts, n_blocks]
+    out = ggml_reshape_2d(ctx0, out, n_vocab, n_tok);
+
+    {
+        ggml_tensor * conf = ggml_reshape_3d(ctx0, cat_conf, 1, n_blocks, block_drafts);
+        conf = ggml_cont(ctx0, ggml_permute(ctx0, conf, 0, 2, 1, 3));
+        conf = ggml_reshape_2d(ctx0, conf, 1, n_tok);
+
+        // note: broadcast the [1, n_tok] confidences to n_embd-wide rows to be able to reuse `llama_get_embeddings_nextn`
+        conf = ggml_repeat(ctx0, conf, res->t_embd);
+        res->t_h_nextn = conf;
+        ggml_build_forward_expand(g.gf, conf);
+    }
+
+    res->t_logits = out;
+    ggml_build_forward_expand(g.gf, out);
+}
+
+// DFlash decoder, dual-mode by batch type:
+//   * embd batch  -> fused target features: project + inject K/V into the cache.
+//   * token batch -> noise-block diffusion: attend over [committed, MASK...] to generate draft tokens
+template <>
+llama_model_dflash::graph<false>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * inp_pos  = build_inp_pos();
+
+    // optional iSWA: pick the matching attention input
+    const bool use_iswa = hparams.swa_type != LLAMA_SWA_TYPE_NONE;
+
+    llm_graph_input_attn_kv      * inp_attn      = nullptr;
+    llm_graph_input_attn_kv_iswa * inp_attn_iswa = nullptr;
+    if (use_iswa) {
+        inp_attn_iswa = build_attn_inp_kv_iswa();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+
+    const float kq_scale = 1.0f/sqrtf(float(n_embd_head));
+
+    // KV cache injection
+    if (ubatch.embd) {
+        auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
+
+        inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, n_tokens);
+        ggml_set_input(inp->embd);
+
+        ggml_tensor * inp_g = inp->embd;
+        cb(inp_g, "inp_g_embeddings", -1);
+
+        res->add_input(std::move(inp));
+
+        for (int il = 0; il < n_layer; ++il) {
+            const auto & layer = model.layers[il];
+
+            ggml_tensor * Kcur = build_lora_mm(layer.wk, inp_g);
+            ggml_tensor * Vcur = build_lora_mm(layer.wv, inp_g);
+
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            Kcur = build_norm(Kcur, layer.attn_k_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+            cb(Kcur, "Kcur_injected", il);
+            cb(Vcur, "Vcur_injected", il);
+
+            if (use_iswa) {
+                // route each layer's K/V to its sub-cache: SWA layers -> sliding cache, full -> dense
+                const bool    is_swa = hparams.is_swa(il);
+                const auto  * kv     = is_swa ? inp_attn_iswa->mctx->get_swa() : inp_attn_iswa->mctx->get_base();
+                ggml_tensor * k_idxs = is_swa ? inp_attn_iswa->get_k_idxs_swa() : inp_attn_iswa->get_k_idxs();
+                ggml_tensor * v_idxs = is_swa ? inp_attn_iswa->get_v_idxs_swa() : inp_attn_iswa->get_v_idxs();
+                // rotate K/V into the cache's rotated space
+                ggml_tensor * k_rot  = is_swa ? inp_attn_iswa->self_k_rot_swa : inp_attn_iswa->self_k_rot;
+                ggml_tensor * v_rot  = is_swa ? inp_attn_iswa->self_v_rot_swa : inp_attn_iswa->self_v_rot;
+                if (k_rot) {
+                    Kcur = llama_mul_mat_hadamard(ctx0, Kcur, k_rot);
+                }
+                if (v_rot) {
+                    Vcur = llama_mul_mat_hadamard(ctx0, Vcur, v_rot);
+                }
+                ggml_build_forward_expand(gf, kv->cpy_k(ctx0, Kcur, k_idxs, il));
+                ggml_build_forward_expand(gf, kv->cpy_v(ctx0, Vcur, v_idxs, il));
+            } else {
+                // rotate K/V into the cache's rotated space
+                if (inp_attn->self_k_rot) {
+                    Kcur = llama_mul_mat_hadamard(ctx0, Kcur, inp_attn->self_k_rot);
+                }
+                if (inp_attn->self_v_rot) {
+                    Vcur = llama_mul_mat_hadamard(ctx0, Vcur, inp_attn->self_v_rot);
+                }
+                ggml_build_forward_expand(gf, inp_attn->mctx->cpy_k(ctx0, Kcur, inp_attn->get_k_idxs(), il));
+                ggml_build_forward_expand(gf, inp_attn->mctx->cpy_v(ctx0, Vcur, inp_attn->get_v_idxs(), il));
+            }
+        }
+
+        res->t_embd = inp_g;
+
+        ggml_build_forward_expand(gf, inp_g);
+        return;
+    }
+
+    // tok_embd from the target model (shared via ctx_other)
+    auto * tok_embd = model.tok_embd;
+    if (tok_embd == nullptr) {
+        GGML_ASSERT(cparams.ctx_other != nullptr);
+        const auto * model_other = llama_get_model(cparams.ctx_other);
+
+        GGML_ASSERT(model_other->tok_embd != nullptr && "DFlash decoder requires the target model's token embeddings");
+        tok_embd = model_other->tok_embd;
+    }
+
+    auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
+
+    inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
+    ggml_set_input(inp->tokens);
+
+    ggml_tensor * inp_tokens = inp->tokens;
+
+    ggml_tensor * inpL = ggml_get_rows(ctx0, tok_embd, inp->tokens);
+    cb(inpL, "inp_noise_embd", -1);
+
+    res->add_input(std::move(inp));
+
+    for (int il = 0; il < n_layer; ++il) {
+        const auto & layer = model.layers[il];
+
+        ggml_tensor * noise_norm = build_norm(inpL, layer.attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(noise_norm, "noise_norm", il);
+
+        ggml_tensor * Qcur = build_lora_mm(layer.wq, noise_norm);
+        ggml_tensor * Kcur = build_lora_mm(layer.wk, noise_norm);
+        ggml_tensor * Vcur = build_lora_mm(layer.wv, noise_norm);
+
+        Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
+        Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+        Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+        Qcur = build_norm(Qcur, layer.attn_q_norm, NULL, LLM_NORM_RMS, il);
+        Kcur = build_norm(Kcur, layer.attn_k_norm, NULL, LLM_NORM_RMS, il);
+
+        Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+        Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+        cb(Qcur, "Qcur", il);
+        cb(Kcur, "Kcur", il);
+        cb(Vcur, "Vcur", il);
+
+        // cache-aware, non-causal attention
+        ggml_tensor * cur = use_iswa
+            ? build_attn(inp_attn_iswa, layer.wo, NULL, NULL, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il)
+            : build_attn(inp_attn,      layer.wo, NULL, NULL, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, layer.ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                layer.ffn_up,   NULL, NULL,
+                layer.ffn_gate, NULL, NULL,
+                layer.ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "l_out", il);
+
+        inpL = cur;
+    }
+
+    ggml_tensor * cur = build_norm(inpL, model.output_norm, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+
+    res->t_embd = cur;
+
+    // lm_head from the target model (shared via ctx_other)
+    auto * output = model.output;
+    if (output == nullptr) {
+        GGML_ASSERT(cparams.ctx_other != nullptr);
+        const auto * model_other = llama_get_model(cparams.ctx_other);
+        GGML_ASSERT(model_other->output != nullptr && "DFlash decoder requires the target model's output projection");
+        output = model_other->output;
+    }
+
+    cur = build_lora_mm(output, cur);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+
+    // DSpark: bias the draft logits with the Markov head
+    if (model.dspark_markov_w1) {
+        build_dspark_markov_head(*this, model, inp_tokens);
+    }
+}

src/models/dots1.cpp

@@ -0,0 +1,193 @@
+#include "models.h"
+
+void llama_model_dots1::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+
+    switch (hparams.n_layer()) {
+        case 62: type = LLM_TYPE_142B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_dots1::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_head_k * n_head, n_embd_head_k * n_head, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (i < (int) hparams.n_layer_dense_lead) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        } else {
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            // MoE branch
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+            // Shared expert branch
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_dots1::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_dots1::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // MoE branch
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/dream.cpp

@@ -0,0 +1,138 @@
+#include "models.h"
+
+void llama_model_dream::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    // Dream models are primarily 7B with 28 layers
+    switch (hparams.n_layer()) {
+        case 28:
+            type = LLM_TYPE_7B;
+            break;
+        default:
+            type = LLM_TYPE_UNKNOWN;
+    }
+    // Set non-causal attention for diffusion models
+    hparams.causal_attn = false;
+}
+
+void llama_model_dream::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    output_b    = create_tensor(tn(LLM_TENSOR_OUTPUT,      "bias"),   {n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_dream::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_dream::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    //copied from qwen2
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+            model.layers[il].ffn_up, NULL, NULL,
+            model.layers[il].ffn_gate, NULL, NULL,
+            model.layers[il].ffn_down, NULL, NULL,
+            NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/eagle3.cpp

@@ -0,0 +1,338 @@
+#include "models.h"
+
+void llama_model_eagle3::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    if (!ml.get_arr(LLM_KV_TARGET_LAYERS, target_layer_ids, false)) {
+        throw std::runtime_error("EAGLE3 model requires 'extract_layers' in GGUF metadata");
+    }
+    if (target_layer_ids.size() != 3) {
+        throw std::runtime_error("EAGLE3 requires exactly 3 entries in 'extract_layers'");
+    }
+    LLAMA_LOG_INFO("%s: EAGLE3 extract_layers = [%d, %d, %d]\n", __func__,
+            target_layer_ids[0],
+            target_layer_ids[1],
+            target_layer_ids[2]);
+
+    uint32_t n_embd_tgt = 0;
+
+    ml.get_key(LLM_KV_TARGET_HIDDEN_SIZE, n_embd_tgt);
+    LLAMA_LOG_INFO("%s: EAGLE3 n_embd_tgt = %u (draft n_embd = %u)\n", __func__, n_embd_tgt, hparams.n_embd);
+
+    hparams.n_embd_inp_enc_impl = (uint32_t) target_layer_ids.size() * n_embd_tgt;
+
+    // eagle3 norm_before_residual (optional, default false)
+    // compatible with Readhat eagle3 speculator model
+    ml.get_key(LLM_KV_NORM_BEFORE_RESIDUAL, hparams.norm_before_residual, false);
+    if (hparams.norm_before_residual) {
+        LLAMA_LOG_INFO("%s: EAGLE3gnorm_before_residual = true\n", __func__);
+    }
+
+    // eagle3 norm_before_fc (optional, default false)
+    // compatible with eagle3.1 (e.g. nvidia/gpt-oss-120b-Eagle3-v3)
+    ml.get_key(LLM_KV_NORM_BEFORE_FC, hparams.norm_before_fc, false);
+
+    type = LLM_TYPE_UNKNOWN;
+}
+
+void llama_model_eagle3::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_embd_inp = hparams.n_embd_inp_enc();
+    const int64_t n_embd_attn_input = 2 * n_embd;
+
+    // Get vocab size from the d2t tensor in the GGUF file (optional - only needed if eagle3 has different vocab_size than target)
+    // d2t: draft to target vocabulary mapping
+    int64_t n_draft_vocab = n_vocab;  // Default: same as target vocab
+    const struct ggml_tensor * d2t_meta = ml->get_tensor_meta("d2t");
+    if (d2t_meta) {
+        n_draft_vocab = d2t_meta->ne[0]; // update draft vocab size
+        d2t = create_tensor(tn(LLM_TENSOR_D2T), {n_draft_vocab}, 0);
+        LLAMA_LOG_INFO("%s: EAGLE3 using d2t mapping (draft_vocab_size = %lld)\n", __func__, (long long)n_draft_vocab);
+    } else {
+        d2t = nullptr; // no d2t, use default vocab size
+        LLAMA_LOG_INFO("%s: EAGLE3 without d2t - sharing same vocab_size with target (vocab_size = %lld)\n", __func__, (long long)n_draft_vocab);
+    }
+
+    // Feature fusion layer: projects 3 target layers to draft hidden size
+    fc = create_tensor(tn(LLM_TENSOR_FC, "weight"), {n_embd_inp, n_embd}, 0);
+
+    // RMSNorm on the fused target features (input to fc), only when norm_before_fc is set.
+    if (hparams.norm_before_fc) {
+        output_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), {n_embd_inp}, 0);
+    }
+
+    // Output layer (uses draft vocab size)
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_draft_vocab}, TENSOR_NOT_REQUIRED);
+
+    // Token embeddings (optional - Llama 3.3 70B EAGLE3 has its own)
+    const struct ggml_tensor * tok_embd_meta = ml->get_tensor_meta(tn(LLM_TENSOR_TOKEN_EMBD, "weight").str().c_str());
+    if (tok_embd_meta) {
+        const int64_t n_target_vocab = tok_embd_meta->ne[1];
+        tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_target_vocab}, 0);
+        LLAMA_LOG_INFO("%s: EAGLE3 using its own token_embd (vocab = %lld)\n", __func__, (long long)n_target_vocab);
+    }
+
+    // Single decoder layer
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // input_layernorm: applied to token embeddings
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        // eagle3 specific: hidden_norm applied to fused target features
+        layer.attn_norm_2 = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, 0);
+
+        // Attention takes input_embeds_normed + fused_target_normed as input
+        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd_attn_input, n_embd_head_k * n_head}, 0);
+        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd_attn_input, n_embd_k_gqa}, 0);
+        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd_attn_input, n_embd_v_gqa}, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+        // rope_freqs for llama3 rope scaling (optional - only if eagle3 config has rope_scaling)
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_eagle3::build_arch_graph(const llm_graph_params & params) const {
+    switch (params.gtype) {
+        case LLM_GRAPH_TYPE_ENCODER:
+            return std::make_unique<graph<true>>(*this, params);
+        case LLM_GRAPH_TYPE_DEFAULT:
+        case LLM_GRAPH_TYPE_DECODER:
+            return std::make_unique<graph<false>>(*this, params);
+        default:
+            GGML_ABORT("invalid graph type");
+    };
+}
+
+template <>
+ggml_tensor * llama_model_eagle3::graph<true>::build_inp_embd_enc() const {
+    ggml_tensor * cur = nullptr;
+
+    // Input: Target model features (3 layers concatenated: low, mid, high)
+    // Data will be provided via ubatch->embd in encode_eagle3_features()
+    auto inp_target = std::make_unique<llm_graph_input_embd>(hparams.n_embd_inp_enc());
+    inp_target->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd_inp_enc(), n_tokens);
+    ggml_set_input(inp_target->embd);
+
+    cur = inp_target->embd;
+    cb(cur, "inp_embd", -1);
+
+    res->add_input(std::move(inp_target));
+
+    return cur;
+}
+
+// eagle3 Encoder: processes target model features through feature fusion layer
+// Input: target_features e.g. [12288, n_tokens] from target model layers low, middle, high
+// Output: g_embeddings e.g. [4096, n_tokens] stored in context
+template <>
+llama_model_eagle3::graph<true>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    ggml_tensor * cur = nullptr;
+
+    cur = build_inp_embd_enc();
+
+    // RMSNorm on the fused target features before fc
+    if (hparams.norm_before_fc) {
+        cur = build_norm(cur, model.output_norm_enc, NULL, LLM_NORM_RMS, -1);
+        cb(cur, "enc_input_norm", -1);
+    }
+
+    // Feature fusion layer
+    cur = build_lora_mm(model.fc, cur);
+    cb(cur, "fc_out", -1);
+
+    // Output: g_embeddings e.g. [4096, n_tokens]
+    // store in t_h_nextn (same as MTP) so can be read via llama_get_embeddings_nextn(ctx_dft)
+    ggml_set_output(cur);
+    res->t_h_nextn = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// eagle3 Decoder: processes draft tokens using g_embeddings from encoder
+// Input: draft tokens + g_embeddings from encoder
+// Output: draft logits
+template <>
+llama_model_eagle3::graph<false>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_layer == 1);  // eagle3 has only one decoder layer
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    // eagle3 Decoder receives:
+    // 1. Token embeddings (e.g.from eagle3's own tok_embd for Llama 3.3 70B, or target model for Llama 3.1 8B)
+    // 2. g_embeddings from encoder
+    auto * tok_embd = model.tok_embd;
+    if (model.tok_embd == nullptr) {
+        GGML_ASSERT(cparams.ctx_other != nullptr);
+        const auto * model_other = llama_get_model(cparams.ctx_other);
+
+        GGML_ASSERT(model_other->tok_embd != nullptr && "EAGLE3 decoder requires token embeddings (own or from target model)");
+        tok_embd = model_other->tok_embd;
+    }
+
+    auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
+
+    inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
+    ggml_set_input(inp->tokens);
+
+    inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, n_tokens);
+    ggml_set_input(inp->embd);
+
+    ggml_tensor * inp_embd = ggml_get_rows(ctx0, tok_embd, inp->tokens);
+    cb(inp_embd, "inp_embd", -1);
+
+    ggml_tensor * inp_g = inp->embd;
+    cb(inp_g, "inp_g_embeddings", -1);
+
+    res->add_input(std::move(inp));
+
+    inpL = inp_g;
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale = 1.0f/sqrtf(float(n_embd_head));
+
+    // Single decoder layer (il = 0)
+    const int il = 0;
+    {
+        // Apply input_layernorm to the token embeddings
+        ggml_tensor * embd_norm = build_norm(inp_embd,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(embd_norm, "embd_norm", il);
+
+        // Apply hidden_norm to inp_g
+        ggml_tensor * g_norm = build_norm(inp_g,
+                model.layers[il].attn_norm_2, NULL,
+                LLM_NORM_RMS, -1);
+        cb(g_norm, "g_norm", il);
+
+        // norm_before_residual: determines what goes into the residual connection (compatible with Readhat eagle3 speculator model)
+        // - false (default): use raw inp_g for residual
+        // - true: use normalized g_norm for residual
+        // inpL is the concatenated input (normalized inp_embd + normalized inp_g)
+        ggml_tensor * inpSA = hparams.norm_before_residual ? g_norm : inpL;
+
+        // Concatenate normalized inp_embd and normalized inp_g
+        cur = ggml_concat(ctx0, embd_norm, g_norm, il);
+        cb(cur, "concat_embd", il);
+
+        // Self-attention with concatenated input
+        ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+        cb(Qcur, "Qcur", il);
+
+        ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+        cb(Kcur, "Kcur", il);
+
+        ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+        cb(Vcur, "Vcur", il);
+
+        Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
+        Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+        Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+        // rope freq factors, returns nullptr if not available
+        ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+        // RoPE
+        Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, rope_factors,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+        Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, rope_factors,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+        cb(Qcur, "Qcur_rope", il);
+        cb(Kcur, "Kcur_rope", il);
+
+        cur = build_attn(inp_attn,
+                model.layers[il].wo, NULL, nullptr,
+                Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+
+        // Add residual and update it
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // Apply FFN norm to the sum
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "post_attn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        // Output norm with residual
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "eagle3_prenorm", il);
+
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    // Output prenorm state (for next token's g_embeddings in autoregressive generation)
+    ggml_set_output(cur);
+    res->t_h_nextn = cur;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+
+    // lm_head - projects to draft vocabulary
+    // if the draft has no own output projection, inherit the target model's lm_head
+    auto * output = model.output;
+    if (output == nullptr) {
+        GGML_ASSERT(cparams.ctx_other != nullptr);
+        const auto * model_other = llama_get_model(cparams.ctx_other);
+
+        GGML_ASSERT(model_other->output != nullptr && "EAGLE3 decoder requires an output projection (own or from target model)");
+        output = model_other->output;
+    }
+    cur = build_lora_mm(output, cur);
+
+    if (model.d2t) {
+        const int64_t n_draft_vocab = cur->ne[0];
+        const int64_t n_outputs     = cur->ne[1];
+        const int64_t n_vocab       = (int64_t) model.vocab.n_tokens();
+
+        GGML_ASSERT(model.d2t->type == GGML_TYPE_I64);
+        GGML_ASSERT(model.d2t->ne[0] == n_draft_vocab);
+
+        ggml_tensor * logits = ggml_fill(ctx0, ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, 1, n_vocab, n_outputs), -INFINITY);
+        cur = ggml_set_rows(ctx0, logits,
+                ggml_reshape_3d(ctx0, cur,       1,             n_draft_vocab, n_outputs),
+                ggml_reshape_3d(ctx0, model.d2t, n_draft_vocab, 1,             1));
+        cur = ggml_reshape_2d(ctx0, cur, n_vocab, n_outputs);
+    }
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/ernie4-5-moe.cpp

@@ -0,0 +1,133 @@
+#include "models.h"
+
+std::unique_ptr<llm_graph_context> llama_model_ernie4_5_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_ernie4_5_moe::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    GGML_ASSERT(hparams.n_moe_layer_step > 0 && "Ernie 4.5 MoE requires n_moe_layer_step > 0");
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+        // norm
+        {
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+        }
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        bool is_moe_layer =
+            static_cast<uint32_t>(il) >= hparams.n_layer_dense_lead && (il + 1) % hparams.n_moe_layer_step == 0;
+
+        if (!is_moe_layer) {
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                                        model.layers[il].ffn_gate_inp,
+                                        model.layers[il].ffn_up_exps,
+                                        model.layers[il].ffn_gate_exps,
+                                        model.layers[il].ffn_down_exps,
+                                        model.layers[il].ffn_exp_probs_b,
+                                        n_expert, n_expert_used,
+                                        LLM_FFN_SILU, true,
+                                        hparams.expert_weights_scale,
+                                        LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                                        il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // Shared expert (if present)
+            if (hparams.n_ff_shexp > 0) {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+            } else {
+                cur = moe_out;
+            }
+            cb(cur, "ffn_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/ernie4-5.cpp

@@ -0,0 +1,164 @@
+#include "models.h"
+
+void llama_model_ernie4_5::load_arch_hparams(llama_model_loader & ml) {
+    // paddleocr need mrope_section
+    ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false);
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    if (arch == LLM_ARCH_ERNIE4_5_MOE) {
+        ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+        ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+        ml.get_key(LLM_KV_INTERLEAVE_MOE_LAYER_STEP,         hparams.n_moe_layer_step);
+        ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,         hparams.n_layer_dense_lead, false);
+    }
+
+    switch (hparams.n_layer()) {
+        case 18: type = LLM_TYPE_0_3B; break;
+        case 28: type = LLM_TYPE_21B_A3B; break;
+        case 54: type = LLM_TYPE_300B_A47B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_ernie4_5::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (arch == LLM_ARCH_ERNIE4_5_MOE && static_cast<uint32_t>(i) >= hparams.n_layer_dense_lead) { // MoE layers
+            int n_ff_exp = hparams.n_ff_exp;
+
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff_exp, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff_exp, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff_exp, n_expert}, 0);
+
+            // Shared expert (if present)
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {    n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd    }, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {    n_embd, hparams.n_ff_shexp}, 0);
+            }
+        } else { // Dense layers
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_ernie4_5::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_ernie4_5::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        {
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+        }
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1) {
+            // skip computing output for unused tokens
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        {
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/eurobert.cpp

@@ -0,0 +1,124 @@
+#include "models.h"
+
+void llama_model_eurobert::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    if (hparams.n_layer() == 12) {
+        type = LLM_TYPE_SMALL;  // 0.2B
+    }
+}
+
+void llama_model_eurobert::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_eurobert::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_eurobert::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    inpL = build_inp_embd(model.tok_embd);
+    cb(inpL, "inp_embd", -1);
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * cur = inpL;
+
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, nullptr, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+            cb(cur, "kqv_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        cur = ggml_add(ctx0, cur, inpL);
+
+        ggml_tensor * ffn_inp = cur;
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_embd", -1);
+    res->t_embd = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/exaone-moe.cpp

@@ -0,0 +1,244 @@
+#include "models.h"
+
+void llama_model_exaone_moe::load_arch_hparams(llama_model_loader & ml) {
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    hparams.n_swa = 128;
+    uint32_t swa_period = 4;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+    hparams.set_swa_pattern(swa_period);
+    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,                hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,          hparams.n_swa);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,               hparams.n_expert_shared, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,                hparams.expert_gating_func);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,              hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,               hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,         hparams.n_layer_dense_lead, false);
+
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl");
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_30B_A3B; break;
+        case 48: type = LLM_TYPE_235B_A22B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_exaone_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_ff_exp       = hparams.n_ff_exp;
+    const int64_t n_ff_shexp     = hparams.n_ff_shexp > 0 ? hparams.n_ff_shexp : n_ff_exp;
+    const int64_t head_dim       = hparams.n_embd_head_k();
+    const int64_t n_qo_dim       = n_head * head_dim;
+    const int64_t n_kv_dim       = n_head_kv * head_dim;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        int flags = 0;
+        if (i >= n_layer) {
+            // skip all tensors in the NextN layers
+            flags |= TENSOR_SKIP;
+        }
+
+        auto & layer = layers[i];
+        create_tensor_qkv(layer, i, n_embd, n_qo_dim, n_kv_dim, n_kv_dim, flags);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_qo_dim, n_embd}, flags);
+
+        layer.rope_freqs   = create_tensor(tn(LLM_TENSOR_ROPE_FREQS,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0) | flags);
+
+        layer.attn_norm    = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "weight", i), {n_embd}, flags);
+        layer.attn_q_norm  = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, flags);
+        layer.attn_k_norm  = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, flags);
+
+        layer.ffn_norm     = create_tensor(tn(LLM_TENSOR_FFN_NORM,    "weight", i), {n_embd}, flags);
+
+        // dense layers for first n_layer_dense_lead layers or nextn_predict_layers layers at the end
+        if (i < (int) hparams.n_layer_dense_lead || (i >= n_layer)) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, flags);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, flags);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, flags);
+        } else {
+            layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, flags);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED | flags);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            layer.ffn_gate_exps  = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS,  "weight", i), {n_embd, n_ff_exp, n_expert}, flags);
+            layer.ffn_down_exps  = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS,  "weight", i), {n_ff_exp, n_embd, n_expert}, flags);
+            layer.ffn_up_exps    = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,    "weight", i), {n_embd, n_ff_exp, n_expert}, flags);
+
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, flags);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, flags);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, flags);
+        }
+
+        // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers
+        if (i >= n_layer) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), {2 * n_embd, n_embd}, flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM,   "weight", i), {n_embd}, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM,   "weight", i), {n_embd}, flags);
+
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), {n_embd}, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS,     "weight", i), {n_embd, n_vocab}, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), {n_embd, n_vocab}, flags | TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_exaone_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_exaone_moe::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_k();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_v());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn_iswa = build_attn_inp_kv_iswa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // use RoPE for SWA layers
+        const bool is_local_layer = hparams.is_swa(il);
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+            cb(Kcur, "Kcur_normed", il);
+
+            if (is_local_layer) {
+                Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base,
+                                     freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base,
+                                     freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
+            }
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn_iswa,
+                model.layers[il].wo, NULL, model.layers[il].wo_s,
+                Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // norm
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+            // dense branch
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL, NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // FFN shared expert
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    // final norm
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/exaone.cpp

@@ -0,0 +1,136 @@
+#include "models.h"
+
+void llama_model_exaone::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_8B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_exaone::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM,   "weight", i), {n_embd}, 0);
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        layer.ffn_gate   = create_tensor(tn(LLM_TENSOR_FFN_GATE,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN,   "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,     "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_exaone::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_exaone::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/exaone4.cpp

@@ -0,0 +1,195 @@
+#include "models.h"
+
+void llama_model_exaone4::load_arch_hparams(llama_model_loader & ml) {
+    if (hparams.n_layer() == 64) {    // 32B
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+        hparams.n_swa = 4096;
+        uint32_t swa_period = 4;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period);
+
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    }
+
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS,        hparams.n_layer_nextn, false);
+
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer");
+
+    switch (hparams.n_layer()) {
+        case 30: type = LLM_TYPE_1_2B; break;
+        case 64: type = LLM_TYPE_32B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_exaone4::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        const bool is_nextn = i >= n_layer;
+        int flags = 0;
+        if (is_nextn) {
+            // NextN/MTP layers are preserved in GGUF but are not executed yet.
+            flags |= TENSOR_SKIP;
+        }
+
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, flags);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, flags);
+
+        if (!is_nextn) {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, flags);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, flags);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, flags);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, flags);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, flags);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, flags);
+        layer.ffn_post_norm  = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, flags);
+
+        if (is_nextn) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), {2 * n_embd, n_embd}, flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM,   "weight", i), {n_embd}, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM,   "weight", i), {n_embd}, flags);
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), {n_embd}, flags | TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_exaone4::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_STANDARD) {
+        return std::make_unique<graph<true>>(*this, params);
+    } else {
+        return std::make_unique<graph<false>>(*this, params);
+    }
+}
+
+template <bool iswa>
+llama_model_exaone4::graph<iswa>::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_k();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_v());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    using inp_attn_type      = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa, llm_graph_input_attn_kv>;
+    inp_attn_type * inp_attn = nullptr;
+
+    if constexpr (iswa) {
+        inp_attn = build_attn_inp_kv_iswa();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // use RoPE for SWA layers or non-SWA models
+        const bool use_rope = hparams.is_swa(il) || hparams.swa_type == LLAMA_SWA_TYPE_NONE;
+
+        cur = inpL;
+
+        // self-attention
+        {
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+            cb(Kcur, "Kcur_normed", il);
+
+            if (use_rope) {
+                Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base,
+                                     freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base,
+                                     freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
+            }
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        cur = build_norm(cur, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_ffn(ffn_inp,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL, NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = build_norm(cur, model.layers[il].ffn_post_norm, NULL, LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", -1);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// Explicit template instantiations
+template struct llama_model_exaone4::graph<false>;
+template struct llama_model_exaone4::graph<true>;

src/models/falcon-h1.cpp

@@ -0,0 +1,209 @@
+#include "models.h"
+
+void llama_model_falcon_h1::load_arch_hparams(llama_model_loader & ml) {
+    // Common parameters
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    // SSM parameters
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
+    ml.get_key(LLM_KV_SSM_STATE_SIZE,     hparams.ssm_d_state);
+    ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
+    ml.get_key(LLM_KV_SSM_GROUP_COUNT,    hparams.ssm_n_group);
+
+    std::fill(hparams.is_recr_impl.begin(), hparams.is_recr_impl.end(), true);
+
+    switch (hparams.n_layer()) {
+        case 36:
+            type = LLM_TYPE_0_5B; break;
+        case 24:
+            type = LLM_TYPE_1_5B; break;
+        case 66:
+            type = LLM_TYPE_1B; break;
+        case 32:
+            type = LLM_TYPE_3B; break;
+        case 44:
+            type = LLM_TYPE_7B; break;
+        case 72:
+            type = LLM_TYPE_34B; break;
+        default:
+            type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_falcon_h1::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    // Common
+    const int64_t hidden_size = hparams.n_embd; // hidden_size
+
+    // mamba2 Mixer SSM params
+    const int64_t ssm_conv_kernel_size  = hparams.ssm_d_conv; // ssm_conv_kernel_size
+    const int64_t ssm_n_groups          = hparams.ssm_n_group; // ssm_n_groups
+    const int64_t ssm_state_size        = hparams.ssm_d_state; // ssm_state_size
+    const int64_t ssm_intermediate_size = hparams.ssm_d_inner; // TODO expand
+    const int64_t ssm_num_heads         = hparams.ssm_dt_rank; // ssm_num_heads
+    const int64_t ssm_conv_dim          = ssm_intermediate_size + 2 * ssm_n_groups * ssm_state_size;
+    const int64_t ssm_projection_size   = ssm_intermediate_size + ssm_conv_dim + ssm_num_heads;
+
+    // attn params
+    const int64_t attn_num_attention_head = hparams.n_head(0); // rename to: attn_num_attention_head
+    const int64_t attn_num_key_value_head = hparams.n_head_kv(0);
+
+    // ffn params
+    const int64_t ffn_intermediate_size = hparams.n_ff(0);
+
+    // embeddings
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {hidden_size, n_vocab}, 0);
+
+    // output
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {hidden_size, n_vocab}, TENSOR_NOT_REQUIRED);
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {hidden_size}, 0);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {hidden_size, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        /*SSM LAYERS*/
+        // ssm in
+        layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {hidden_size, ssm_projection_size}, 0);
+        // ssm 1d conv
+        layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {ssm_conv_kernel_size, ssm_conv_dim}, 0);
+        layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {ssm_conv_dim}, TENSOR_NOT_REQUIRED);
+        // ssm_dt
+        layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {ssm_num_heads}, 0);
+        // no "weight" suffix for these
+        layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {1, ssm_num_heads}, 0);
+        layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {1, ssm_num_heads}, 0);
+        // ssm_norm
+        layer.ssm_norm = create_tensor(tn(LLM_TENSOR_SSM_NORM, "weight", i), {ssm_intermediate_size / ssm_n_groups, ssm_n_groups}, TENSOR_NOT_REQUIRED);
+        // out_proj
+        layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {ssm_intermediate_size, hidden_size}, 0);
+
+        /*ATTENTION LAYERS*/
+        // attention layers (with optional bias)
+        create_tensor_qkv(layer, i, hidden_size, n_embd_head_k * attn_num_attention_head, attn_num_key_value_head * n_embd_head_k, attn_num_key_value_head * n_embd_head_v, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * attn_num_attention_head, hidden_size}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {hidden_size}, TENSOR_NOT_REQUIRED);
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {hidden_size}, 0);
+
+
+        // feed forward (w/ optional biases)
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, i), {hidden_size}, 0);
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {hidden_size,   ffn_intermediate_size}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  ffn_intermediate_size, hidden_size}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {hidden_size,   ffn_intermediate_size}, 0);
+
+        layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {ffn_intermediate_size}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {hidden_size}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {ffn_intermediate_size}, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_falcon_h1::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_falcon_h1::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_build_mamba_base(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // Build the inputs in the recurrent & kv cache
+    auto * inp = build_inp_mem_hybrid();
+
+    const float kq_scale =
+        hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                n_embd_head, n_head, n_head_kv, il);
+
+        Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, hparams.rope_type, n_ctx_orig, freq_base, freq_scale,
+                             ext_factor, attn_factor, beta_fast, beta_slow);
+
+        Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, hparams.rope_type, n_ctx_orig, freq_base, freq_scale,
+                             ext_factor, attn_factor, beta_fast, beta_slow);
+
+        cb(Qcur, "Qcur-post-rope", il);
+        cb(Kcur, "Kcur-post-rope", il);
+        cb(Vcur, "Vcur-post-rope", il);
+
+        ggml_tensor * attn_out = build_attn(inp->get_attn(),
+                                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+        cb(attn_out, "attn_out", il);
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        // Mamba2 layer
+        cb(cur, "ssm_in", il);
+
+        ggml_tensor * ssm_out = build_mamba2_layer(inp->get_recr(), cur, model, ubatch, il);
+        cb(ssm_out, "ssm_out", il);
+
+        // // Aggregation
+        cur   = ggml_add(ctx0, attn_out, ssm_out);
+        inpSA = ggml_add(ctx0, cur, inpSA);
+        cb(cur, "layer_out", il);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = inpSA;
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
+                model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, inpSA);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/falcon.cpp

@@ -0,0 +1,161 @@
+#include "models.h"
+
+void llama_model_falcon::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_7B; break;
+        case 60: type = LLM_TYPE_40B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_falcon::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    {
+        output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+        output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+
+        output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+        if (!output) {
+            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // needs to be on GPU
+        }
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.attn_norm_2   = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.attn_norm_2_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_falcon::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_falcon::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * attn_norm;
+
+        attn_norm = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(attn_norm, "attn_norm", il);
+
+        // self-attention
+        {
+            if (model.layers[il].attn_norm_2) {
+                // Falcon-40B
+                cur = build_norm(inpL,
+                        model.layers[il].attn_norm_2,
+                        model.layers[il].attn_norm_2_b,
+                        LLM_NORM, il);
+                cb(cur, "attn_norm_2", il);
+            } else {
+                cur = attn_norm;
+            }
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // using mode = 2 for neox mode
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur       = ggml_get_rows(ctx0,       cur, inp_out_ids);
+            inpL      = ggml_get_rows(ctx0,      inpL, inp_out_ids);
+            attn_norm = ggml_get_rows(ctx0, attn_norm, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = cur;
+
+        // feed forward
+        {
+            cur = build_ffn(attn_norm, // !! use the attn norm, not the result
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    NULL,                      NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cur = ggml_add(ctx0, cur, inpL);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    // norm
+    cur = build_norm(cur,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/gemma-embedding.cpp

@@ -0,0 +1,178 @@
+#include "models.h"
+
+void llama_model_gemma_embedding::load_arch_hparams(llama_model_loader & ml) {
+    hparams.swa_type = LLAMA_SWA_TYPE_SYMMETRIC;
+    uint32_t swa_period = 6;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+    hparams.set_swa_pattern(swa_period);
+
+    hparams.causal_attn = false; // embeddings do not use causal attention
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    //applied only if model converted with --sentence-transformers-dense-modules
+    ml.get_key(LLM_KV_DENSE_2_FEAT_IN, hparams.dense_2_feat_in, false);
+    ml.get_key(LLM_KV_DENSE_2_FEAT_OUT, hparams.dense_2_feat_out, false);
+    ml.get_key(LLM_KV_DENSE_3_FEAT_IN, hparams.dense_3_feat_in, false);
+    ml.get_key(LLM_KV_DENSE_3_FEAT_OUT, hparams.dense_3_feat_out, false);
+
+    GGML_ASSERT((hparams.dense_2_feat_in == 0 || hparams.dense_2_feat_in == hparams.n_embd) && "dense_2_feat_in must be equal to n_embd");
+    GGML_ASSERT((hparams.dense_3_feat_out == 0 || hparams.dense_3_feat_out == hparams.n_embd) && "dense_3_feat_out must be equal to n_embd");
+
+    switch (hparams.n_layer()) {
+        case 24: type = LLM_TYPE_0_3B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+    hparams.f_attention_scale = 1.0f / std::sqrt(float(hparams.n_embd_head_k()));
+
+}
+
+void llama_model_gemma_embedding::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,   "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    // Dense linear weights
+    dense_2_out_layers = create_tensor(tn(LLM_TENSOR_DENSE_2_OUT, "weight"), {n_embd, hparams.dense_2_feat_out}, TENSOR_NOT_REQUIRED);
+    dense_3_out_layers = create_tensor(tn(LLM_TENSOR_DENSE_3_OUT, "weight"), {hparams.dense_3_feat_in, n_embd}, TENSOR_NOT_REQUIRED);
+
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_k_norm    = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM,    "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm    = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma_embedding::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_gemma_embedding::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_k();
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // important: do not normalize weights for raw embeddings input (i.e. encoded image embeddings)
+    inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const float freq_base_l  = model.get_rope_freq_base(cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            // ref: https://github.com/google/gemma_pytorch/blob/014acb7ac4563a5f77c76d7ff98f31b568c16508/gemma/model.py#L315
+            Qcur = ggml_scale(ctx0, Qcur, hparams.f_attention_scale);
+
+            cur =
+                build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f, il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        cur = build_norm(cur, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
+        cb(sa_out, "sa_out", il);
+
+        cur = build_norm(sa_out, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        {
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL, LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = build_norm(cur, model.layers[il].ffn_post_norm, NULL, LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", -1);
+
+        cur = ggml_add(ctx0, cur, sa_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/gemma.cpp

@@ -0,0 +1,139 @@
+#include "models.h"
+
+void llama_model_gemma::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 18: type = LLM_TYPE_2B; break;
+        case 28: type = LLM_TYPE_7B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+   }
+}
+
+void llama_model_gemma::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_gemma::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd_head)));
+            cb(Qcur, "Qcur_scaled", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
+        cb(sa_out, "sa_out", il);
+
+        cur = build_norm(sa_out,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = ggml_add(ctx0, cur, sa_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/gemma2.cpp

@@ -0,0 +1,177 @@
+#include "models.h"
+
+void llama_model_gemma2::load_arch_hparams(llama_model_loader & ml) {
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    hparams.n_swa = 4096; // default value of gemma 2
+    uint32_t swa_period = 2;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+    hparams.set_swa_pattern(swa_period);
+    hparams.attn_soft_cap = true;
+    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,          hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_ATTN_LOGIT_SOFTCAPPING,      hparams.f_attn_logit_softcapping, false);
+    ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING,     hparams.f_final_logit_softcapping, false);
+
+    switch (hparams.n_layer()) {
+        case 26: type = LLM_TYPE_2B; break;
+        case 42: type = LLM_TYPE_9B; break;
+        case 46: type = LLM_TYPE_27B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+   }
+
+    // ref: https://github.com/google/gemma_pytorch/blob/014acb7ac4563a5f77c76d7ff98f31b568c16508/gemma/config.py#L173
+    hparams.f_attention_scale = type == LLM_TYPE_27B
+        ? 1.0f / std::sqrt(float(hparams.n_embd / hparams.n_head(0)))
+        : 1.0f / std::sqrt(float(hparams.n_embd_head_k()));
+}
+
+void llama_model_gemma2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_gemma2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_k();
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv_iswa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const float freq_base_l  = model.get_rope_freq_base (cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            Qcur = ggml_scale(ctx0, Qcur, hparams.f_attention_scale);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
+        cb(sa_out, "sa_out", il);
+
+        cur = build_norm(sa_out,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, NULL,
+                LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", -1);
+
+        cur = ggml_add(ctx0, cur, sa_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    // final logit soft-capping
+    cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_final_logit_softcapping);
+    cur = ggml_tanh(ctx0, cur);
+    cur = ggml_scale(ctx0, cur, hparams.f_final_logit_softcapping);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/gemma3.cpp

@@ -0,0 +1,225 @@
+#include "models.h"
+
+void llama_model_gemma3::load_arch_hparams(llama_model_loader & ml) {
+    const bool found_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false);
+    if (found_swa && hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+        uint32_t swa_period = 6;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period);
+
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    } else {
+        hparams.swa_type = LLAMA_SWA_TYPE_NONE;
+    }
+
+    hparams.f_final_logit_softcapping = 0.0f;
+    ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING, hparams.f_final_logit_softcapping, false);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 18: type = LLM_TYPE_270M; break;
+        case 26: type = LLM_TYPE_1B; break;
+        case 32: type = LLM_TYPE_8B; break; // Rnj-1
+        case 34: type = LLM_TYPE_4B; break;
+        case 48: type = LLM_TYPE_12B; break;
+        case 62: type = LLM_TYPE_27B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+
+    // ref: https://github.com/google/gemma_pytorch/blob/014acb7ac4563a5f77c76d7ff98f31b568c16508/gemma/config.py#L289
+    hparams.f_attention_scale = type == LLM_TYPE_27B
+        ? 1.0f / std::sqrt(float(hparams.n_embd / hparams.n_head(0)))
+        : 1.0f / std::sqrt(float(hparams.n_embd_head_k()));
+}
+
+void llama_model_gemma3::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,   "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    // Dense linear weights
+    dense_2_out_layers = create_tensor(tn(LLM_TENSOR_DENSE_2_OUT, "weight"), {n_embd, hparams.dense_2_feat_out}, TENSOR_NOT_REQUIRED);
+    dense_3_out_layers = create_tensor(tn(LLM_TENSOR_DENSE_3_OUT, "weight"), {hparams.dense_3_feat_in, n_embd}, TENSOR_NOT_REQUIRED);
+
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_k_norm    = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM,    "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm    = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma3::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_STANDARD) {
+        return std::make_unique<graph<true>>(*this, params);
+    } else {
+        return std::make_unique<graph<false>>(*this, params);
+    }
+}
+
+template <bool iswa>
+llama_model_gemma3::graph<iswa>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_k();
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // important: do not normalize weights for raw embeddings input (i.e. encoded image embeddings)
+    inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // TODO: is causal == true correct? might need some changes
+    using inp_attn_type = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa, llm_graph_input_attn_kv>;
+    inp_attn_type * inp_attn = nullptr;
+
+    if constexpr (iswa) {
+        inp_attn = build_attn_inp_kv_iswa();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        float freq_base_l  = 0.0f;
+        float freq_scale_l = 0.0f;
+
+        if constexpr (iswa) {
+            freq_base_l  = model.get_rope_freq_base (cparams, il);
+            freq_scale_l = model.get_rope_freq_scale(cparams, il);
+        } else {
+            freq_base_l  = freq_base;
+            freq_scale_l = freq_scale;
+        }
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            // ref: https://github.com/google/gemma_pytorch/blob/014acb7ac4563a5f77c76d7ff98f31b568c16508/gemma/model.py#L315
+            Qcur = ggml_scale(ctx0, Qcur, hparams.f_attention_scale);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
+        cb(sa_out, "sa_out", il);
+
+        cur = build_norm(sa_out,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        {
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, NULL,
+                LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", il);
+
+        cur = ggml_add(ctx0, cur, sa_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    if (hparams.f_final_logit_softcapping) {
+        cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_final_logit_softcapping);
+        cur = ggml_tanh(ctx0, cur);
+        cur = ggml_scale(ctx0, cur, hparams.f_final_logit_softcapping);
+    }
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+template struct llama_model_gemma3::graph<false>;
+template struct llama_model_gemma3::graph<true>;

src/models/gemma3n.cpp

@@ -0,0 +1,459 @@
+#include "models.h"
+
+void llama_model_gemma3n::load_arch_hparams(llama_model_loader & ml) {
+    uint32_t swa_period = 5;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    hparams.set_swa_pattern(swa_period);
+
+    hparams.n_layer_kv_from_start = 20;
+    hparams.f_attention_scale     = 1.0f;
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,          hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 30: type = LLM_TYPE_E2B; break;
+        case 35: type = LLM_TYPE_E4B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_gemma3n::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_altup      = hparams.n_altup;
+    const int64_t laurel_rank  = hparams.laurel_rank;
+    const int64_t n_embd_altup = hparams.n_embd_altup;
+
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    altup_proj        = create_tensor(tn(LLM_TENSOR_ALTUP_PROJ,        "weight"), {n_embd, n_embd, n_altup - 1}, 0);
+    altup_unembd_proj = create_tensor(tn(LLM_TENSOR_ALTUP_UNEMBD_PROJ, "weight"), {n_embd, n_embd, n_altup - 1}, 0);
+
+    per_layer_tok_embd   = create_tensor(tn(LLM_TENSOR_PER_LAYER_TOKEN_EMBD, "weight"), {n_embd_altup * n_layer, n_vocab}, 0);
+    per_layer_model_proj = create_tensor(tn(LLM_TENSOR_PER_LAYER_MODEL_PROJ, "weight", 0), {n_embd, n_embd_altup * n_layer}, 0);
+    per_layer_proj_norm  = create_tensor(tn(LLM_TENSOR_PER_LAYER_PROJ_NORM,  "weight", 0), {n_embd_altup}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_q_norm    = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), {n_embd_head_k}, 0);
+        layer.attn_k_norm    = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM,    "weight", i), {n_embd_head_k}, 0);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        // altup & laurel
+        layer.per_layer_inp_gate   = create_tensor(tn(LLM_TENSOR_PER_LAYER_INP_GATE,  "weight", i), {n_embd, n_embd_altup}, 0);
+        layer.per_layer_proj       = create_tensor(tn(LLM_TENSOR_PER_LAYER_PROJ,      "weight", i), {n_embd_altup, n_embd}, 0);
+        layer.per_layer_post_norm  = create_tensor(tn(LLM_TENSOR_PER_LAYER_POST_NORM, "weight", i), {n_embd}, 0);
+        layer.altup_correct_coef   = create_tensor(tn(LLM_TENSOR_ALTUP_CORRECT_COEF,  "weight", i), {n_altup, n_altup}, 0);
+        layer.altup_correct_scale  = create_tensor(tn(LLM_TENSOR_ALTUP_CORRECT_SCALE, "weight", i), {n_embd}, 0);
+        layer.altup_predict_coef   = create_tensor(tn(LLM_TENSOR_ALTUP_PREDICT_COEF,  "weight", i), {n_altup, n_altup * n_altup}, 0);
+        layer.altup_router         = create_tensor(tn(LLM_TENSOR_ALTUP_ROUTER,        "weight", i), {n_embd, n_altup}, 0);
+        layer.altup_router_norm    = create_tensor(tn(LLM_TENSOR_ALTUP_ROUTER_NORM,   "weight", i), {n_embd}, 0);
+        layer.laurel_l             = create_tensor(tn(LLM_TENSOR_LAUREL_L,            "weight", i), {n_embd, laurel_rank}, 0);
+        layer.laurel_r             = create_tensor(tn(LLM_TENSOR_LAUREL_R,            "weight", i), {laurel_rank, n_embd}, 0);
+        layer.laurel_post_norm     = create_tensor(tn(LLM_TENSOR_LAUREL_POST_NORM,    "weight", i), {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma3n::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+// get 2D slice view from a 3D tensor, the idx corresponds to the 3rd dim
+static ggml_tensor * ggml_view_2d_slice(ggml_context * ctx0, ggml_tensor * x, int idx) {
+    GGML_ASSERT(idx < (int) x->ne[2]);
+    return ggml_view_2d(ctx0, x, x->ne[0], x->ne[1], ggml_row_size(x->type, x->ne[0]),
+                        idx * x->ne[0] * x->ne[1] * ggml_element_size(x));
+}
+
+llama_model_gemma3n::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params),
+    model(model),
+    n_embd_head(model.hparams.n_embd_head_k()),
+    n_embd_altup(model.hparams.n_embd_altup),
+    n_altup(model.hparams.n_altup),
+    i_altup_act(model.hparams.i_altup_act) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // important: do not normalize weights for raw embeddings input (i.e. encoded image embeddings)
+    inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // TODO: is causal == true correct? might need some changes
+    auto * inp_attn = build_attn_inp_kv_iswa();
+
+    ggml_tensor * inp_per_layer = build_inp_per_layer();
+    ggml_build_forward_expand(gf, inp_per_layer);
+
+    // inp_per_layer now has shape: [n_embd_altup, n_tokens, n_layer]
+    inp_per_layer = project_per_layer_inputs(inpL, inp_per_layer);
+
+    // inpL now has only 1 altup, project it to the rest of the altups
+    // these "added" altups will be concat to the last dim of inpL
+    {
+        ggml_tensor * target_magnitude = calc_magnitude(inpL);
+        ggml_tensor * inp_repeated     = ggml_repeat_4d(ctx0, inpL, n_embd, n_tokens, n_altup - 1, 1);
+        ggml_tensor * altup_added =
+            ggml_mul_mat(ctx0, model.altup_proj, inp_repeated);  // shape: [n_embd, n_tokens, n_altup - 1]
+        ggml_tensor * new_magnitude = calc_magnitude(altup_added);
+        altup_added                 = ggml_div(ctx0, ggml_mul(ctx0, altup_added, target_magnitude), new_magnitude);
+        inpL                        = ggml_concat(ctx0, inpL, altup_added, 2);  // shape: [n_embd, n_tokens, n_altup]
+        cb(inpL, "inp_stacked", -1);
+    }
+    // inpL now has shape: [n_embd, n_tokens, n_altup]
+
+    for (int il = 0; il < n_layer; ++il) {
+        // this block is made to be closely resemble Gemma3p5DecoderLayer on python code
+        const float freq_base_l  = model.get_rope_freq_base(cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        ggml_tensor * cur         = inpL;                    // [n_embd, n_tokens, n_altup]
+        ggml_tensor * predictions = altup_predict(cur, il);  // [n_embd, n_tokens, n_altup]
+
+        // predicted value will go through self-attention and laurel
+        ggml_tensor * active_prediction = ggml_view_2d_slice(ctx0, predictions, i_altup_act);  // [n_embd, n_tokens]
+        cur = active_prediction;
+        cb(cur, "active_prediction", il);
+
+        // norm
+        cur = build_norm(cur, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // laurel
+        ggml_tensor * laurel_out = laurel(cur, il);  // [n_embd, n_tokens]
+
+        // self-attention
+        if (hparams.has_kv(il)) {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur, n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            Vcur = ggml_rms_norm(ctx0, Vcur, hparams.f_norm_rms_eps);
+
+            cb(Qcur, "Qcur_normed", il);
+            cb(Kcur, "Kcur_normed", il);
+            cb(Vcur, "Vcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur_pos", il);
+            cb(Kcur, "Kcur_pos", il);
+
+            cur = build_attn(inp_attn, model.layers[il].wo,
+                    NULL, model.layers[il].wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr,
+                    hparams.f_attention_scale, il);
+        } else {
+            // reuse KV cache of earlier layers
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            cb(Qcur, "Qcur", il);
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(Qcur, "Qcur_pos", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, nullptr, nullptr, nullptr, nullptr, nullptr, hparams.f_attention_scale, il);
+        }
+        cur = build_norm(cur, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        cur = ggml_add(ctx0, cur, active_prediction);  // [n_embd, n_tokens]
+        cb(cur, "attn_gated", il);
+
+        ggml_tensor * attn_laurel = ggml_scale(ctx0, ggml_add(ctx0, cur, laurel_out),
+                                               1.0f / sqrtf(2.0f));  // [n_embd, n_tokens]
+        cb(attn_laurel, "attn_laurel", il);
+
+        cur = build_norm(attn_laurel, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        {
+            ggml_tensor * up_proj   = build_lora_mm(model.layers[il].ffn_up, cur);
+            ggml_tensor * gate_proj = build_lora_mm(model.layers[il].ffn_gate, cur);
+
+            if (il < n_layer_sparsity) {
+                // apply activation sparsity
+                gate_proj = gaussian_topk(gate_proj);
+            }
+            gate_proj = ggml_gelu(ctx0, gate_proj);
+
+            cur = ggml_mul(ctx0, up_proj, gate_proj);
+            cur = build_lora_mm(model.layers[il].ffn_down, cur);
+            cb(cur, "ffn_out", il);
+        }
+        cur = build_norm(cur, model.layers[il].ffn_post_norm, NULL, LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", il);
+
+        ggml_tensor * attn_ffw_laurel_gated = ggml_add(ctx0, cur, attn_laurel);  // [n_embd, n_tokens]
+        cb(attn_ffw_laurel_gated, "attn_ffw_laurel_gated", il);
+
+        ggml_tensor * corrected = altup_correct(predictions, attn_ffw_laurel_gated, il);  // [n_embd, n_tokens, n_altup]
+
+        ggml_tensor * first_prediction;                                                   // [n_embd, n_tokens]
+        {
+            first_prediction = ggml_view_2d_slice(ctx0, corrected, i_altup_act);          // [n_embd, n_tokens]
+            first_prediction = ggml_mul(ctx0, first_prediction, model.layers[il].altup_correct_scale);
+            first_prediction = build_lora_mm(model.layers[il].per_layer_inp_gate, first_prediction);
+            first_prediction = ggml_gelu(ctx0, first_prediction);                 // [n_embd_altup, n_tokens]
+            cb(first_prediction, "first_prediction_gated", il);
+
+            ggml_tensor * inp_this_layer = ggml_view_2d_slice(ctx0, inp_per_layer, il);   // [n_embd_altup, n_tokens]
+            first_prediction = ggml_mul(ctx0, first_prediction, inp_this_layer);  // [n_embd_altup, n_tokens]
+            cb(first_prediction, "first_prediction_scaled", il);
+
+            first_prediction = build_lora_mm(model.layers[il].per_layer_proj, first_prediction);  // [n_embd, n_tokens]
+            first_prediction =
+                build_norm(first_prediction, model.layers[il].per_layer_post_norm, NULL, LLM_NORM_RMS, il);
+            cb(first_prediction, "first_prediction_out", il);
+        }
+        // equivalent to python code: corrected_predictions[1:] += first_prediction
+        {
+            ggml_tensor * slice_first = ggml_view_2d_slice(ctx0, corrected, 0);
+            ggml_tensor * slice_rest  = ggml_view_3d(
+                ctx0, corrected, n_embd, n_tokens, n_altup - 1, ggml_row_size(corrected->type, n_embd),
+                ggml_row_size(corrected->type, n_embd * n_tokens), n_embd * n_tokens * ggml_element_size(corrected));
+            ggml_tensor * tmp = ggml_add(ctx0, slice_rest, first_prediction);  // [n_embd, n_tokens, n_altup - 1]
+            corrected         = ggml_concat(ctx0, slice_first, tmp, 2);        // [n_embd, n_tokens, n_altup]
+        }
+        cur = corrected;                                                       // [n_embd, n_tokens, n_altup]
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;  // [n_embd, n_tokens, n_altup]
+
+    // cur now has multiple altup(s), we want to merge them back to 1 altup
+    {
+        ggml_tensor * target_magnitude = calc_magnitude(ggml_view_2d_slice(ctx0, cur, i_altup_act));  // [n_embd, n_tokens]
+        // do a view to skip the first slice (active altup)
+        ggml_tensor * alt_slice =
+            ggml_view_3d(ctx0, cur, n_embd, n_tokens, n_altup - 1, ggml_row_size(cur->type, n_embd),
+                         ggml_row_size(cur->type, n_embd * n_tokens), n_embd * n_tokens * ggml_element_size(cur));
+        ggml_tensor * altup_unembd =
+            ggml_mul_mat(ctx0, model.altup_unembd_proj, alt_slice);  // shape: [n_embd, n_tokens, n_altup - 1]
+        ggml_tensor * new_magnitude = calc_magnitude(altup_unembd);
+        altup_unembd                = ggml_div(ctx0, ggml_mul(ctx0, altup_unembd, target_magnitude), new_magnitude);
+        cb(altup_unembd, "altup_unembd", -1);
+
+        // equivalent to torch.mean(hidden_states, dim=0)
+        cur = ggml_view_2d_slice(ctx0, cur, 0);  // [n_embd, n_tokens]
+        for (int i = 0; i < n_altup - 1; ++i) {
+            cur = ggml_add(ctx0, cur, ggml_view_2d_slice(ctx0, altup_unembd, i));
+        }
+        cur = ggml_scale(ctx0, cur, 1.0f / float(n_altup));  // [n_embd, n_tokens]
+        cb(cur, "unembd_merged", -1);
+    }
+    // cur now has shape: [n_embd, n_tokens]
+
+    // TODO: move this to right after the last KV layer
+    {
+        // skip computing output for unused tokens
+        ggml_tensor * inp_out_ids = build_inp_out_ids();
+        cur                       = ggml_get_rows(ctx0, cur, inp_out_ids);
+    }
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    {
+        // final logit soft-capping
+        cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_final_logit_softcapping);
+        cur = ggml_tanh(ctx0, cur);
+        cur = ggml_scale(ctx0, cur, hparams.f_final_logit_softcapping);
+    }
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+ggml_tensor * llama_model_gemma3n::graph::calc_magnitude(ggml_tensor * x) {
+    return ggml_sqrt(ctx0, ggml_sum_rows(ctx0, ggml_sqr(ctx0, x)));
+}
+
+// equivalent to get_per_layer_inputs() in python code
+// output shape: [n_embd_altup, n_layer, n_tokens]
+ggml_tensor * llama_model_gemma3n::graph::build_inp_per_layer() {
+    auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
+    ggml_tensor * inp_per_layer;
+    float tok_embd_scale = sqrtf((float) n_embd_altup);
+    if (ubatch.token) {
+        inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
+        ggml_set_input(inp->tokens);
+        res->t_inp_tokens = inp->tokens;
+        inp_per_layer = ggml_get_rows  (ctx0, model.per_layer_tok_embd, inp->tokens);
+        inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_altup, n_layer, n_tokens);
+        inp_per_layer = ggml_scale     (ctx0, inp_per_layer, tok_embd_scale);
+        cb(inp_per_layer, "inp_per_layer_selected", -1);
+        res->add_input(std::move(inp));
+    } else {
+        // Multimodal embedding path: use padding token (ID=0) embedding
+        // TODO: verify if this is the correct behavior in transformers implementation
+        const int64_t embd_size = model.per_layer_tok_embd->ne[0];  // n_embd_altup * n_layer
+
+        // Extract and dequantize padding token embedding (row 0)
+        ggml_tensor * padding = ggml_view_1d(ctx0, model.per_layer_tok_embd, embd_size, 0);
+        inp_per_layer = ggml_cast (ctx0, padding, GGML_TYPE_F32);
+        inp_per_layer = ggml_scale(ctx0, inp_per_layer, tok_embd_scale);
+
+        // Reshape to [n_embd_altup, n_layer, 1]
+        inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_altup, n_layer, 1);
+        cb(inp_per_layer, "inp_per_layer_multimodal", -1);
+    }
+    return inp_per_layer;
+}
+
+// equivalent to project_per_layer_inputs() in python code
+// this calculates the per-layer inputs, so the final tensor shape will have n_layer as the last dim
+// output shape: [n_embd_altup, n_tokens, n_layer]
+ggml_tensor * llama_model_gemma3n::graph::project_per_layer_inputs(ggml_tensor * inp_batch, ggml_tensor * inp_per_layer) {
+    const float per_layer_projection_scale = 1.0f / sqrtf((float) n_embd);
+    const float per_layer_input_scale      = 1.0f / sqrtf(2.0f);
+
+    ggml_tensor * per_layer_proj;
+    per_layer_proj = ggml_mul_mat   (ctx0, model.per_layer_model_proj, inp_batch);
+    per_layer_proj = ggml_scale     (ctx0, per_layer_proj, per_layer_projection_scale);
+    per_layer_proj = ggml_reshape_3d(ctx0, per_layer_proj, n_embd_altup, n_layer, n_tokens);
+
+    per_layer_proj = build_norm(per_layer_proj, model.per_layer_proj_norm, NULL, LLM_NORM_RMS, -1);
+    cb(per_layer_proj, "per_layer_proj", -1);
+
+    inp_per_layer = ggml_add  (ctx0, per_layer_proj, inp_per_layer);
+    inp_per_layer = ggml_scale(ctx0, inp_per_layer, per_layer_input_scale);
+    cb(inp_per_layer, "inp_per_layer", -1);
+
+    // permute to shape: [n_embd_altup, n_tokens, n_layer]
+    inp_per_layer = ggml_cont(ctx0, ggml_permute(ctx0, inp_per_layer, 0, 2, 1, 3));
+    return inp_per_layer;
+}
+
+// input cur shape: [n_altup, n_tokens]
+// output    shape: [n_altup, n_tokens]
+ggml_tensor * llama_model_gemma3n::graph::laurel(ggml_tensor * cur, int il) {
+    ggml_tensor * tmp = cur;
+    tmp               = build_lora_mm(model.layers[il].laurel_l, tmp);
+    tmp               = build_lora_mm(model.layers[il].laurel_r, tmp);
+    tmp               = build_norm(tmp, model.layers[il].laurel_post_norm, NULL, LLM_NORM_RMS, il);
+    tmp               = ggml_add(ctx0, tmp, cur);
+    cb(tmp, "laurel_out", il);
+    return tmp;
+}
+
+// input x shape: [n_embd, n_tokens]
+// output  shape: [n_embd, n_tokens]
+ggml_tensor * llama_model_gemma3n::graph::gaussian_topk(ggml_tensor * x) {
+    ggml_tensor * mean = ggml_mean(ctx0, x);
+    ggml_tensor * std  = ggml_sqrt(ctx0, ggml_scale(ctx0, ggml_sum_rows(ctx0, ggml_sqr(ctx0, ggml_sub(ctx0, x, mean))),
+                                                    1.0f / (float) (x->ne[0] - 1)));
+    ggml_tensor * cutoff_x = ggml_add(ctx0, mean, ggml_scale(ctx0, std, f_sparsity_std_mul));
+    return ggml_relu(ctx0, ggml_sub(ctx0, x, cutoff_x));
+}
+
+//
+// altup functions
+//
+
+// equivalent to compute_router_modalities() in python code
+// input x shape: [n_embd,  n_tokens]
+// output  shape: [n_altup, n_tokens]
+ggml_tensor * llama_model_gemma3n::graph::altup_compute_router_modalities(ggml_tensor * x, int il) {
+    ggml_tensor * router_inputs = build_norm(x, model.layers[il].altup_router_norm, NULL, LLM_NORM_RMS, il);
+
+    // router_input_scale
+    router_inputs = ggml_scale(ctx0, router_inputs, 1.0f / (float) n_embd);
+
+    ggml_tensor * output = ggml_mul_mat(ctx0, model.layers[il].altup_router, router_inputs);
+    return ggml_tanh(ctx0, output);  // [n_altup, n_tokens]
+}
+
+// input cur shape: [n_embd, n_tokens, n_altup]
+// output    shape: [n_embd, n_tokens, n_altup]
+ggml_tensor * llama_model_gemma3n::graph::altup_predict(ggml_tensor * cur, int il) {
+    ggml_tensor * activated  = ggml_view_2d_slice(ctx0, cur, i_altup_act);      // [n_embd, n_tokens]
+    ggml_tensor * modalities = altup_compute_router_modalities(activated, il);  // [n_altup, n_tokens]
+    cb(modalities, "modalities", il);
+
+    ggml_tensor * all_coefs = build_lora_mm(model.layers[il].altup_predict_coef, modalities);
+    cb(all_coefs, "all_coefs", il);
+    // first dim now having n_altup^2 elements, we reshape it to 2D (so we end up with 3D tensor)
+    all_coefs = ggml_reshape_3d(ctx0, all_coefs, n_altup, n_altup, n_tokens);
+
+    // permute to [n_altup, n_embd, n_tokens]
+    ggml_tensor * cur_permuted = ggml_cont(ctx0, ggml_permute(ctx0, cur, 1, 2, 0, 3));
+    ggml_tensor * predictions  = ggml_mul_mat(ctx0, cur_permuted, all_coefs);  // [n_altup, n_embd, n_tokens]
+
+    // final shape must be the same as cur: [n_embd, n_tokens, n_altup]
+    predictions = ggml_cont(ctx0, ggml_permute(ctx0, predictions, 0, 2, 1, 3));
+    predictions = ggml_add(ctx0, predictions, cur);
+    cb(predictions, "predictions", il);
+
+    return predictions;
+}
+
+// input predictions       shape: [n_embd, n_tokens, n_altup]
+// input activated         shape: [n_embd, n_tokens]
+// output                  shape: [n_embd, n_tokens, n_altup]
+ggml_tensor * llama_model_gemma3n::graph::altup_correct(ggml_tensor * predictions, ggml_tensor * activated, int il) {
+    ggml_tensor * modalities = altup_compute_router_modalities(activated, il);  // [n_altup, n_tokens]
+    cb(modalities, "modalities", il);
+
+    ggml_tensor * active_prediction = ggml_view_2d_slice(ctx0, predictions, i_altup_act);
+    ggml_tensor * innovation        = ggml_sub(ctx0, activated, active_prediction);  // [n_embd, n_tokens]
+    cb(innovation, "innovation", il);
+
+    ggml_tensor * all_coefs = build_lora_mm(model.layers[il].altup_correct_coef, modalities);  // [n_altup, n_tokens]
+    all_coefs               = ggml_scale_bias(ctx0, all_coefs, 1.0f, 1.0f);                    // + 1.0
+    cb(all_coefs, "all_coefs", il);
+    all_coefs = ggml_transpose(ctx0, all_coefs);                                               // [n_tokens, n_altup]
+    all_coefs = ggml_cont_3d(ctx0, all_coefs, 1, n_tokens, n_altup);                           // [1, n_tokens, n_altup]
+
+    innovation              = ggml_repeat_4d(ctx0, innovation, n_embd, n_tokens, n_altup, 1);
+    ggml_tensor * corrected = ggml_mul(ctx0, innovation, all_coefs);   // [n_embd, n_tokens, n_altup]
+    corrected               = ggml_add(ctx0, corrected, predictions);  // [n_embd, n_tokens, n_altup]
+    cb(corrected, "corrected", il);
+
+    return corrected;
+}

src/models/gemma4-assistant.cpp

@@ -0,0 +1,203 @@
+#include "models.h"
+
+void llama_model_gemma4_assistant::load_arch_hparams(llama_model_loader & ml) {
+    hparams.n_embd_inp_impl = hparams.n_embd_out();
+
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+
+    uint32_t n_kv_shared_layers = 0;
+    ml.get_key(LLM_KV_ATTENTION_SHARED_KV_LAYERS, n_kv_shared_layers, false);
+
+    hparams.f_attention_scale = 1.0f;
+
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn == hparams.n_layer_all && "n_layer_nextn must be == n_layer_impl");
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,           hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,     hparams.n_swa);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,  hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH_SWA,     hparams.n_embd_head_k_swa);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_SWA,   hparams.n_embd_head_v_swa);
+}
+
+void llama_model_gemma4_assistant::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_embd_head_k != n_embd_head_v) {
+        throw std::runtime_error("Gemma 4 assistant requires n_embd_head_k == n_embd_head_v");
+    }
+    if (hparams.n_embd_head_k_swa != hparams.n_embd_head_v_swa) {
+        throw std::runtime_error("Gemma 4 assistant requires n_embd_head_k_swa == n_embd_head_v_swa");
+    }
+    if (hparams.n_embd_out() == n_embd) {
+        throw std::runtime_error("Gemma 4 assistant requires embedding_length_out to carry the target hidden size");
+    }
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+    output   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+
+    create_tensor(tn(LLM_TENSOR_MASKED_EMBD_CENTROIDS, "weight"), {}, TENSOR_NOT_REQUIRED);
+    create_tensor(tn(LLM_TENSOR_MASKED_EMBD_ORDERING),  {}, TENSOR_NOT_REQUIRED);
+
+    const int64_t n_embd_backbone = hparams.n_embd_inp();
+    nextn_proj_post = create_tensor(tn(LLM_TENSOR_NEXTN_PROJ_POST, "weight"), { n_embd, n_embd_backbone }, 0);
+
+    int rope_freqs_flag = 0;
+
+    for (int i = 0; i < n_layer_nextn; ++i) {
+        auto & layer = layers[i];
+
+        const int64_t n_head      = hparams.n_head(i);
+        const int64_t n_embd_head = hparams.n_embd_head_k(i);
+        const int64_t n_ff        = hparams.n_ff(i);
+
+        if (i == 0) {
+            nextn_proj_pre = create_tensor(tn(LLM_TENSOR_NEXTN_PROJ_PRE, "weight", i), { 2*n_embd_backbone, n_embd }, 0);
+        }
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+        layer.wq        = create_tensor(tn(LLM_TENSOR_ATTN_Q,    "weight", i), { n_embd, n_embd_head*n_head }, 0);
+        layer.wo        = create_tensor(tn(LLM_TENSOR_ATTN_OUT,  "weight", i), { n_embd_head*n_head, n_embd }, 0);
+
+        layer.attn_q_norm    = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), { n_embd_head }, 0);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), { n_embd }, 0);
+
+        layer.out_scale = create_tensor(tn(LLM_TENSOR_LAYER_OUT_SCALE, "weight", i), { 1u }, 0);
+
+        if (!hparams.is_swa(i)) {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), { n_embd_head/2 }, rope_freqs_flag);
+            rope_freqs_flag = TENSOR_DUPLICATED;
+        }
+
+        layer.ffn_norm      = create_tensor(tn(LLM_TENSOR_FFN_NORM,      "weight", i), { n_embd }, 0);
+        layer.ffn_gate      = create_tensor(tn(LLM_TENSOR_FFN_GATE,      "weight", i), { n_embd, n_ff }, 0);
+        layer.ffn_up        = create_tensor(tn(LLM_TENSOR_FFN_UP,        "weight", i), { n_embd, n_ff }, 0);
+        layer.ffn_down      = create_tensor(tn(LLM_TENSOR_FFN_DOWN,      "weight", i), { n_ff, n_embd }, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), { n_embd }, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma4_assistant::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_gemma4_assistant::graph::graph(const llama_model & model, const llm_graph_params & params) :
+        llm_graph_context(params) {
+    const int64_t n_embd_backbone = hparams.n_embd_inp();
+
+    ggml_tensor * inp_tokens;
+    ggml_tensor * inp_h;
+    {
+        auto inp = std::make_unique<llm_graph_input_embd>(n_embd_backbone);
+
+        inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
+        cb(inp->tokens, "inp_tokens", -1);
+        ggml_set_input(inp->tokens);
+        inp_tokens = inp->tokens;
+        res->t_inp_tokens = inp->tokens;
+
+        inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd_backbone, ubatch.n_tokens);
+        cb(inp->embd, "inp_h", -1);
+        ggml_set_input(inp->embd);
+        inp_h = inp->embd;
+        res->t_inp_embd = inp->embd;
+
+        res->add_input(std::move(inp));
+    }
+
+    GGML_ASSERT(cparams.ctx_other != nullptr);
+    const auto * model_other = llama_get_model(cparams.ctx_other);
+
+    ggml_tensor * x = ggml_get_rows(ctx0, model_other->tok_embd, inp_tokens);
+    x = ggml_scale(ctx0, x, sqrtf((float) n_embd_backbone));
+    cb(x, "inp_embd_target", -1);
+
+    ggml_tensor * xh = ggml_concat(ctx0, x, inp_h, 0);
+    cb(xh, "inp_xh", -1);
+
+    ggml_tensor * cur = ggml_mul_mat(ctx0, model.nextn_proj_pre, xh);
+    cb(cur, "pre_proj", -1);
+
+    auto *        inp_attn    = build_attn_inp_kv_iswa();
+    ggml_tensor * inp_pos     = build_inp_pos();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    ggml_tensor * inpL = cur;
+
+    for (int il = 0; il < n_layer_nextn; ++il) {
+        const bool is_swa = hparams.is_swa(il);
+
+        const int64_t n_embd_head = hparams.n_embd_head_k(il);
+        const int64_t n_head      = hparams.n_head(il);
+
+        const float freq_base_l  = model.get_rope_freq_base(cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+        const int   n_rot_l      = hparams.n_rot(il);
+
+        ggml_tensor * cur_norm = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur_norm, "attn_norm", il);
+
+        ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur_norm);
+        Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+        Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, nullptr, LLM_NORM_RMS, il);
+        cb(Qcur, "Qcur_normed", il);
+
+        ggml_tensor * freq_factors = is_swa ? nullptr : model.layers[il].rope_freqs;
+        Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, freq_factors, n_rot_l, rope_type, n_ctx_orig,
+                             freq_base_l, freq_scale_l, ext_factor, attn_factor, beta_fast, beta_slow);
+        cb(Qcur, "Qcur_pos", il);
+
+        cur = build_attn(inp_attn, model.layers[il].wo, nullptr, nullptr,
+                Qcur, nullptr, nullptr, nullptr, nullptr, nullptr, hparams.f_attention_scale, il);
+
+        if (il == n_layer_nextn - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0, cur,  inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        cur = build_norm(cur, model.layers[il].attn_post_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * attn_out = ggml_add(ctx0, cur, inpL);
+        cb(attn_out, "attn_out", il);
+
+        cur = build_norm(attn_out, model.layers[il].ffn_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   nullptr, nullptr,
+                model.layers[il].ffn_gate, nullptr, nullptr,
+                model.layers[il].ffn_down, nullptr, nullptr,
+                nullptr,
+                LLM_FFN_GELU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = build_norm(cur, model.layers[il].ffn_post_norm, nullptr, LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", il);
+
+        cur = ggml_add(ctx0, cur, attn_out);
+
+        cur = ggml_mul(ctx0, cur, model.layers[il].out_scale);
+        cb(cur, "out_scaled", il);
+
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, nullptr, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+
+    ggml_tensor * logits = build_lora_mm(model.output, cur);
+    cb(logits, "result_output", -1);
+    res->t_logits = logits;
+
+    ggml_tensor * h_next = ggml_mul_mat(ctx0, model.nextn_proj_post, cur);
+    cb(h_next, "h_nextn", -1);
+    res->t_h_nextn = h_next;
+
+    ggml_build_forward_expand(gf, logits);
+    ggml_build_forward_expand(gf, h_next);
+}

src/models/gemma4.cpp

@@ -0,0 +1,508 @@
+#include "models.h"
+
+void llama_model_gemma4::load_arch_hparams(llama_model_loader & ml) {
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+
+    uint32_t n_kv_shared_layers = 0;
+    ml.get_key(LLM_KV_ATTENTION_SHARED_KV_LAYERS, n_kv_shared_layers, false);
+
+    hparams.n_layer_kv_from_start = hparams.n_layer_all - (int32_t)n_kv_shared_layers;
+    hparams.f_attention_scale     = 1.0f; // Gemma4 uses self.scaling = 1.0 (no pre-attn scaling)
+
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,          hparams.rope_freq_base_train_swa, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp, false);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EMBEDDING_LENGTH_PER_LAYER,  hparams.n_embd_per_layer);
+    ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH_SWA,    hparams.n_embd_head_k_swa);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_SWA,  hparams.n_embd_head_v_swa);
+    ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING,     hparams.f_final_logit_softcapping, false);
+
+    switch (hparams.n_layer()) {
+        case 30: type = LLM_TYPE_26B_A4B; break;
+        case 35: type = LLM_TYPE_E2B; break;
+        case 42: type = LLM_TYPE_E4B; break;
+        case 60: type = LLM_TYPE_31B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_gemma4::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const uint32_t n_embd_per_layer = hparams.n_embd_per_layer;
+    const int64_t  n_ff_exp         = hparams.n_ff_exp;
+
+    if (n_embd_head_k != n_embd_head_v) {
+        throw std::runtime_error("Gemma 4 requires n_embd_head_k == n_embd_head_v");
+    }
+    if (hparams.n_embd_head_k_swa != hparams.n_embd_head_v_swa) {
+        throw std::runtime_error("Gemma 4 requires n_embd_head_k_swa == n_embd_head_v_swa");
+    }
+
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    if (n_embd_per_layer > 0) {
+        per_layer_tok_embd   = create_tensor(tn(LLM_TENSOR_PER_LAYER_TOKEN_EMBD, "weight"),    {n_embd_per_layer * n_layer, n_vocab}, 0);
+        per_layer_model_proj = create_tensor(tn(LLM_TENSOR_PER_LAYER_MODEL_PROJ, "weight", 0), {n_embd, n_embd_per_layer * n_layer}, 0);
+        per_layer_proj_norm  = create_tensor(tn(LLM_TENSOR_PER_LAYER_PROJ_NORM,  "weight", 0), {n_embd_per_layer}, 0);
+    }
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    int rope_freqs_flag = 0;
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+        const int64_t n_head      = hparams.n_head(i);
+        const int64_t n_embd_head = hparams.n_embd_head_k(i);
+        const int64_t n_embd_k    = hparams.n_embd_k_gqa(i);
+        const int64_t n_embd_v    = hparams.n_embd_v_gqa(i);
+        const int     kv_flags    = hparams.has_kv(i) ? 0 : TENSOR_NOT_REQUIRED;
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        // note: use_alternative_attention (v_proj is optional, if it's not present, use k_proj)
+        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head * n_head}, 0);
+        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k}, kv_flags);
+        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v}, TENSOR_NOT_REQUIRED);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head * n_head, n_embd}, 0);
+
+        layer.attn_q_norm    = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), {n_embd_head}, 0);
+        layer.attn_k_norm    = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM,    "weight", i), {n_embd_head}, kv_flags);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        layer.out_scale = create_tensor(tn(LLM_TENSOR_LAYER_OUT_SCALE, "weight", i), {1u}, TENSOR_NOT_REQUIRED);
+
+        if (!hparams.is_swa(i)) {
+            // full_attention layers use rope_freqs for proportional rope
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_embd_head/2}, rope_freqs_flag);
+            rope_freqs_flag = TENSOR_DUPLICATED;
+        }
+
+        // handle use_double_wide_mlp
+        int64_t n_ff_cur = hparams.n_ff(i);
+
+        // for expert layers, we use normal FFN as shared expert (same as python code)
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff_cur}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff_cur}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff_cur, n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        // MoE router
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, TENSOR_NOT_REQUIRED);
+        bool has_expert = layer.ffn_gate_inp != nullptr;
+
+        // norm
+        if (has_expert) {
+            layer.ffn_gate_inp_s = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "scale", i), {n_embd}, 0);
+
+            layer.ffn_pre_norm_2  = create_tensor(tn(LLM_TENSOR_FFN_PRE_NORM_2,  "weight", i), {n_embd}, 0);
+            layer.ffn_post_norm_1 = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM_1, "weight", i), {n_embd}, 0);
+            layer.ffn_post_norm_2 = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM_2, "weight", i), {n_embd}, 0);
+
+            // MoE FFN
+            layer.ffn_gate_up_exps  = create_tensor(tn(LLM_TENSOR_FFN_GATE_UP_EXPS,  "weight", i), {n_embd, n_ff_exp * 2, n_expert}, TENSOR_NOT_REQUIRED);
+
+            if (layer.ffn_gate_up_exps == nullptr) {
+                layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+                layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+            }
+
+            layer.ffn_down_exps     = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS,     "weight", i), {n_ff_exp, n_embd, n_expert}, 0);
+
+            // per-expert scale will be loaded as down_exps_s at the end of the current switch case
+        }
+
+        // per-layer embeddings
+        if (n_embd_per_layer > 0) {
+            layer.per_layer_inp_gate   = create_tensor(tn(LLM_TENSOR_PER_LAYER_INP_GATE,  "weight", i), {n_embd, n_embd_per_layer}, 0);
+            layer.per_layer_proj       = create_tensor(tn(LLM_TENSOR_PER_LAYER_PROJ,      "weight", i), {n_embd_per_layer, n_embd}, 0);
+            layer.per_layer_post_norm  = create_tensor(tn(LLM_TENSOR_PER_LAYER_POST_NORM, "weight", i), {n_embd}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gemma4::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+// get 2D slice view from a 3D tensor, the idx corresponds to the 3rd dim
+static ggml_tensor * ggml_view_2d_slice(ggml_context * ctx0, ggml_tensor * x, int idx) {
+    GGML_ASSERT(idx < (int) x->ne[2]);
+    return ggml_view_2d(ctx0, x, x->ne[0], x->ne[1], ggml_row_size(x->type, x->ne[0]),
+                        idx * x->ne[0] * x->ne[1] * ggml_element_size(x));
+}
+
+// TODO @ngxson : maybe improve this in the future
+class llm_graph_input_logits_bias : public llm_graph_input_i {
+public:
+    llm_graph_input_logits_bias(const llama_vocab & vocab) {
+        arr.resize(vocab.n_tokens(), 0.0f);
+        for (llama_token id : vocab.get_suppress_tokens()) {
+            if (0 <= id && id < (int32_t)vocab.n_tokens()) {
+                arr[id] = -INFINITY;
+            }
+        }
+    }
+    virtual ~llm_graph_input_logits_bias() = default;
+
+    void set_input(const llama_ubatch * /*ubatch*/) override {
+        const int64_t n_vocab = arr.size();
+        ggml_backend_tensor_set(logits_bias, arr.data(), 0, n_vocab*ggml_element_size(logits_bias));
+    }
+
+    bool can_reuse(const llm_graph_params & /*params*/) override {
+        return true;
+    }
+
+    ggml_tensor * logits_bias = nullptr; // F32 [n_vocab]
+
+    std::vector<float> arr;
+};
+
+llama_model_gemma4::graph::graph(const llama_model & model, const llm_graph_params & params) :
+        llm_graph_context(params),
+        model(model),
+        n_embd_per_layer(model.hparams.n_embd_per_layer) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // important: do not normalize weights for raw embeddings input (i.e. encoded image emdeddings)
+    inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // TODO: is causal == true correct? might need some changes
+    auto * inp_attn = build_attn_inp_kv_iswa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    ggml_tensor * inp_per_layer = nullptr;
+    if (model.per_layer_tok_embd) {
+        inp_per_layer = build_inp_per_layer();
+        ggml_build_forward_expand(gf, inp_per_layer);
+
+        // inp_per_layer shape: [n_embd_per_layer, n_tokens, n_layer]
+        inp_per_layer = project_per_layer_inputs(inpL, inp_per_layer);
+    }
+
+    for (int il = 0; il < n_layer; ++il) {
+        const int64_t n_embd_head = hparams.n_embd_head_k(il);
+        GGML_ASSERT(n_embd_head == hparams.n_embd_head_v(il));
+
+        const int64_t n_head    = hparams.n_head(il);
+        const int64_t n_head_kv = hparams.n_head_kv(il);
+
+        const float freq_base_l  = model.get_rope_freq_base(cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+        const int   n_rot_l      = hparams.n_rot(il);
+
+        res->t_layer_inp[il] = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        ggml_tensor * freq_factors = nullptr;
+        if (!hparams.is_swa(il)) {
+            // full_attention layers use rope_freqs for proportional rope
+            freq_factors = model.layers[il].rope_freqs;
+        }
+
+        // Q projection (shared for both non-KV and KV layers)
+        // this is to mirror Gemma4Attention in pytorch code
+        ggml_tensor * Qcur;
+        {
+            Qcur = build_lora_mm(model.layers[il].wq, cur, model.layers[il].wq_s);
+            cb(Qcur, "Qcur", il);
+
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, nullptr, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, freq_factors, n_rot_l, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(Qcur, "Qcur_pos", il);
+        }
+
+        // self-attention
+        if (hparams.has_kv(il)) {
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur, model.layers[il].wk_s);
+            cb(Kcur, "Kcur", il);
+
+            ggml_tensor * Vcur = model.layers[il].wv
+                                    ? build_lora_mm(model.layers[il].wv, cur, model.layers[il].wv_s)
+                                    : Kcur; // if v_proj is not present, use Kcur as Vcur
+            cb(Vcur, "Vcur", il);
+
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, nullptr, LLM_NORM_RMS, il);
+            Vcur = ggml_rms_norm(ctx0, Vcur, hparams.f_norm_rms_eps);
+
+            cb(Kcur, "Kcur_normed", il);
+            cb(Vcur, "Vcur_normed", il);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, freq_factors, n_rot_l, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Kcur, "Kcur_pos", il);
+
+            cur = build_attn(inp_attn, model.layers[il].wo,
+                    nullptr, model.layers[il].wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr,
+                    hparams.f_attention_scale, il);
+        } else {
+            // reuse KV cache of earlier layers
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, nullptr, model.layers[il].wo_s,
+                    Qcur, nullptr, nullptr, nullptr, nullptr, nullptr, hparams.f_attention_scale, il);
+        }
+
+        // TODO @ngxson : strip unused token right after the last KV layer to speed up prompt processing
+        // keep all rows when extracting unmasked nextn embeddings (MTP target needs the hidden state for every token)
+        if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, nullptr,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * attn_out = ggml_add(ctx0, cur, inpL);
+        cb(attn_out, "attn_out", il);
+
+        // feed-forward network
+        const bool is_moe_layer = model.layers[il].ffn_gate_inp != nullptr;
+        if (is_moe_layer) {
+            // MLP (shared exp)
+            ggml_tensor * cur_mlp = build_norm(attn_out,
+                    model.layers[il].ffn_norm, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur_mlp, "ffn_norm_1", il);
+
+            cur_mlp = build_ffn(cur_mlp,
+                    model.layers[il].ffn_up,   nullptr, model.layers[il].ffn_up_s,
+                    model.layers[il].ffn_gate, nullptr, model.layers[il].ffn_gate_s,
+                    model.layers[il].ffn_down, nullptr, model.layers[il].ffn_down_s,
+                    nullptr,
+                    LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cur_mlp = build_norm(cur_mlp,
+                    model.layers[il].ffn_post_norm_1, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur_mlp, "ffn_mlp", il);
+
+            // Expert FFN
+            ggml_tensor * cur_moe = build_norm(attn_out,
+                    model.layers[il].ffn_pre_norm_2, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur_moe, "ffn_norm_2", il);
+
+            // custom MoE logits calculation (router operates on attn_out, not cur)
+            ggml_tensor * tmp = ggml_rms_norm(ctx0, attn_out, hparams.f_norm_rms_eps);
+            tmp = ggml_scale(ctx0, tmp, 1.0f / sqrtf((float) n_embd));
+            tmp = ggml_mul(ctx0, tmp, model.layers[il].ffn_gate_inp_s);
+            ggml_tensor * logits = build_lora_mm(model.layers[il].ffn_gate_inp, tmp); // [n_expert, n_tokens]
+            cb(logits, "ffn_moe_logits", il);
+
+            cur_moe = build_moe_ffn(cur_moe,
+                    nullptr, // gate_inp
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr, // exp_probs_b (not used for gemma4)
+                    n_expert, n_expert_used,
+                    LLM_FFN_GELU, true,
+                    1.0f,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il, logits,
+                    model.layers[il].ffn_gate_up_exps,
+                    model.layers[il].ffn_up_exps_s,
+                    model.layers[il].ffn_gate_exps_s,
+                    model.layers[il].ffn_down_exps_s);
+            cur_moe = build_norm(cur_moe,
+                    model.layers[il].ffn_post_norm_2, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur_moe, "ffn_moe", il);
+
+            cur = ggml_add(ctx0, cur_mlp, cur_moe);
+            cb(cur, "ffn_moe_combined", il);
+        } else {
+            cur = build_norm(attn_out,
+                    model.layers[il].ffn_norm, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   nullptr, model.layers[il].ffn_up_s,
+                    model.layers[il].ffn_gate, nullptr, model.layers[il].ffn_gate_s,
+                    model.layers[il].ffn_down, nullptr, model.layers[il].ffn_down_s,
+                    nullptr,
+                    LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, nullptr,
+                LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", il);
+
+        // residual connection
+        cur = ggml_add(ctx0, cur, attn_out);
+
+        // per-layer embedding
+        if (inp_per_layer) {
+            ggml_tensor * pe_in = cur;
+            cb(cur, "pe_in", il);
+
+            cur = build_lora_mm(model.layers[il].per_layer_inp_gate, cur); // [n_embd_per_layer, n_tokens]
+            cur = ggml_gelu(ctx0, cur);
+
+            ggml_tensor * inp_this_layer = ggml_view_2d_slice(ctx0, inp_per_layer, il); // [n_embd_per_layer, n_tokens]
+
+            // TODO @ngxson : improve this
+            if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) {
+                inp_this_layer = ggml_get_rows(ctx0, inp_this_layer, inp_out_ids);
+            }
+
+            cur = ggml_mul(ctx0, cur, inp_this_layer);
+            cur = build_lora_mm(model.layers[il].per_layer_proj, cur); // [n_embd, n_tokens]
+            cur = build_norm(cur, model.layers[il].per_layer_post_norm, nullptr, LLM_NORM_RMS, il);
+            cb(cur, "per_layer_embd_out", il);
+
+            // residual connection
+            cur = ggml_add(ctx0, pe_in, cur);
+        }
+
+        // layer_scalar
+        if (model.layers[il].out_scale) {
+            cur = ggml_mul(ctx0, cur, model.layers[il].out_scale);
+            cb(cur, "out_scaled", il);
+        }
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, nullptr,
+            LLM_NORM_RMS, -1);
+
+    // Expose the post-output-norm hidden state (the LM-head input feature) so that
+    // MTP draft contexts can read it via llama_get_embeddings_nextn_ith() as the
+    // recurrent h input. This matches the reference (transformers/vLLM/SGLang),
+    // which feeds the drafter the target's post-final-norm hidden state.
+    cb(cur, "h_nextn", -1);
+    res->t_h_nextn = cur;
+
+    if (!cparams.embeddings_nextn_masked && inp_out_ids) {
+        cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+    }
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    if (hparams.f_final_logit_softcapping) {
+        cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_final_logit_softcapping);
+        cur = ggml_tanh(ctx0, cur);
+        cur = ggml_scale(ctx0, cur, hparams.f_final_logit_softcapping);
+    }
+
+    // apply logits bias if needed (e.g. for gemma4_unified patch)
+    // this is to mirror the suppress_tokens patch on transformers, to avoid model from outputing <image|> and <audio|> tokens (which is a known issue related to the checkpoint)
+    // TODO: maybe handle this inside the sampling system in the future
+    if (!model.vocab.get_suppress_tokens().empty()) {
+        auto inp_bias = std::make_unique<llm_graph_input_logits_bias>(model.vocab);
+        inp_bias->logits_bias = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, inp_bias->arr.size());
+        cur = ggml_add(ctx0, cur, inp_bias->logits_bias);
+        res->add_input(std::move(inp_bias));
+    }
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// equivalent to get_per_layer_inputs() in python code
+// output shape: [n_embd_per_layer, n_layer, n_tokens]
+ggml_tensor * llama_model_gemma4::graph::build_inp_per_layer() {
+    auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
+
+    ggml_tensor * inp_per_layer;
+    float tok_embd_scale = sqrtf((float) n_embd_per_layer);
+    if (ubatch.token) {
+        inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
+        ggml_set_input(inp->tokens);
+        res->t_inp_tokens = inp->tokens;
+
+        inp_per_layer = ggml_get_rows  (ctx0, model.per_layer_tok_embd, inp->tokens);
+        inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_per_layer, n_layer, n_tokens);
+        inp_per_layer = ggml_scale     (ctx0, inp_per_layer, tok_embd_scale);
+        cb(inp_per_layer, "inp_per_layer_selected", -1);
+
+        res->add_input(std::move(inp));
+    } else {
+        // Multimodal embedding path: use padding token (ID=0) embedding
+        // TODO: verify if this is the correct behavior in transformers implementation
+        const int64_t embd_size = model.per_layer_tok_embd->ne[0];  // n_embd_per_layer * n_layer
+
+        // Extract and dequantize padding token embedding (row 0)
+        ggml_tensor * padding = ggml_view_1d(ctx0, model.per_layer_tok_embd, embd_size, 0);
+        inp_per_layer = ggml_cast (ctx0, padding, GGML_TYPE_F32);
+        inp_per_layer = ggml_scale(ctx0, inp_per_layer, tok_embd_scale);
+
+        // Reshape to [n_embd_per_layer, n_layer, 1]
+        inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_per_layer, n_layer, 1);
+        cb(inp_per_layer, "inp_per_layer_multimodal", -1);
+    }
+    return inp_per_layer;
+}
+
+// equivalent to project_per_layer_inputs() in python code
+// this calculates the per-layer inputs, so the final tensor shape will have n_layer as the last dim
+// inp_batch     shape: [n_embd, n_tokens]
+// inp_per_layer shape: [n_embd_per_layer, n_layer, n_tokens] (from build_inp_per_layer)
+// output shape: [n_embd_per_layer, n_tokens, n_layer]
+ggml_tensor * llama_model_gemma4::graph::project_per_layer_inputs(ggml_tensor * inp_batch, ggml_tensor * inp_per_layer) {
+    const float per_layer_projection_scale = 1.0f / sqrtf((float) n_embd);
+    const float per_layer_input_scale      = 1.0f / sqrtf(2.0f);
+
+    // note: this matrix multiplication will be performed in the input layer (i.e. on the CPU)
+    ggml_tensor * per_layer_proj;
+    per_layer_proj = ggml_mul_mat   (ctx0, model.per_layer_model_proj, inp_batch);
+    per_layer_proj = ggml_scale     (ctx0, per_layer_proj, per_layer_projection_scale);
+    per_layer_proj = ggml_reshape_3d(ctx0, per_layer_proj, n_embd_per_layer, n_layer, n_tokens);
+
+    per_layer_proj = build_norm(per_layer_proj, model.per_layer_proj_norm, nullptr, LLM_NORM_RMS, -1);
+    cb(per_layer_proj, "per_layer_proj", -1);
+
+    inp_per_layer = ggml_add  (ctx0, per_layer_proj, inp_per_layer);
+    inp_per_layer = ggml_scale(ctx0, inp_per_layer, per_layer_input_scale);
+    cb(inp_per_layer, "inp_per_layer", -1);
+
+    // permute to shape: [n_embd_per_layer, n_tokens, n_layer]
+    inp_per_layer = ggml_cont(ctx0, ggml_permute(ctx0, inp_per_layer, 0, 2, 1, 3));
+    return inp_per_layer;
+}

src/models/glm-dsa.cpp

@@ -0,0 +1,545 @@
+#include "models.h"
+
+#include "llama-kv-cache-dsa.h"
+
+// https://huggingface.co/zai-org/GLM-5.2/blob/main/config.json#L26
+const std::array<uint32_t, LLAMA_MAX_LAYERS> GLM_5_2_DEFAULT_INDEXER_TYPES = {
+    1, 1,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+    1, 0, 0, 0,
+};
+
+void llama_model_glm_dsa::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,     hparams.n_ff_exp);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,    hparams.f_norm_rms_eps);
+    ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false);
+
+    // MoE parameters
+    ml.get_key(LLM_KV_EXPERT_COUNT,                hparams.n_expert);
+    ml.get_key(LLM_KV_EXPERT_USED_COUNT,           hparams.n_expert_used);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+
+    // deepseek MLA parameters
+    ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK,      hparams.n_lora_q);
+    ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK,     hparams.n_lora_kv);
+    ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH_MLA,   hparams.n_embd_head_k_mla_impl, false);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_MLA, hparams.n_embd_head_v_mla_impl, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,        hparams.n_expert_shared);
+
+    // DSA parameters
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_HEAD_COUNT, hparams.indexer_n_head);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_KEY_LENGTH, hparams.indexer_head_size);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_TOP_K,      hparams.indexer_top_k);
+
+    // Expert gating function (GLM-4.5 uses sigmoid)
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func =  LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+    }
+
+    // NextN/MTP parameters
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_all");
+
+    // BC for GLM 5, 5.1 (full indexers) without indexer_types metadata
+    const bool is_pre_5_2 = hparams.n_ctx_train < 1048576;
+    if (is_pre_5_2) {
+        std::fill(hparams.is_indexer_full_impl.begin(), hparams.is_indexer_full_impl.end(), 1);
+    } else {
+        hparams.is_indexer_full_impl = GLM_5_2_DEFAULT_INDEXER_TYPES;
+    }
+    ml.get_key_or_arr(LLM_KV_ATTENTION_INDEXER_TYPES, hparams.is_indexer_full_impl, hparams.n_layer(), false);
+
+    switch (hparams.n_layer()) {
+        case 78: type = LLM_TYPE_744B_A40B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+    const bool is_mla = hparams.is_mla();
+    if (!is_mla) {
+        throw std::runtime_error("GLM_DSA architecture requires MLA");
+    }
+
+    // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+    const int64_t n_embd_head_k_mla = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v_mla = hparams.n_embd_head_v_mla();
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = n_embd_head_k_mla - n_embd_head_qk_rope;
+
+    const int64_t q_lora_rank  = hparams.n_lora_q;
+    const int64_t kv_lora_rank = hparams.n_lora_kv;
+
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    // try to load output.weight, if not found, use token_embd (tied embeddings)
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        int flags = 0;
+        if (i >= n_layer) {
+            // skip all tensors in the NextN layers
+            // TODO @ngxson : TENSOR_NOT_REQUIRED was a hack, need to remove it later
+            flags |= TENSOR_SKIP | TENSOR_NOT_REQUIRED;
+        }
+
+        auto & layer = layers[i];
+
+        layer.attn_norm      = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, flags);
+        layer.attn_q_a_norm  = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, flags);
+        layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, flags);
+
+        layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, flags);
+        layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k_mla}, flags);
+
+        layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + n_embd_head_qk_rope}, flags);
+
+        // note: only old legacy GGUF files will have the unsplit wkv_b tensor in
+        layer.wk_b = create_tensor(tn(LLM_TENSOR_ATTN_K_B, "weight", i), {n_embd_head_qk_nope, kv_lora_rank, n_head}, flags);
+        layer.wv_b = create_tensor(tn(LLM_TENSOR_ATTN_V_B, "weight", i), {kv_lora_rank, n_embd_head_v_mla, n_head}, flags);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_embd_head_v_mla, n_embd}, flags);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, flags);
+
+        // DSA indexer
+        layer.indexer_k_norm   = create_tensor(tn(LLM_TENSOR_INDEXER_K_NORM,   "weight", i), {hparams.indexer_head_size}, flags | TENSOR_NOT_REQUIRED);
+        layer.indexer_k_norm_b = create_tensor(tn(LLM_TENSOR_INDEXER_K_NORM,   "bias",   i), {hparams.indexer_head_size}, flags | TENSOR_NOT_REQUIRED);
+        layer.indexer_proj     = create_tensor(tn(LLM_TENSOR_INDEXER_PROJ,     "weight", i), {n_embd, hparams.indexer_n_head}, flags | TENSOR_NOT_REQUIRED);
+        layer.indexer_attn_k   = create_tensor(tn(LLM_TENSOR_INDEXER_ATTN_K,   "weight", i), {n_embd, hparams.indexer_head_size}, flags | TENSOR_NOT_REQUIRED);
+        layer.indexer_attn_q_b = create_tensor(tn(LLM_TENSOR_INDEXER_ATTN_Q_B, "weight", i), {q_lora_rank, hparams.indexer_n_head * hparams.indexer_head_size}, flags | TENSOR_NOT_REQUIRED);
+        if (i < (int) hparams.n_layer_dense_lead) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, flags);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, flags);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, flags);
+        } else {
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, flags);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+
+            if (n_expert == 0) {
+                throw std::runtime_error("n_expert must be > 0");
+            }
+            if (n_expert_used == 0) {
+                throw std::runtime_error("n_expert_used must be > 0");
+            }
+
+            // MoE branch
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, flags);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, flags);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, flags);
+
+            // Shared expert branch
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, flags);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, flags);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, flags);
+        }
+
+        // NextN/MTP tensors (preserved but unused) - conditionally load for last n_layer_nextn
+        if (i >= n_layer) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags);
+
+            // Optional tensors
+            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, flags | TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_glm_dsa::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_glm_dsa::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const bool is_mla = hparams.is_mla();
+    GGML_ASSERT(is_mla);
+
+    // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+    const int64_t n_embd_head_k = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v = hparams.n_embd_head_v_mla();
+    GGML_UNUSED(n_embd_head_v);
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = n_embd_head_k - n_embd_head_qk_rope;
+
+    const int64_t n_indexer_head = hparams.indexer_n_head;
+    const int64_t n_embd_indexer_head = hparams.indexer_head_size;
+    const int64_t n_embd_indexer_head_rope = hparams.n_rot();
+    const int64_t n_embd_indexer_head_nope = n_embd_indexer_head - n_embd_indexer_head_rope;
+    const uint32_t n_indexer_top_k = hparams.indexer_top_k;
+
+    const uint32_t kv_lora_rank = hparams.n_lora_kv;
+
+    // We have to pre-scale kq_scale and attn_factor to make the YaRN RoPE work correctly.
+    // See https://github.com/ggml-org/llama.cpp/discussions/7416 for detailed explanation.
+    // And also: https://github.com/ggml-org/llama.cpp/pull/17945 [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
+
+    // first cancel the adjustment from llama_hparams::yarn_attn_factor_adjust to get the original attn_factor
+    GGML_ASSERT(ext_factor >= 0.0f);
+    const float attn_factor_org = attn_factor * (1.0f + 0.1f * logf(1.0f / freq_scale));
+
+    // use the original attn_factor to pre-scale the kq_scale
+    const float mscale   = attn_factor_org * (1.0f + 0.1f * hparams.rope_yarn_log_mul * logf(1.0f / freq_scale));
+    const float kq_scale = 1.0f * mscale * mscale / sqrtf(float(n_embd_head_k));
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    // {n_embd, n_tokens}
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    llm_graph_input_attn_k_dsa * inp_attn_dsa = build_attn_inp_k_dsa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    // Difference vs Deepseek 3.2: shared indexer layers reuse the top_k from the previous full indexer layers
+    // See https://huggingface.co/zai-org/GLM-5.2/blob/main/config.json#L30
+    ggml_tensor * prev_top_k = nullptr;
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            ggml_tensor * qr = ggml_mul_mat(ctx0, model.layers[il].wq_a, cur);
+            cb(qr, "qr", il);
+
+            qr = build_norm(qr, model.layers[il].attn_q_a_norm, nullptr, LLM_NORM_RMS, il);
+            cb(qr, "qr", il);
+
+            ggml_tensor * top_k = nullptr;
+
+            // lightning indexer
+            if (hparams.is_indexer_full(il)) {
+                // "full" layer
+                ggml_tensor * indexer_q = ggml_mul_mat(ctx0, model.layers[il].indexer_attn_q_b, qr);
+                cb(indexer_q, "indexer_q", il);
+
+                // split into {n_embd_indexer_head_rope, n_indexer_head, n_tokens}
+                ggml_tensor * indexer_q_pe =
+                    ggml_view_3d(ctx0, indexer_q, n_embd_indexer_head_rope, n_indexer_head, n_tokens,
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head) * n_indexer_head, 0);
+                cb(indexer_q_pe, "indexer_q_pe", il);
+
+                // and {n_embd_indexer_head_nope, n_indexer_head, n_tokens}
+                ggml_tensor * indexer_q_nope =
+                    ggml_view_3d(ctx0, indexer_q, n_embd_indexer_head_nope, n_indexer_head, n_tokens,
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head) * n_indexer_head,
+                                 ggml_row_size(indexer_q->type, n_embd_indexer_head_nope));
+                cb(indexer_q_nope, "indexer_q_nope", il);
+
+                indexer_q_pe = ggml_rope_ext(ctx0, indexer_q_pe, inp_pos, nullptr, n_rot,
+                                     LLAMA_ROPE_TYPE_NORM, n_ctx_orig, freq_base, freq_scale,
+                                     ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(indexer_q_pe, "indexer_q_pe", il);
+
+                // {n_embd_indexer_head_rope + n_embd_indexer_head_nope, n_head, n_tokens}
+                indexer_q = ggml_concat(ctx0, indexer_q_pe, indexer_q_nope, 0);
+                cb(indexer_q, "indexer_q", il);
+
+                ggml_tensor * indexer_k = ggml_mul_mat(ctx0, model.layers[il].indexer_attn_k, cur);
+                cb(indexer_k, "indexer_k", il);
+
+                indexer_k = build_norm(indexer_k, model.layers[il].indexer_k_norm, model.layers[il].indexer_k_norm_b, LLM_NORM, il);
+                cb(indexer_k, "indexer_k", il);
+
+                // split into {n_embd_indexer_head_rope, 1, n_tokens}
+                ggml_tensor * indexer_k_pe =
+                    ggml_view_3d(ctx0, indexer_k, n_embd_indexer_head_rope, 1, n_tokens,
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head) * 1, 0);
+                cb(indexer_k_pe, "indexer_k_pe", il);
+
+                // and {n_embd_indexer_head_nope, 1, n_tokens}
+                ggml_tensor * indexer_k_nope =
+                    ggml_view_3d(ctx0, indexer_k, n_embd_indexer_head_nope, 1, n_tokens,
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head),
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head) * 1,
+                                 ggml_row_size(indexer_k->type, n_embd_indexer_head_nope));
+                cb(indexer_k_nope, "indexer_k_nope", il);
+
+                indexer_k_pe = ggml_rope_ext(ctx0, indexer_k_pe, inp_pos, nullptr, n_rot,
+                                     LLAMA_ROPE_TYPE_NORM, n_ctx_orig, freq_base, freq_scale,
+                                     ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(indexer_k_pe, "indexer_k_pe", il);
+
+                // {n_embd_indexer_head_rope + n_embd_indexer_head_nope, 1, n_tokens}
+                indexer_k = ggml_concat(ctx0, indexer_k_pe, indexer_k_nope, 0);
+                cb(indexer_k, "indexer_k", il);
+
+                // perform Hadamard transform on indexer q and k
+                indexer_q = ggml_mul_mat(ctx0, inp_attn_dsa->self_k_rot_lid, indexer_q);
+                cb(indexer_q, "indexer_q", il);
+                indexer_k = ggml_mul_mat(ctx0, inp_attn_dsa->self_k_rot_lid, indexer_k);
+                cb(indexer_k, "indexer_k", il);
+
+                // store indexer keys to KV cache
+                const auto * mctx_lid = inp_attn_dsa->mctx->get_lid();
+                const auto & k_idxs_lid = inp_attn_dsa->get_k_idxs_lid();
+                ggml_build_forward_expand(gf, mctx_lid->cpy_k(ctx0, indexer_k, k_idxs_lid, il));
+
+                // prepare indexer weights
+                ggml_tensor * indexer_weights = ggml_mul_mat(ctx0, model.layers[il].indexer_proj, cur);
+                cb(indexer_weights, "indexer_weights", il);
+
+                // get cached indexer keys
+                indexer_k = mctx_lid->get_k(ctx0, il);
+
+                // split the batch into streams if needed
+                const auto n_stream = indexer_k->ne[3];
+                indexer_q = ggml_view_4d(ctx0, indexer_q, indexer_q->ne[0], indexer_q->ne[1], indexer_q->ne[2]/n_stream, n_stream, indexer_q->nb[1], indexer_q->nb[2], indexer_q->nb[3]/n_stream, 0);
+                indexer_weights = ggml_view_4d(ctx0, indexer_weights, indexer_weights->ne[0], indexer_weights->ne[1]/n_stream, indexer_weights->ne[2], n_stream, indexer_weights->nb[1], indexer_weights->nb[2]/n_stream, indexer_weights->nb[3]/n_stream, 0);
+
+                // pre-scale weights to avoid scaling operations on huge indexer_score tensor
+                indexer_weights = ggml_scale(ctx0, indexer_weights, 1.0f / sqrtf(float(n_embd_indexer_head * n_indexer_head)));
+                cb(indexer_weights, "indexer_weights", il);
+
+                ggml_tensor * indexer_score = nullptr;
+                if (cparams.fused_lid) {
+                    indexer_score = ggml_lightning_indexer(ctx0, indexer_q, indexer_k, indexer_weights, inp_attn_dsa->get_kq_mask_lid());
+                    cb(indexer_score, "indexer_score", il);
+                    res->add_fused_node({LLM_FUSED_OP_LIGHTNING_INDEXER, indexer_score, il});
+                } else {
+                    // calculate indexer kq
+                    indexer_q = ggml_permute(ctx0, indexer_q, 0, 2, 1, 3);
+                    cb(indexer_q, "indexer_q", il);
+                    indexer_k = ggml_permute(ctx0, indexer_k, 0, 2, 1, 3);
+                    cb(indexer_k, "indexer_k", il);
+
+                    ggml_tensor * indexer_kq = ggml_mul_mat(ctx0, indexer_k, indexer_q);
+                    cb(indexer_kq, "indexer_kq", il);
+
+                    // ReLU requires contiguous tensors
+                    indexer_kq = ggml_cont(ctx0, ggml_permute(ctx0, indexer_kq, 2, 1, 0, 3));
+                    cb(indexer_kq, "indexer_kq", il);
+
+                    // apply ReLU
+                    indexer_score = ggml_relu(ctx0, indexer_kq);
+                    cb(indexer_score, "indexer_score", il);
+
+                    // multiply scores by indexer weights
+                    indexer_score = ggml_mul(ctx0, indexer_score, indexer_weights);
+                    cb(indexer_score, "indexer_score", il);
+
+                    // sum by q n_indexer_head dimension
+                    indexer_score = ggml_sum_rows(ctx0, indexer_score);
+                    cb(indexer_score, "indexer_score", il);
+
+                    // permute result to match KQ mask
+                    indexer_score = ggml_cont(ctx0, ggml_permute(ctx0, indexer_score, 2, 1, 0, 3));
+                    cb(indexer_score, "indexer_score", il);
+
+                    // mask indexer scores
+                    ggml_tensor * indexer_kq_mask = inp_attn_dsa->get_kq_mask_lid();
+                    indexer_score = ggml_add(ctx0, indexer_score, indexer_kq_mask);
+                    cb(indexer_score, "indexer_score", il);
+                }
+
+                // get indices of top k indexer scores
+                uint32_t n_top_k = indexer_score->ne[0] < n_indexer_top_k ? indexer_score->ne[0] : n_indexer_top_k;
+                top_k = ggml_cont(ctx0, ggml_top_k(ctx0, indexer_score, n_top_k));
+                prev_top_k = top_k;
+                cb(top_k, "top_k", il);
+            } else {
+                // "shared" indexer layer - reuse top-k from a previous full layer
+                GGML_ASSERT(prev_top_k != nullptr && "shared indexer layer must follow a previous full indexer layer");
+                top_k = prev_top_k;
+                cb(top_k, "top_k", il);
+            }
+
+            ggml_tensor * q = ggml_mul_mat(ctx0, model.layers[il].wq_b, qr);
+            cb(q, "q", il);
+
+            // split into {n_embd_head_qk_nope, n_head, n_tokens}
+            ggml_tensor * q_nope =
+                ggml_view_3d(ctx0, q, n_embd_head_qk_nope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+                             ggml_row_size(q->type, n_embd_head_k) * n_head, 0);
+            cb(q_nope, "q_nope", il);
+
+            // and {n_embd_head_qk_rope, n_head, n_tokens}
+            ggml_tensor * q_pe = ggml_view_3d(
+                ctx0, q, n_embd_head_qk_rope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+                ggml_row_size(q->type, n_embd_head_k) * n_head, ggml_row_size(q->type, n_embd_head_qk_nope));
+            cb(q_pe, "q_pe", il);
+
+            ggml_tensor * kv_cmpr_pe = ggml_mul_mat(ctx0, model.layers[il].wkv_a_mqa, cur);
+            cb(kv_cmpr_pe, "kv_cmpr_pe", il);
+
+            // split into {kv_lora_rank, n_tokens}
+            ggml_tensor * kv_cmpr =
+                ggml_view_2d(ctx0, kv_cmpr_pe, kv_lora_rank, n_tokens,
+                             ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope), 0);
+            cb(kv_cmpr, "kv_cmpr", il);
+
+            // and {n_embd_head_qk_rope, 1, n_tokens}
+            ggml_tensor * k_pe = ggml_view_3d(ctx0, kv_cmpr_pe, n_embd_head_qk_rope, 1, n_tokens,
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                                              ggml_row_size(kv_cmpr_pe->type, kv_lora_rank));
+            cb(k_pe, "k_pe", il);
+
+            q_pe = ggml_rope_ext(ctx0, q_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(q_pe, "q_pe", il);
+
+            k_pe = ggml_rope_ext(ctx0, k_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+            cb(k_pe, "k_pe", il);
+
+            kv_cmpr = build_norm(kv_cmpr, model.layers[il].attn_kv_a_norm, nullptr, LLM_NORM_RMS, il);
+            cb(kv_cmpr, "kv_cmpr", il);
+
+            // MLA attention
+            {
+                // {n_embd_head_qk_nope, n_tokens, n_head}
+                q_nope = ggml_permute(ctx0, q_nope, 0, 2, 1, 3);
+                cb(q_nope, "q_nope_perm", il);
+
+                // {n_embd_head_qk_nope, kv_lora_rank, n_head} x {n_embd_head_qk_nope, n_tokens, n_head}
+                ggml_tensor * q_nope_absorbed = ggml_mul_mat(ctx0, model.layers[il].wk_b, q_nope);
+                cb(q_nope_absorbed, "q_nope_absorbed", il);
+
+                // {kv_lora_rank, n_head, n_tokens}
+                q_nope_absorbed = ggml_permute(ctx0, q_nope_absorbed, 0, 2, 1, 3);
+                cb(q_nope_absorbed, "q_nope_absorbed_perm", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, n_head, n_tokens}
+                // note: rope must go first for in-place context shifting in build_rope_shift()
+                ggml_tensor * Qcur = ggml_concat(ctx0, q_nope_absorbed, q_pe, 0);
+                cb(Qcur, "Qcur", il);
+
+                kv_cmpr = ggml_reshape_3d(ctx0, kv_cmpr, kv_lora_rank, 1, n_tokens);
+                cb(kv_cmpr, "kv_cmpr_reshape", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Kcur = ggml_concat(ctx0, kv_cmpr, k_pe, 0);
+                cb(Kcur, "Kcur", il);
+
+                // {kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Vcur = kv_cmpr;
+                cb(Vcur, "Vcur", il);
+
+                // note: MLA with the absorption optimization converts into MQA (ie: GQA with 1 group)
+                cur = build_attn(inp_attn_dsa,
+                        model.layers[il].wo, NULL, model.layers[il].wo_s,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, model.layers[il].wv_b, top_k, kq_scale, il);
+            }
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, model.layers[il].ffn_up_s,
+                model.layers[il].ffn_gate, NULL, model.layers[il].ffn_gate_s,
+                model.layers[il].ffn_down, NULL, model.layers[il].ffn_down_s,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il,
+                nullptr,
+                model.layers[il].ffn_gate_up_exps,
+                model.layers[il].ffn_up_exps_s,
+                model.layers[il].ffn_gate_exps_s,
+                model.layers[il].ffn_down_exps_s);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // FFN shared expert
+            {
+                ggml_tensor * ffn_shexp =
+                    build_ffn(cur,
+                        model.layers[il].ffn_up_shexp, NULL, model.layers[il].ffn_up_shexp_s,
+                        model.layers[il].ffn_gate_shexp, NULL, model.layers[il].ffn_gate_shexp_s,
+                        model.layers[il].ffn_down_shexp, NULL, model.layers[il].ffn_down_shexp_s,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = ggml_mul_mat(ctx0, model.output, cur);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/glm4-moe.cpp

@@ -0,0 +1,280 @@
+#include "models.h"
+
+void llama_model_glm4_moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,     hparams.n_ff_exp);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,    hparams.f_norm_rms_eps);
+    ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false);
+
+    // MoE parameters
+    ml.get_key(LLM_KV_EXPERT_COUNT,                hparams.n_expert);
+    ml.get_key(LLM_KV_EXPERT_USED_COUNT,           hparams.n_expert_used);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+
+    // Expert gating function (GLM-4.5 uses sigmoid)
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func =  LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+    }
+
+    // NextN/MTP parameters
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl");
+
+    switch (hparams.n_layer()) {
+        case 46: type = LLM_TYPE_106B_A12B; break; // GLM-4.5-Air
+        case 48: type = LLM_TYPE_102B_A12B; break; // Solar Open
+        case 92: type = LLM_TYPE_355B_A32B; break; // GLM-4.5
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_glm4_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+
+    GGML_ASSERT(hparams.n_expert > 0 && "n_expert must be > 0 for GLM4_MOE MoE layers");
+    GGML_ASSERT(hparams.n_expert_used > 0 && "n_expert_used must be > 0 for GLM4_MOE MoE layers");
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
+    }
+
+    // Load ALL tensors including NextN layer to satisfy total tensor count
+    // but only PROCESS up to last layer (skipping final NextN layer) in forward pass
+    for (int i = 0; i < n_layer_all; ++i) {
+        int flags = 0;
+        if (i >= n_layer) {
+            // skip all tensors in the NextN layers
+            flags |= TENSOR_SKIP;
+        }
+
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, flags);
+
+        // GLM-style attention with bias terms
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, flags);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, flags);
+
+        // K/Q norm tensors (optional for GLM-4.5 355B variant)
+        layer.attn_q_norm = create_tensor(
+            tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), { n_embd_head_k }, TENSOR_NOT_REQUIRED | flags);
+        layer.attn_k_norm = create_tensor(
+            tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), { n_embd_head_k }, TENSOR_NOT_REQUIRED | flags);
+
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), { n_embd }, flags);
+
+        // Check if this layer uses MoE or dense FFN based on n_layer_dense_lead
+        // GLM 4.5 uses hybrid architecture: layer 0 is dense, layers 1+ are MoE
+        const bool use_moe = (static_cast<uint32_t>(i) >= hparams.n_layer_dense_lead);
+
+        if (use_moe) {
+            // MoE layers
+            layer.ffn_gate_inp =
+                create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), { n_embd, n_expert }, flags);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), { n_expert }, flags);
+
+            // MoE branch
+            const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+
+            layer.ffn_gate_exps = create_tensor(
+                tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert }, flags);
+            layer.ffn_down_exps = create_tensor(
+                tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff_exp, n_embd, n_expert }, flags);
+            layer.ffn_up_exps = create_tensor(
+                tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert }, flags);
+
+            // Shared expert
+            if (n_expert_shared > 0) {
+                const int64_t n_ff_shexp = n_ff_exp * n_expert_shared;
+                layer.ffn_gate_shexp = create_tensor(
+                    tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), { n_embd, n_ff_shexp }, flags);
+                layer.ffn_down_shexp = create_tensor(
+                    tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_shexp, n_embd }, flags);
+                layer.ffn_up_shexp = create_tensor(
+                    tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), { n_embd, n_ff_shexp }, flags);
+            }
+        } else {
+            // Dense layers (first k layers) - GLM uses separate gate/up projections
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, flags);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, flags);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), { n_embd, n_ff }, flags);
+        }
+
+        // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers
+        if (i >= n_layer) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags);
+
+            // Optional tensors
+            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, flags | TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_glm4_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_glm4_moe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    int sections[4];
+    std::copy(std::begin(hparams.rope_sections), std::begin(hparams.rope_sections) + 4, sections);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    bool use_mrope = hparams.use_mrope();
+    if (ubatch.embd && !use_mrope) {
+        // unfortunately, we need to forcefully stop here, to avoid users complaining about wrong results
+        GGML_ABORT("This GGUF does not support multimodal. Please reconvert it.");
+    }
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    // Only process up to last layer (skip final NextN layer)
+    // Final layer tensors are loaded but not processed in forward pass
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // Pre-attention norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // Apply Q/K norm if available (GLM-4.5 355B variant)
+            if (model.layers[il].attn_q_norm) {
+                Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+                cb(Qcur, "Qcur_normed", il);
+            }
+            if (model.layers[il].attn_k_norm) {
+                Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+                cb(Kcur, "Kcur_normed", il);
+            }
+
+            if (use_mrope) {
+                Qcur = ggml_rope_multi(ctx0, Qcur, inp_pos, nullptr,
+                            n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_multi(ctx0, Kcur, inp_pos, nullptr,
+                            n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow);
+            } else {
+                // Normal RoPE
+                Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot,
+                                    rope_type, n_ctx_orig, freq_base, freq_scale,
+                                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot,
+                                    rope_type, n_ctx_orig, freq_base, freq_scale,
+                                    ext_factor, attn_factor, beta_fast, beta_slow);
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // Post-attention norm
+        cur = build_norm(ffn_inp, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "post_attn_norm", il);
+
+        // Check if this is a dense layer (n_layer_dense_lead=1, so layer 0 is dense)
+        if (static_cast<uint32_t>(il) < hparams.n_layer_dense_lead) {
+            // Dense FFN layer
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // Process routed experts using existing MoE infrastructure
+            ggml_tensor * routed_out = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il);
+            cb(routed_out, "ffn_moe_out", il);
+
+            // Process shared expert on original input
+            ggml_tensor * shared_out = build_ffn(cur,
+                    model.layers[il].ffn_up_shexp,   NULL, NULL,
+                    model.layers[il].ffn_gate_shexp, NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(shared_out, "ffn_shexp_out", il);
+
+            // Final output: routed_output + shared_output
+            cur = ggml_add(ctx0, routed_out, shared_out);
+            cb(cur, "ffn_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/glm4.cpp

@@ -0,0 +1,190 @@
+#include "models.h"
+
+void llama_model_glm4::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,    hparams.f_norm_rms_eps);
+    ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false);
+
+    // NextN/MTP parameters (GLM-OCR)
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl");
+
+    switch (hparams.n_layer()) {
+        case 17: type = LLM_TYPE_1B; break; // GLM-OCR
+        case 40: type = LLM_TYPE_9B; break;
+        case 61: type = LLM_TYPE_32B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_glm4::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        int flags = 0;
+        if (i >= n_layer) {
+            // skip all tensors in the NextN layers
+            flags |= TENSOR_SKIP;
+        }
+
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, flags);
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, flags);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, flags);
+
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, flags);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, flags);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, flags);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff * 2}, flags);
+
+        layer.ffn_post_norm  = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, flags);
+
+        // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers
+        if (i >= n_layer) {
+            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
+            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
+            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags);
+
+            // Optional tensors
+            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, flags | TENSOR_NOT_REQUIRED);
+            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, flags | TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_glm4::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_glm4::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    int sections[4];
+    std::copy(std::begin(hparams.rope_sections), std::begin(hparams.rope_sections) + 4, sections);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    bool use_mrope = hparams.use_mrope();
+    if (ubatch.embd && !use_mrope) {
+        // unfortunately, we need to forcefully stop here, to avoid users complaining about wrong results
+        GGML_ABORT("This GGUF does not support multimodal. Please reconvert it.");
+    }
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    // Only process up to last layer (skip final NextN layer)
+    // Final layer tensors are loaded but not processed in forward pass
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // Pre-attention norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (use_mrope) {
+                Qcur = ggml_rope_multi(ctx0, Qcur, inp_pos, nullptr,
+                            n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_multi(ctx0, Kcur, inp_pos, nullptr,
+                            n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow);
+            } else {
+                // Normal RoPE
+                Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot,
+                                    rope_type, n_ctx_orig, freq_base, freq_scale,
+                                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+                Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot,
+                                    rope_type, n_ctx_orig, freq_base, freq_scale,
+                                    ext_factor, attn_factor, beta_fast, beta_slow);
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        // Post-attention norm (new!)
+        cur = build_norm(cur, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "post_attn_norm", il);
+
+        // Add the input (residual connection after post-attention norm)
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FF
+        {
+            // Pre-MLP norm
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            // MLP
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up, NULL, NULL,
+                    NULL, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL, LLM_FFN_SWIGLU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+
+            // Post-MLP norm
+            cur = build_norm(cur, model.layers[il].ffn_post_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "post_mlp_norm", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    // Final norm
+    cur = build_norm(inpL, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // Output projection
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/gpt2.cpp

@@ -0,0 +1,148 @@
+#include "models.h"
+
+void llama_model_gpt2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 12: type = LLM_TYPE_SMALL; break;
+        case 24: type = LLM_TYPE_MEDIUM; break;
+        case 36: type = LLM_TYPE_LARGE; break;
+        case 48: type = LLM_TYPE_XL; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_gpt2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+    pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,   "weight"), {n_embd, n_ctx_train}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "bias", i),   {n_embd}, 0);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wqkv_b = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, 0);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gpt2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_gpt2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * pos;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    pos = ggml_get_rows(ctx0, model.pos_embd, inp_pos);
+    cb(pos, "pos_embd", -1);
+
+    inpL = ggml_add(ctx0, inpL, pos);
+    cb(inpL, "inpL", -1);
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // add the input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FF
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    NULL,                      NULL,                        NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = build_norm(inpL,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/gptneox.cpp

@@ -0,0 +1,219 @@
+#include "models.h"
+
+void llama_model_gptneox::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+    ml.get_key(LLM_KV_USE_PARALLEL_RESIDUAL,   hparams.use_par_res);
+
+    switch (hparams.n_layer()) {
+        case 6:
+            switch (hparams.n_ff()) {
+                case 512:  type = LLM_TYPE_14M; break;
+                case 2048: type = LLM_TYPE_70M; break;
+                default:   type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 12:
+            switch (hparams.n_ff()) {
+                case 3072: type = LLM_TYPE_160M; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 16:
+            switch (hparams.n_ff()) {
+                case 8192: type = LLM_TYPE_1B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 24:
+            switch (hparams.n_ff()) {
+                case 4096: type = LLM_TYPE_410M; break;
+                case 8192: type = LLM_TYPE_1_4B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 32:
+            switch (hparams.n_ff()) {
+                case 10240: type = LLM_TYPE_2_8B; break;
+                case 16384: type = LLM_TYPE_6_9B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 36:
+            switch (hparams.n_ff()) {
+                case 20480: type = LLM_TYPE_12B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 44:
+            switch (hparams.n_ff()) {
+                case 24576: type = LLM_TYPE_20B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_gptneox::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wqkv_b = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, 0);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_gptneox::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_gptneox::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // ffn
+        if (hparams.use_par_res) {
+            // attention and ffn are computed in parallel
+            // x = x + attn(ln1(x)) + ffn(ln2(x))
+
+            ggml_tensor * attn_out = cur;
+
+            cur = build_norm(inpL,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    NULL,                      NULL,                        NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+
+            cur = ggml_add(ctx0, cur, inpL);
+            cb(cur, "ffn_out", il);
+
+            cur = ggml_add(ctx0, cur, attn_out);
+
+            cur = build_cvec(cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        } else {
+            // attention and ffn are computed sequentially
+            // x = x + attn(ln1(x))
+            // x = x + ffn(ln2(x))
+
+            ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+            cb(ffn_inp, "ffn_inp", il);
+
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    NULL,                      NULL,                        NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+
+            cur = ggml_add(ctx0, cur, ffn_inp);
+
+            cur = build_cvec(cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        }
+    }
+
+    cur = build_norm(inpL,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/granite-hybrid.cpp

@@ -0,0 +1,302 @@
+#include "models.h"
+
+void llama_model_granite_hybrid::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LOGIT_SCALE,                 hparams.f_logit_scale, /* required */ false);
+    ml.get_key(LLM_KV_RESIDUAL_SCALE,              hparams.f_residual_scale, /* required */ false);
+    ml.get_key(LLM_KV_EMBEDDING_SCALE,             hparams.f_embedding_scale, /* required */ false);
+    ml.get_key(LLM_KV_ATTENTION_SCALE,             hparams.f_attention_scale, /* required */ false);
+
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
+    ml.get_key(LLM_KV_SSM_STATE_SIZE,     hparams.ssm_d_state);
+    ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
+    ml.get_key(LLM_KV_SSM_GROUP_COUNT,    hparams.ssm_n_group);
+
+    // Granite uses rope_finetuned as a switch for rope, so default to true
+    bool rope_finetuned = true;
+    ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false);
+    hparams.rope_finetuned = rope_finetuned;
+
+    // A layer is recurrent IFF the n_head_kv value is set to 0
+    for (uint32_t i = 0; i < hparams.n_layer(); ++i) {
+        hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0;
+    }
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_embd) {
+        case 768: type = LLM_TYPE_350M; break;
+        case 1536: type = (hparams.n_ff() == 512 ? LLM_TYPE_7B_A1B : LLM_TYPE_1B); break;
+        case 2048: case 2560: type = LLM_TYPE_3B; break;
+        case 4096: type = LLM_TYPE_32B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+
+    // For Granite MoE Shared
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, /* required */ false);
+}
+
+void llama_model_granite_hybrid::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    // mamba2 Mixer SSM params
+    // NOTE: int64_t for tensor dimensions
+    const int64_t d_conv     = hparams.ssm_d_conv;
+    const int64_t d_inner    = hparams.ssm_d_inner;
+    const int64_t d_state    = hparams.ssm_d_state;
+    const int64_t n_ssm_head = hparams.ssm_dt_rank;
+    const int64_t n_group    = hparams.ssm_n_group;
+    const int64_t d_in_proj  = 2*d_inner + 2*n_group*d_state + n_ssm_head;
+
+    // only an expansion factor of 2 is supported for now
+    GGML_ASSERT(2 * n_embd == d_inner);
+
+    // embeddings
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    {
+        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+        output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+        // if output is NULL, init from the input tok embed, duplicated to allow offloading
+        if (output == NULL) {
+            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+        }
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // norm
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.is_recr(i)) {
+            // ssm layers
+            layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, d_in_proj}, 0);
+
+            layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner + 2*n_group*d_state}, 0);
+            layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner + 2*n_group*d_state}, TENSOR_NOT_REQUIRED);
+
+            layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {n_ssm_head}, 0);
+
+            // no "weight" suffix for these
+            layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {1, n_ssm_head}, 0);
+            layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {1, n_ssm_head}, 0);
+
+            layer.ssm_norm = create_tensor(tn(LLM_TENSOR_SSM_NORM, "weight", i), {d_inner / n_group, n_group}, 0);
+
+            // out_proj
+            layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd}, 0);
+        } else {
+            // attention layers (with optional bias)
+            const int64_t n_head_i = hparams.n_head(i);
+            const int64_t n_embd_k_gqa_i = hparams.n_embd_k_gqa(i);
+            const int64_t n_embd_v_gqa_i = hparams.n_embd_v_gqa(i);
+            create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head_i, n_embd_k_gqa_i, n_embd_v_gqa_i, 0);
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head_i, n_embd}, 0);
+            layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        }
+
+        // feed forward (w/ optional biases)
+        if (n_expert > 0) {
+            // MoE FFN
+            layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+            // For Granite MoE Shared
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd}, 0);
+            }
+        } else {
+            layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_granite_hybrid::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_granite_hybrid::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_build_mamba_base(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * inp = build_inp_mem_hybrid();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    // Positional embeddings populated if rope enabled
+    ggml_tensor * inp_pos = nullptr;
+    if (hparams.rope_finetuned) {
+        inp_pos = build_inp_pos();
+    }
+
+    for (int il = 0; il < n_layer; ++il) {
+        struct ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        if (hparams.is_recr(il)) {
+            // ssm layer //
+            cur = build_mamba2_layer(inp->get_recr(), cur, model, ubatch, il);
+        } else {
+            // attention layer //
+            cur = build_attention_layer(cur, inp_pos, inp->get_attn(), model, n_embd_head, il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        // ffn
+        cur = build_layer_ffn(cur, inpSA, model, il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    // For Granite architectures - scale logits
+    if (hparams.f_logit_scale) {
+        cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_logit_scale);
+    }
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+ggml_tensor * llama_model_granite_hybrid::graph::build_attention_layer(ggml_tensor *             cur,
+                                                              ggml_tensor *             inp_pos,
+                                                              llm_graph_input_attn_kv * inp_attn,
+                                                              const llama_model &       model,
+                                                              const int64_t             n_embd_head,
+                                                              const int                 il) {
+    auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur, n_embd_head, hparams.n_head(il), hparams.n_head_kv(il), il);
+
+    const bool use_rope = hparams.rope_finetuned;
+    if (use_rope) {
+        ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+        Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                             ext_factor, attn_factor, beta_fast, beta_slow);
+
+        Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                             ext_factor, attn_factor, beta_fast, beta_slow);
+    }
+
+    cb(Qcur, "Qcur", il);
+    cb(Kcur, "Kcur", il);
+    cb(Vcur, "Vcur", il);
+
+    const float kq_scale =
+        hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+    cur = build_attn(inp_attn,
+            model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+            Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+    cb(cur, "attn_out", il);
+    return cur;
+}
+
+ggml_tensor * llama_model_granite_hybrid::graph::build_layer_ffn(ggml_tensor *       cur,
+                                                        ggml_tensor *       inpSA,
+                                                        const llama_model & model,
+                                                        const int           il) {
+    // For Granite architectures - scale residual
+    if (hparams.f_residual_scale) {
+        cur = ggml_scale(ctx0, cur, hparams.f_residual_scale);
+    }
+    ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+    cb(ffn_inp, "ffn_inp", il);
+
+    // feed-forward network (non-MoE)
+    if (model.layers[il].ffn_gate_inp == nullptr) {
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
+                model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+    } else {
+        // MoE branch
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        ggml_tensor * moe_out =
+            build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(moe_out, "ffn_moe_out", il);
+
+        // For Granite MoE Shared
+        if (hparams.n_ff_shexp > 0) {
+            ggml_tensor * ffn_shexp =
+                build_ffn(cur,
+                    model.layers[il].ffn_up_shexp, NULL, NULL,
+                    model.layers[il].ffn_gate_shexp, NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, NULL,
+                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(ffn_shexp, "ffn_shexp", il);
+
+            cur = ggml_add(ctx0, moe_out, ffn_shexp);
+            cb(cur, "ffn_out", il);
+        } else {
+            cur = moe_out;
+        }
+    }
+
+    // For Granite architectures - scale residual
+    if (hparams.f_residual_scale) {
+        cur = ggml_scale(ctx0, cur, hparams.f_residual_scale);
+    }
+    cur = ggml_add(ctx0, cur, ffn_inp);
+    cb(cur, "ffn_out", il);
+
+    cur = build_cvec(cur, il);
+    cb(cur, "l_out", il);
+
+    return cur;
+}

src/models/granite-moe.cpp

@@ -0,0 +1,89 @@
+#include "models.h"
+
+void llama_model_granite_moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LOGIT_SCALE,                 hparams.f_logit_scale);
+    ml.get_key(LLM_KV_RESIDUAL_SCALE,              hparams.f_residual_scale, false);
+    ml.get_key(LLM_KV_EMBEDDING_SCALE,             hparams.f_embedding_scale, false);
+    ml.get_key(LLM_KV_ATTENTION_SCALE,             hparams.f_attention_scale, false);
+
+    // Granite uses rope_finetuned as a switch for rope, so default to true
+    bool rope_finetuned = true;
+    ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false);
+    hparams.rope_finetuned = rope_finetuned;
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_3B; break;
+        case 40: type = LLM_TYPE_3B; break;
+        // Add additional layer/vocab/etc checks here for other model sizes
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+
+    // For Granite MoE Shared
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, /* required */ false);
+}
+
+void llama_model_granite_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+        else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        if (n_expert == 0) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+            // optional MLP bias
+            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        } else {
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+            // For Granite MoE Shared
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd}, 0);
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_granite_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/granite.cpp

@@ -0,0 +1,311 @@
+#include "models.h"
+
+#include <sstream>
+
+void llama_model_granite::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LOGIT_SCALE,                 hparams.f_logit_scale);
+    ml.get_key(LLM_KV_RESIDUAL_SCALE,              hparams.f_residual_scale, false);
+    ml.get_key(LLM_KV_EMBEDDING_SCALE,             hparams.f_embedding_scale, false);
+    ml.get_key(LLM_KV_ATTENTION_SCALE,             hparams.f_attention_scale, false);
+
+    // Granite4 Vision uses array deepstack_mapping
+    ml.get_arr(LLM_KV_DEEPSTACK_MAPPING, hparams.deepstack_mapping_arr, false);
+
+    // Count the unique deepstack input indices
+    std::unordered_set<uint32_t> unique_deepstack_idxs;
+    for (const auto val : hparams.deepstack_mapping_arr) {
+        if (val >= 0) {
+            unique_deepstack_idxs.insert(val);
+        }
+    }
+    hparams.n_deepstack_layers = unique_deepstack_idxs.size();
+
+    // Ensure all values are valid (avoid overflow attacks)
+    for (const auto val : unique_deepstack_idxs) {
+        if (val > hparams.n_deepstack_layers) {
+            std::stringstream ss;
+            ss << "Invalid deepstack index: " << val << " > " << hparams.n_deepstack_layers;
+            throw std::runtime_error(ss.str());
+        }
+    }
+
+    // Granite uses rope_finetuned as a switch for rope, so default to true
+    bool rope_finetuned = true;
+    ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false);
+    hparams.rope_finetuned = rope_finetuned;
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_3B; break;
+        case 40: type = LLM_TYPE_3B; break;
+        // Add additional layer/vocab/etc checks here for other model sizes
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+
+    // For Granite MoE Shared
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, /* required */ false);
+}
+
+void llama_model_granite::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+        else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        if (n_expert == 0) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+            // optional MLP bias
+            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        } else {
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+            // For Granite MoE Shared
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd}, 0);
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_granite::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_granite::graph::graph(
+    const llama_model & model,
+    const llm_graph_params & params)
+    : llm_graph_context(params) {
+
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - built only if rope enabled
+    ggml_tensor * inp_pos = nullptr;
+    if (hparams.rope_finetuned) {
+        inp_pos = build_inp_pos();
+    }
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+
+        // Granite Vision 4.1 deepstack: inject the projector stream that
+        // targets decoder layer `il` before the decoder runs.
+        // NOTE: skip the first deepstack layer since that's inpL
+        const auto & deepstack_emb_idx = hparams.deepstack_mapping_arr[il];
+        if (il > 0 && deepstack_emb_idx >= 0) {
+            ggml_tensor * ds = ggml_view_2d(ctx0,
+                res->t_inp_embd, n_embd, n_tokens,
+                res->t_inp_embd->nb[1],
+                deepstack_emb_idx * n_embd * sizeof(float));
+            inpL = ggml_add(ctx0, inpL, ds);
+            cb(inpL, "deepstack_in", il);
+        }
+
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        cur = build_attention_layer(
+            cur, inp_pos, inp_attn,
+            model, n_embd_head, il);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        // ffn
+        cur = build_layer_ffn(cur, inpSA, model, il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    // For Granite architectures - scale logits
+    cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_logit_scale);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+ggml_tensor * llama_model_granite::graph::build_attention_layer(
+          ggml_tensor             * cur,
+          ggml_tensor             * inp_pos,
+          llm_graph_input_attn_kv * inp_attn,
+    const llama_model             & model,
+    const int64_t                 n_embd_head,
+    const int                     il) {
+
+    auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+            n_embd_head, hparams.n_head(il), hparams.n_head_kv(il), il);
+
+    const bool use_rope = hparams.rope_finetuned;
+    if (use_rope) {
+        ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+        Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, rope_factors,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+        Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, rope_factors,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+    }
+
+    cb(Qcur, "Qcur", il);
+    cb(Kcur, "Kcur", il);
+    cb(Vcur, "Vcur", il);
+
+    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+    cur = build_attn(inp_attn,
+            model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+            Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+    return cur;
+}
+
+ggml_tensor * llama_model_granite::graph::build_layer_ffn(
+          ggml_tensor       * cur,
+          ggml_tensor       * inpSA,
+    const llama_model       & model,
+    const int                 il) {
+
+    // For Granite architectures - scale residual
+    if (hparams.f_residual_scale) {
+        cur = ggml_scale(ctx0, cur, hparams.f_residual_scale);
+    }
+    ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+    cb(ffn_inp, "ffn_inp", il);
+
+    // feed-forward network (non-MoE)
+    if (model.layers[il].ffn_gate_inp == nullptr) {
+
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+                cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(cur, "ffn_out", il);
+
+    } else {
+        // MoE branch
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+                cb(cur, "ffn_norm", il);
+
+        ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(moe_out, "ffn_moe_out", il);
+
+        // For Granite MoE Shared
+        if (hparams.n_ff_shexp > 0) {
+            ggml_tensor * ffn_shexp = build_ffn(cur,
+                model.layers[il].ffn_up_shexp,   NULL, NULL,
+                model.layers[il].ffn_gate_shexp, NULL, NULL,
+                model.layers[il].ffn_down_shexp, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(ffn_shexp, "ffn_shexp", il);
+
+            cur = ggml_add(ctx0, moe_out, ffn_shexp);
+            cb(cur, "ffn_out", il);
+        } else {
+            cur = moe_out;
+        }
+    }
+
+    // For Granite architectures - scale residual
+    if (hparams.f_residual_scale) {
+        cur = ggml_scale(ctx0, cur, hparams.f_residual_scale);
+    }
+    cur = ggml_add(ctx0, cur, ffn_inp);
+    cb(cur, "ffn_out", il);
+
+    cur = build_cvec(cur, il);
+    cb(cur, "l_out", il);
+
+    return cur;
+}

src/models/grok.cpp

@@ -0,0 +1,223 @@
+#include "models.h"
+
+void llama_model_grok::load_arch_hparams(llama_model_loader & ml) {
+    // defaults for old GGUFs
+    hparams.yarn_beta_fast = 8.0f;
+    hparams.f_logit_scale = 0.5773502691896257f;
+    hparams.f_embedding_scale = 78.38367176906169f;
+    hparams.f_attn_out_scale = 0.08838834764831845f;
+    hparams.f_attn_logit_softcapping = 30.0f;
+    hparams.f_router_logit_softcapping = 30.0f;
+    // no final_logit_softcapping in grok-1
+    hparams.f_final_logit_softcapping = 0.0f;
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,  hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,   hparams.n_ff_exp, false);
+    ml.get_key(LLM_KV_LOGIT_SCALE,                  hparams.f_logit_scale, false);
+    ml.get_key(LLM_KV_EMBEDDING_SCALE,              hparams.f_embedding_scale, false);
+    ml.get_key(LLM_KV_ATTENTION_OUTPUT_SCALE,       hparams.f_attn_out_scale, false);
+    ml.get_key(LLM_KV_ATTN_LOGIT_SOFTCAPPING,       hparams.f_attn_logit_softcapping, false);
+    ml.get_key(LLM_KV_ROUTER_LOGIT_SOFTCAPPING,     hparams.f_router_logit_softcapping, false);
+    ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING,      hparams.f_final_logit_softcapping, false);
+
+    ml.get_key(LLM_KV_ATTENTION_TEMPERATURE_LENGTH,  hparams.attn_temp_length, false);
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_EXT_FACTOR,  hparams.yarn_ext_factor, false);
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_ATTN_FACTOR, hparams.yarn_attn_factor, false);
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_FAST,   hparams.yarn_beta_fast, false);
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_SLOW,   hparams.yarn_beta_slow, false);
+
+    switch (hparams.n_layer()) {
+        case 64: type = LLM_TYPE_314B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_grok::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_expert == 0) {
+        throw std::runtime_error(arch_name() + " model cannot have zero experts");
+    }
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff/* / n_expert_used*/; // grok-1 n_ff_exp == n_ff
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff,   n_embd}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff_exp, n_expert}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd,   n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff_exp, n_expert}, 0);
+
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        if (!layer.ffn_post_norm) {
+            layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_grok::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_grok::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        cur = build_norm(cur,
+                model.layers[il].attn_out_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_out_norm", il);
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // MoE branch
+        ggml_tensor * moe_out = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_GELU, true,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(moe_out, "ffn_moe_out", il);
+
+        if (model.layers[il].ffn_up) {
+            ggml_tensor * ffn_out = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_PAR, il);
+            cb(ffn_out, "ffn_out", il);
+
+            cur = ggml_scale(ctx0, ggml_add(ctx0, ffn_out, moe_out), std::sqrt(2) / 2);
+            cb(cur, "ffn_out", il);
+        } else {
+            cur = moe_out;
+        }
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_post_norm", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cur = ggml_scale(ctx0, cur, hparams.f_logit_scale);
+
+    // final logit soft-capping
+    if (hparams.f_final_logit_softcapping) {
+        cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_final_logit_softcapping);
+        cur = ggml_tanh(ctx0, cur);
+        cur = ggml_scale(ctx0, cur, hparams.f_final_logit_softcapping);
+    }
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/grovemoe.cpp

@@ -0,0 +1,193 @@
+#include "models.h"
+
+void llama_model_grovemoe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_CHUNK_FEED_FORWARD_LENGTH,  hparams.n_ff_chexp, false);
+    ml.get_key(LLM_KV_EXPERT_GROUP_SCALE,                hparams.expert_group_scale);
+    ml.get_key(LLM_KV_EXPERTS_PER_GROUP,                 hparams.n_group_experts);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 48: type = LLM_TYPE_30B_A3B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_grovemoe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    GGML_ASSERT(n_expert > 0 && "n_expert must be > 0 for GROVEMOE");
+    GGML_ASSERT(n_expert_used > 0 && "n_expert_used must be > 0 for GROVEMOE");
+    GGML_ASSERT(hparams.n_group_experts > 0 && "n_group_experts must be > 0 for GROVEMOE");
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+        // MoE branch
+        const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+        const int64_t n_ff_chexp = hparams.n_ff_chexp ? hparams.n_ff_chexp : n_embd_head_k;
+        const int64_t n_chunk_expert = n_expert / hparams.n_group_experts;
+
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+        layer.ffn_gate_chexps = create_tensor(tn(LLM_TENSOR_FFN_GATE_CHEXPS, "weight", i), {  n_embd, n_ff_chexp, n_chunk_expert}, 0);
+        layer.ffn_down_chexps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_CHEXPS, "weight", i), {n_ff_chexp,   n_embd, n_chunk_expert}, 0);
+        layer.ffn_up_chexps   = create_tensor(tn(LLM_TENSOR_FFN_UP_CHEXPS,   "weight", i), {  n_embd, n_ff_chexp, n_chunk_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_grovemoe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_grovemoe::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    const int64_t n_embd_head    = hparams.n_embd_head_v();
+    const int64_t n_chunk_expert = n_expert / hparams.n_group_experts;
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                 ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // MoE branch
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        ggml_tensor * probs = build_lora_mm(model.layers[il].ffn_gate_inp, cur);  // [n_expert, n_tokens]
+        cb(probs, "ffn_moe_logits", il);
+
+        ggml_tensor * moe_out =
+            build_moe_ffn(cur,
+                nullptr,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il,
+                probs);
+        cb(moe_out, "ffn_moe_out", il);
+        cur = moe_out;
+
+        // TODO: Only do the expert selection and weights once
+        moe_out = build_moe_ffn(cur,
+                    nullptr,
+                    model.layers[il].ffn_up_chexps,
+                    model.layers[il].ffn_gate_chexps,
+                    model.layers[il].ffn_down_chexps,
+                    nullptr,
+                    n_chunk_expert, n_expert_used > n_chunk_expert ? n_chunk_expert : n_expert_used,
+                    LLM_FFN_SILU, true,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il,
+                    probs);
+        cb(moe_out, "ffn_adj_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ggml_scale(ctx0, moe_out, hparams.expert_group_scale));
+        cb(cur, "ffn_final_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/hunyuan-dense.cpp

@@ -0,0 +1,6 @@
+#include "models.h"
+
+std::unique_ptr<llm_graph_context> llama_model_hunyuan_dense::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/hunyuan-moe.cpp

@@ -0,0 +1,187 @@
+#include "models.h"
+
+void llama_model_hunyuan_moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_A13B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_hunyuan_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+        const uint32_t n_ff_shexp = hparams.n_ff_shexp > 0 ? hparams.n_ff_shexp : hparams.n_ff(i);
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+        layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0);
+        layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, 0);
+        layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_hunyuan_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_hunyuan_moe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale = 1.0f / sqrtf(float(n_embd_head));
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = build_norm(Kcur,
+                    model.layers[il].attn_k_norm, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_norm", il);
+
+            Qcur = build_norm(Qcur,
+                    model.layers[il].attn_q_norm, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_norm", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp,
+            model.layers[il].ffn_norm, NULL,
+            LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network (non-MoE)
+        ggml_tensor * cur_mlp = build_ffn(cur,
+                model.layers[il].ffn_up_shexp,   NULL, NULL,
+                model.layers[il].ffn_gate_shexp, NULL, NULL,
+                model.layers[il].ffn_down_shexp, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur_mlp, "ffn_mlp", il);
+
+        // MoE branch
+        ggml_tensor * cur_moe = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU,
+                true, // norm_topk_prob
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(cur_moe, "ffn_moe_out", il);
+
+        ggml_tensor * ffn_out = ggml_add(ctx0, cur_moe, cur_mlp);
+        cb(ffn_out, "ffn_out", il);
+
+        cur = ggml_add(ctx0, ffn_out, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/hunyuan-vl.cpp

@@ -0,0 +1,189 @@
+#include "models.h"
+
+void llama_model_hunyuan_vl::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false);
+
+    // XDRoPE / NTK-aware scaling: base = rope_theta * alpha^(dim / (dim - 2))
+    if (hparams.rope_scaling_alpha > 0.0f) {
+        const int dim = hparams.n_embd_head_k();
+        hparams.rope_freq_base_train = hparams.rope_freq_base_train
+            * powf(hparams.rope_scaling_alpha, (float)dim / (float)(dim - 2));
+    }
+
+    switch (hparams.n_embd) {
+        case 1024: type = LLM_TYPE_0_5B; break;
+        case 2048: type = LLM_TYPE_1_8B; break;
+        case 3072: type = LLM_TYPE_4B; break;
+        case 4096: type = LLM_TYPE_7B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_hunyuan_vl::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_hunyuan_vl::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_hunyuan_vl::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    const bool use_mrope = hparams.use_mrope();
+
+    int sections[4];
+    std::copy(std::begin(hparams.rope_sections), std::begin(hparams.rope_sections) + 4, sections);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale = 1.0f / sqrtf(float(n_embd_head));
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (use_mrope) {
+                Qcur = ggml_rope_multi(
+                            ctx0, Qcur, inp_pos, rope_factors,
+                            n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow
+                            );
+
+                Kcur = ggml_rope_multi(
+                            ctx0, Kcur, inp_pos, rope_factors,
+                            n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow
+                            );
+            } else {
+                Qcur = ggml_rope_ext(
+                            ctx0, Qcur, inp_pos, rope_factors,
+                            n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow
+                            );
+
+                Kcur = ggml_rope_ext(
+                            ctx0, Kcur, inp_pos, rope_factors,
+                            n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                            ext_factor, attn_factor, beta_fast, beta_slow
+                            );
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            Kcur = build_norm(Kcur,
+                        model.layers[il].attn_k_norm, nullptr,
+                        LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_norm", il);
+
+            Qcur = build_norm(Qcur,
+                        model.layers[il].attn_q_norm, nullptr,
+                        LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_norm", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+        // feed-forward network (non-MoE)
+        ggml_tensor * cur_mlp = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur_mlp, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur_mlp, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/hy-v3.cpp

@@ -0,0 +1,390 @@
+#include "models.h"
+
+void llama_model_hy_v3::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,                hparams.expert_gating_func, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,              hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,               hparams.expert_weights_norm, false);
+
+    // HY V3 uses a sigmoid router with expert selection bias by default
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+    }
+
+    // NextN/MTP (HY V3): extra decoder block(s) appended beyond the main stack
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_all");
+
+    switch (hparams.n_layer()) {
+        case 48: type = LLM_TYPE_30B_A3B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_hy_v3::load_arch_tensors(llama_model_loader & ml) {
+    LLAMA_LOAD_LOCALS;
+
+    const bool mtp_only = (hparams.n_layer_nextn > 0) && (ml.get_weight("blk.0.attn_norm.weight") == nullptr);
+    // Trunk-only: the GGUF declares MTP layers in metadata but the actual MTP
+    // tensors live in a separate file (e.g. user split target/draft). Mark
+    // MTP tensors NOT_REQUIRED so the trunk loads cleanly.
+    const std::string mtp_probe = "blk." + std::to_string(n_layer) + ".nextn.eh_proj.weight";
+    const bool trunk_only = (hparams.n_layer_nextn > 0) && (ml.get_weight(mtp_probe.c_str()) == nullptr);
+    const int trunk_flags = mtp_only   ? TENSOR_NOT_REQUIRED : 0;
+    const int mtp_flags   = trunk_only ? TENSOR_NOT_REQUIRED : 0;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    auto load_block = [&](int i, int flags) {
+        auto & layer = layers[i];
+        const int64_t n_ff_exp   = hparams.n_ff_exp   ? hparams.n_ff_exp   : n_ff / (n_expert_used > 0 ? n_expert_used : 1);
+        const int64_t n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff_exp;
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, flags);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, flags);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, flags);
+
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, flags);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, flags);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, flags);
+
+        // dense FFN (leading dense blocks, first_k_dense_replace)
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, TENSOR_NOT_REQUIRED);
+
+        // MoE routed experts (sigmoid router + expert selection bias)
+        layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,    "weight", i), {n_embd, n_expert}, TENSOR_NOT_REQUIRED);
+        layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B,           i), {n_expert}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS,   "weight", i), {n_ff_exp, n_embd, n_expert}, TENSOR_NOT_REQUIRED);
+        create_tensor_gate_up_exps(layer, i, n_embd, n_ff_exp, n_expert, TENSOR_NOT_REQUIRED);
+
+        // shared expert (always active, no gate)
+        layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, TENSOR_NOT_REQUIRED);
+        layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, TENSOR_NOT_REQUIRED);
+    };
+
+    for (int i = 0; i < n_layer; ++i) {
+        load_block(i, trunk_flags);
+    }
+
+    // NextN/MTP block(s): a full hy_v3 decoder block plus the NextN projections.
+    for (int i = n_layer; i < n_layer_all; ++i) {
+        auto & layer = layers[i];
+
+        load_block(i, mtp_flags);
+
+        layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ,          "weight", i), { 2 * n_embd, n_embd }, mtp_flags);
+        layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM,            "weight", i), { n_embd },             mtp_flags);
+        layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM,            "weight", i), { n_embd },             mtp_flags);
+        layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS,     "weight", i), { n_embd, n_vocab },    TENSOR_NOT_REQUIRED);
+        layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab },    TENSOR_NOT_REQUIRED);
+        // hy_v3 stores the MTP block's trailing final_layernorm here (applied
+        // after the decoder block, before the shared LM head).
+        layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd },             TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_hy_v3::build_arch_graph(const llm_graph_params & params) const {
+    if (params.gtype == LLM_GRAPH_TYPE_DECODER_MTP) {
+        return std::make_unique<graph_mtp>(*this, params);
+    }
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_hy_v3::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+    ggml_tensor * inp_pos = build_inp_pos();
+    auto * inp_attn = build_attn_inp_kv();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    const float kq_scale = 1.0f / sqrtf(float(n_embd_head));
+
+    // MTP/NextN layers are loaded as extra decoder blocks but not executed in the main pass.
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur, n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, nullptr, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, nullptr, LLM_NORM_RMS, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, nullptr, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+            // dense FFN (leading dense blocks)
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   model.layers[il].ffn_up_s,
+                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, model.layers[il].ffn_gate_s,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, model.layers[il].ffn_down_s,
+                    nullptr,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_dense_out", il);
+        } else {
+            // MoE routed experts (sigmoid gating + expert selection bias)
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU,
+                    hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il,
+                    nullptr, model.layers[il].ffn_gate_up_exps,
+                    model.layers[il].ffn_up_exps_s,
+                    model.layers[il].ffn_gate_exps_s,
+                    model.layers[il].ffn_down_exps_s);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // shared expert (always active, no gate)
+            ggml_tensor * sh_out = build_ffn(cur,
+                    model.layers[il].ffn_up_shexp,   nullptr, model.layers[il].ffn_up_shexp_s,
+                    model.layers[il].ffn_gate_shexp, nullptr, model.layers[il].ffn_gate_shexp_s,
+                    model.layers[il].ffn_down_shexp, nullptr, model.layers[il].ffn_down_shexp_s,
+                    nullptr,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(sh_out, "ffn_shared_out", il);
+
+            cur = ggml_add(ctx0, moe_out, sh_out);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        inpL = cur;
+    }
+
+    cur = build_norm(inpL, model.output_norm, nullptr, LLM_NORM_RMS, -1);
+
+    // Post-final-norm hidden state: what the MTP draft head's hnorm consumes.
+    // vLLM feeds the target model's normed output states, and the MTP layer
+    // itself returns final_layernorm(h), so the chained state is post-norm.
+    cb(cur, "h_nextn", -1);
+    res->t_h_nextn = cur;
+
+    if (!cparams.embeddings_nextn_masked && inp_out_ids) {
+        cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+    }
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// LLM_GRAPH_TYPE_DECODER_MTP draft head for HY V3 (MoE).
+// Semantics mirror vLLM's HYV3MultiTokenPredictorLayer (hy_v3_mtp.py):
+//   enorm(embed) + hnorm(prev_hidden) -> concat(e, h) -> eh_proj ->
+//   hy_v3 decoder block -> final_layernorm (stored as nextn.shared_head_norm) ->
+//   shared LM head (the main model's lm_head; the checkpoint has no separate
+//   MTP head or MTP embeddings).
+llama_model_hy_v3::graph_mtp::graph_mtp(const llama_model & model, const llm_graph_params & params)
+    : llm_graph_context(params) {
+    GGML_ASSERT(hparams.n_layer_nextn > 0 && "HY_V3 MTP requires n_layer_nextn > 0");
+
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    const int il = hparams.n_layer() + cparams.nextn_layer_offset;
+    GGML_ASSERT(cparams.nextn_layer_offset >= 0 &&
+                cparams.nextn_layer_offset < (int) hparams.n_layer_nextn &&
+                "nextn_layer_offset out of range [0, n_layer_nextn)");
+    const auto & layer = model.layers[il];
+
+    GGML_ASSERT(layer.nextn.eh_proj && "MTP block missing nextn.eh_proj");
+    GGML_ASSERT(layer.nextn.enorm   && "MTP block missing nextn.enorm");
+    GGML_ASSERT(layer.nextn.hnorm   && "MTP block missing nextn.hnorm");
+
+    auto inp = std::make_unique<llm_graph_input_embd>(hparams.n_embd);
+
+    inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
+    ggml_set_input(inp->tokens);
+
+    inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd, n_tokens);
+    ggml_set_input(inp->embd);
+    ggml_set_name(inp->embd, "mtp_h_input");
+
+    ggml_tensor * tok_embd_w = layer.nextn.embed_tokens ? layer.nextn.embed_tokens : model.tok_embd;
+
+    ggml_tensor * h_input  = inp->embd;
+    ggml_tensor * tok_embd = ggml_get_rows(ctx0, tok_embd_w, inp->tokens);
+    cb(tok_embd, "mtp_tok_embd", il);
+
+    res->add_input(std::move(inp));
+
+    ggml_tensor * inp_pos     = build_inp_pos();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+    auto * inp_attn           = build_attn_inp_kv();
+
+    ggml_tensor * h_norm = build_norm(h_input, layer.nextn.hnorm, nullptr, LLM_NORM_RMS, il);
+    cb(h_norm, "mtp_hnorm", il);
+
+    ggml_tensor * e_norm = build_norm(tok_embd, layer.nextn.enorm, nullptr, LLM_NORM_RMS, il);
+    cb(e_norm, "mtp_enorm", il);
+
+    ggml_tensor * concat = ggml_concat(ctx0, e_norm, h_norm, /*dim=*/ 0);
+    cb(concat, "mtp_concat", il);
+
+    ggml_tensor * cur = build_lora_mm(layer.nextn.eh_proj, concat);
+    cb(cur, "mtp_eh_proj", il);
+
+    ggml_tensor * inpSA = cur;
+
+    // mtp_block: a full hy_v3 decoder layer (mirrors the trunk graph)
+    cur = build_norm(cur, layer.attn_norm, nullptr, LLM_NORM_RMS, il);
+    cb(cur, "mtp_attn_norm", il);
+
+    {
+        ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+        auto [Qcur, Kcur, Vcur] = build_qkv(layer, cur, n_embd_head, n_head, n_head_kv, il);
+
+        Qcur = build_norm(Qcur, layer.attn_q_norm, nullptr, LLM_NORM_RMS, il);
+        Kcur = build_norm(Kcur, layer.attn_k_norm, nullptr, LLM_NORM_RMS, il);
+
+        Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow);
+        Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow);
+
+        const float kq_scale = 1.0f / sqrtf(float(n_embd_head));
+
+        cur = build_attn(inp_attn,
+                layer.wo, layer.wo_b, layer.wo_s,
+                Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+        cb(cur, "mtp_attn_out", il);
+    }
+
+    ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+    cb(ffn_inp, "mtp_ffn_inp", il);
+
+    cur = build_norm(ffn_inp, layer.ffn_norm, nullptr, LLM_NORM_RMS, il);
+    cb(cur, "mtp_ffn_norm", il);
+
+    if (layer.ffn_gate_inp == nullptr) {
+        cur = build_ffn(cur,
+                layer.ffn_up,   layer.ffn_up_b,   layer.ffn_up_s,
+                layer.ffn_gate, layer.ffn_gate_b, layer.ffn_gate_s,
+                layer.ffn_down, layer.ffn_down_b, layer.ffn_down_s,
+                nullptr,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "mtp_ffn_dense_out", il);
+    } else {
+        ggml_tensor * moe_out = build_moe_ffn(cur,
+                layer.ffn_gate_inp,
+                layer.ffn_up_exps,
+                layer.ffn_gate_exps,
+                layer.ffn_down_exps,
+                layer.ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU,
+                hparams.expert_weights_norm,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il,
+                nullptr, layer.ffn_gate_up_exps,
+                layer.ffn_up_exps_s,
+                layer.ffn_gate_exps_s,
+                layer.ffn_down_exps_s);
+        cb(moe_out, "mtp_ffn_moe_out", il);
+
+        ggml_tensor * sh_out = build_ffn(cur,
+                layer.ffn_up_shexp,   nullptr, layer.ffn_up_shexp_s,
+                layer.ffn_gate_shexp, nullptr, layer.ffn_gate_shexp_s,
+                layer.ffn_down_shexp, nullptr, layer.ffn_down_shexp_s,
+                nullptr,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(sh_out, "mtp_ffn_shared_out", il);
+
+        cur = ggml_add(ctx0, moe_out, sh_out);
+        cb(cur, "mtp_ffn_out", il);
+    }
+
+    cur = ggml_add(ctx0, cur, ffn_inp);
+    cb(cur, "mtp_post_ffn", il);
+
+    // final_layernorm applied after the decoder block, before the shared head.
+    // The post-norm hidden state seeds the next MTP step (matches vLLM, where
+    // HYV3MultiTokenPredictorLayer returns final_layernorm(h)).
+    ggml_tensor * head_norm_w = layer.nextn.shared_head_norm
+            ? layer.nextn.shared_head_norm
+            : model.output_norm;
+    GGML_ASSERT(head_norm_w && "HY_V3 MTP: missing both nextn.shared_head_norm and output_norm");
+    cur = build_norm(cur, head_norm_w, nullptr, LLM_NORM_RMS, -1);
+
+    cb(cur, "h_nextn", -1);
+    res->t_h_nextn = cur;
+
+    cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+    cb(cur, "mtp_shared_head_norm", -1);
+
+    ggml_tensor * head_w = layer.nextn.shared_head_head ? layer.nextn.shared_head_head : model.output;
+    ggml_tensor * head_s = layer.nextn.shared_head_head ? layer.nextn.shared_head_head_s : model.output_s;
+    GGML_ASSERT(head_w && "HY_V3 MTP: missing LM head (nextn.shared_head_head or model.output)");
+    cur = build_lora_mm(head_w, cur, head_s);
+    cb(cur, "result_output", -1);
+
+    res->t_logits = cur;
+    ggml_build_forward_expand(gf, cur);
+}

src/models/internlm2.cpp

@@ -0,0 +1,139 @@
+#include "models.h"
+
+void llama_model_internlm2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_7B; break;
+        case 48: type = LLM_TYPE_20B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_internlm2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        // layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_internlm2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_internlm2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/jais.cpp

@@ -0,0 +1,132 @@
+#include "models.h"
+
+void llama_model_jais::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+    ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias, false);
+
+    switch (hparams.n_layer()) {
+        case 24: type = LLM_TYPE_1_3B; break;
+        case 40: type = LLM_TYPE_13B; break;
+        /* TODO: add variants */
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_jais::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "bias", i),   {n_embd}, 0);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wqkv_b = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, 0);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+        layer.ffn_gate   = create_tensor(tn(LLM_TENSOR_FFN_GATE,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE,   "bias", i),   {n_ff}, 0);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_jais::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_jais::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/float(n_embd_head), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        // add the input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FF
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = build_norm(inpL,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/jais2.cpp

@@ -0,0 +1,161 @@
+#include "models.h"
+
+void llama_model_jais2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_8B; break;
+        case 68: type = LLM_TYPE_70B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_jais2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}, 0);
+        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // attention biases - all have shape n_embd (output dimension of projections)
+        layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}, 0);
+        layer.wk_b = create_tensor(tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd}, 0);
+        layer.wv_b = create_tensor(tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+        // Jais-2 uses simple MLP (no gate) with biases
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_jais2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+// JAIS-2 model graph builder
+// Uses: LayerNorm (not RMSNorm), relu2 activation, separate Q/K/V, RoPE embeddings
+llama_model_jais2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // KV input for attention
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        // Pre-attention LayerNorm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // Self-attention with separate Q, K, V projections
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // Apply RoPE
+            Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+            );
+
+            Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+            );
+
+            cb(Qcur, "Qcur_rope", il);
+            cb(Kcur, "Kcur_rope", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // Residual connection
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // Pre-FFN LayerNorm
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm,
+                model.layers[il].ffn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "ffn_norm", il);
+
+        // FFN with relu2 activation (ReLU squared) - no gate projection
+        // up -> relu2 -> down
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                NULL, NULL, NULL,  // no gate
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL,
+                LLM_FFN_RELU_SQR, LLM_FFN_SEQ, il);
+        cb(cur, "ffn_out", il);
+
+        // Residual connection
+        inpL = ggml_add(ctx0, cur, ffn_inp);
+        inpL = build_cvec(inpL, il);
+        cb(inpL, "l_out", il);
+    }
+
+    // Final LayerNorm
+    cur = build_norm(inpL,
+            model.output_norm,
+            model.output_norm_b,
+            LLM_NORM, -1);
+    cb(cur, "result_norm", -1);
+
+    res->t_embd = cur;
+
+    // Output projection
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/jamba.cpp

@@ -0,0 +1,198 @@
+#include "models.h"
+
+void llama_model_jamba::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
+    ml.get_key(LLM_KV_SSM_STATE_SIZE,     hparams.ssm_d_state);
+    ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    for (uint32_t i = 0; i < hparams.n_layer(); ++i) {
+        hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0;
+    }
+
+    switch (hparams.n_layer()) {
+        // TODO: Jamba layers are a bit heterogeneous, so naming this is hard.
+        case 12: // 900M  8x???M
+        case 32: // 51B  16x?B
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_jamba::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t d_conv  = hparams.ssm_d_conv;
+    const int64_t d_inner = hparams.ssm_d_inner;
+    const int64_t d_state = hparams.ssm_d_state;
+    const int64_t dt_rank = hparams.ssm_dt_rank;
+
+    // only an expansion factor of 2 is supported for now
+    GGML_ASSERT(2 * n_embd == d_inner);
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    {
+        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+        output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+        // if output is NULL, init from the input tok embed, duplicated to allow offloading
+        if (output == NULL) {
+            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+        }
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        const int64_t n_head_kv = hparams.n_head_kv(i);
+        const int64_t n_embd_gqa = hparams.n_embd_v_gqa(i);
+
+        auto & layer = layers[i];
+
+        // norm
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (n_head_kv == 0) {
+            // Mamba layer
+            layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, 2*d_inner}, 0);
+
+            layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner}, 0);
+            layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner}, 0);
+
+            layer.ssm_x = create_tensor(tn(LLM_TENSOR_SSM_X, "weight", i), {d_inner, dt_rank + 2*d_state}, 0);
+
+            layer.ssm_dt_norm = create_tensor(tn(LLM_TENSOR_SSM_DT_NORM, "weight", i), {dt_rank}, 0);
+
+            layer.ssm_dt = create_tensor(tn(LLM_TENSOR_SSM_DT, "weight", i), {dt_rank, d_inner}, 0);
+            layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {d_inner}, 0);
+
+            layer.ssm_b_norm = create_tensor(tn(LLM_TENSOR_SSM_B_NORM, "weight", i), {d_state}, 0);
+            layer.ssm_c_norm = create_tensor(tn(LLM_TENSOR_SSM_C_NORM, "weight", i), {d_state}, 0);
+
+            // no "weight" suffix for these
+            layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {d_state, d_inner}, 0);
+            layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {d_inner}, 0);
+
+            // out_proj
+            layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd}, 0);
+        } else {
+            // Attention layers
+
+            create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        }
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, TENSOR_NOT_REQUIRED);
+
+        if (layer.ffn_gate_inp) {
+            // MoE
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff, n_expert}, 0);
+        } else {
+            // FFN (no MoE)
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_jamba::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_jamba::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_build_mamba_base(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    // {n_embd, n_tokens}
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * inp_hybrid = build_inp_mem_hybrid();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const int64_t n_head_kv = hparams.n_head_kv(il);
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        if (n_head_kv == 0) {
+            cur = build_mamba_layer(inp_hybrid->get_recr(), cur, model, ubatch, il);
+        } else {
+            // Attention
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // No RoPE :)
+            cur = build_attn(inp_hybrid->get_attn(),
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, NULL, NULL, NULL, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        // residual
+        struct ggml_tensor * ffn_inp = ggml_add(ctx0, inpL, cur);
+        cb(cur, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+            // FFN
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            cur = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, false,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il);
+            cb(cur, "ffn_moe_out", il);
+        }
+        // residual
+        cur = ggml_add(ctx0, ffn_inp, cur);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    // final rmsnorm
+    cur = build_norm(inpL, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/jina-bert-v2.cpp

@@ -0,0 +1,66 @@
+#include "models.h"
+
+void llama_model_jina_bert_v2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,    hparams.f_norm_eps);
+    hparams.f_max_alibi_bias = 8.0f;
+
+    switch (hparams.n_layer()) {
+        case 4:  type = LLM_TYPE_33M;  break; // jina-embeddings-small
+        case 12: type = LLM_TYPE_137M; break; // jina-embeddings-base
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_jina_bert_v2::load_arch_tensors(llama_model_loader & ml) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd  = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0); // word_embeddings
+    type_embd = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, 0); // token_type_embeddings
+
+    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0); // LayerNorm
+    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias",   0), {n_embd}, 0); // LayerNorm bias
+
+    cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, 1}, TENSOR_NOT_REQUIRED);
+    cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {1},         TENSOR_NOT_REQUIRED);
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i]; // JinaBertLayer
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.attn_q_norm   = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.attn_k_norm   = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0); //output_dens
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0); //output_dens
+
+        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0); //output_norm
+        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias",   i), {n_embd}, 0);
+
+        layer.attn_norm_2   = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.attn_norm_2_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, TENSOR_NOT_REQUIRED);
+
+        const auto tn_ffn_up_weight = tn(LLM_TENSOR_FFN_UP, "weight", i);
+        ggml_tensor * t_ffn_up = ml.get_tensor_meta(tn_ffn_up_weight.str().c_str());
+        const int64_t n_ffn_up = t_ffn_up ? t_ffn_up->ne[1] : n_ff;
+
+        GGML_ASSERT(n_ffn_up == n_ff || n_ffn_up == n_ff * 2);
+        layer.ffn_up   = create_tensor(tn_ffn_up_weight, {n_embd, n_ffn_up}, 0);
+        layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ffn_up}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, 0);
+
+        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias",   i), {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_jina_bert_v2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/jina-bert-v3.cpp

@@ -0,0 +1,69 @@
+#include "models.h"
+
+void llama_model_jina_bert_v3::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,    hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 24:
+            type = LLM_TYPE_558M; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_jina_bert_v3::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_token_types == 0) {
+        throw std::runtime_error(arch_name() + " model needs to define token type count");
+    }
+    tok_embd     = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0);
+    type_embd    = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, TENSOR_NOT_REQUIRED);
+
+    if (arch == LLM_ARCH_BERT) {
+        pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,    "weight"), {n_embd, n_ctx_train}, 0);
+
+        cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+        cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {n_embd},         TENSOR_NOT_REQUIRED);
+
+        cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+        cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+    }
+
+    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0);
+    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias",   0), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i),   {n_embd}, 0);
+
+        if (hparams.moe_every_n_layers > 0 && i % hparams.moe_every_n_layers == 1) {
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff,   n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff,   n_embd, n_expert}, 0);
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,   "weight", i), {n_embd, n_expert}, 0);
+        } else {
+            layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+            if (arch == LLM_ARCH_NOMIC_BERT) {
+                layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            }
+        }
+
+        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i),   {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_jina_bert_v3::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/kimi-linear.cpp

@@ -0,0 +1,550 @@
+#include "models.h"
+#include "llama-memory-recurrent.h"
+
+void llama_model_kimi_linear::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH_MLA,    hparams.n_embd_head_k_mla_impl);
+    ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_MLA,  hparams.n_embd_head_v_mla_impl);
+    ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK,      hparams.n_lora_kv);
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,             hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_KDA_HEAD_DIM,                hparams.n_embd_head_kda);
+
+    // MLA qk_rope_head_dim (for reference)
+    // qk_rope_head_dim = 64, qk_nope_head_dim = 128, qk_head_dim = 192
+
+    // Mark KDA layers as recurrent using n_head_kv pattern (like Jamba)
+    // Set n_head_kv = 0 for KDA layers (recurrent), n_head_kv = n_head for MLA layers (attention)
+    for (uint32_t i = 0; i < hparams.n_layer(); ++i) {
+        hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0;  // KDA layers are recurrent
+    }
+
+    // MoE parameters - Kimi uses moe_intermediate_size = 1024
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,               hparams.n_expert_shared);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,         hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,              hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,                hparams.expert_gating_func);
+
+    switch (hparams.n_layer()) {
+        case 27: type = LLM_TYPE_48B_A3B; break; // Kimi-Linear-48B-A3B
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_kimi_linear::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        // Check for KDA specific tensors to determine layer type or if it's a mixed model
+        // Assuming KDA layer if KDA tensors are present
+
+        // KDA uses head_dim = 128 (from linear_attn_config.head_dim)
+        const int64_t n_embd_head_k_kda = hparams.n_embd_head_kda;
+        const int64_t n_embd_head_v_kda = hparams.n_embd_head_kda;
+        const int64_t ssm_d_conv = hparams.ssm_d_conv;
+
+        if (hparams.is_recr(i)) {
+            // Conv1d weights: try 4D first, then 3D (quantization may remove trailing 1)
+            // 4D: [d_conv, 1, d_inner, 1], 3D: [d_conv, 1, d_inner]
+            layer.ssm_q_conv = create_tensor(tn(LLM_TENSOR_SSM_CONV1D_Q, "weight", i), {ssm_d_conv, 1, n_embd_head_k_kda * n_head, 1}, TENSOR_NOT_REQUIRED);
+            if (!layer.ssm_q_conv) {
+                layer.ssm_q_conv = create_tensor(tn(LLM_TENSOR_SSM_CONV1D_Q, "weight", i), {ssm_d_conv, 1, n_embd_head_k_kda * n_head}, 0);
+            }
+
+             // KDA Layer - Conv1d weights may be 3D or 4D
+             layer.ssm_k_conv = create_tensor(tn(LLM_TENSOR_SSM_CONV1D_K, "weight", i), {ssm_d_conv, 1, n_embd_head_k_kda * n_head, 1}, TENSOR_NOT_REQUIRED);
+             if (!layer.ssm_k_conv) {
+                 layer.ssm_k_conv = create_tensor(tn(LLM_TENSOR_SSM_CONV1D_K, "weight", i), {ssm_d_conv, 1, n_embd_head_k_kda * n_head}, 0);
+             }
+             layer.ssm_v_conv = create_tensor(tn(LLM_TENSOR_SSM_CONV1D_V, "weight", i), {ssm_d_conv, 1, n_embd_head_v_kda * n_head, 1}, TENSOR_NOT_REQUIRED);
+             if (!layer.ssm_v_conv) {
+                 layer.ssm_v_conv = create_tensor(tn(LLM_TENSOR_SSM_CONV1D_V, "weight", i), {ssm_d_conv, 1, n_embd_head_v_kda * n_head}, 0);
+             }
+
+             // q, k, v projections
+             // Python: q_proj, k_proj, v_proj
+             create_tensor_qkv(layer, i, n_embd, n_embd_head_k_kda * n_head, n_embd_head_k_kda * n_head, n_embd_head_v_kda * n_head, 0);
+
+             // KDA specific projections
+             // f_a_proj, f_b_proj
+             layer.ssm_f_a = create_tensor(tn(LLM_TENSOR_SSM_F_A, "weight", i), {n_embd, n_embd_head_k_kda}, 0); // head_dim
+             layer.ssm_f_b = create_tensor(tn(LLM_TENSOR_SSM_F_B, "weight", i), {n_embd_head_k_kda, n_embd_head_k_kda * n_head}, 0); // projection_size
+
+             // b_proj (beta mixing coefficient)
+             layer.ssm_beta = create_tensor(tn(LLM_TENSOR_SSM_BETA, "weight", i), {n_embd, n_head}, 0);
+
+             // A_log - Shape in GGUF: [1, num_heads, 1, 1] (4D) or [1, num_heads] (2D after quantization) Note: -exp(A_log) is applied in convert_hf_to_gguf.py
+             layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {1, n_head, 1, 1}, TENSOR_NOT_REQUIRED);
+             if (!layer.ssm_a) {
+                 layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {1, n_head}, 0);
+             }
+
+             // dt_bias - shape [n_embd_head_k_kda * n_head] = [4096]
+             layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {n_embd_head_k_kda * n_head}, 0);
+
+             // g_a_proj, g_b_proj (output gate)
+             layer.ssm_g_a = create_tensor(tn(LLM_TENSOR_SSM_G_A, "weight", i), {n_embd, n_embd_head_k_kda}, 0);
+             layer.ssm_g_b = create_tensor(tn(LLM_TENSOR_SSM_G_B, "weight", i), {n_embd_head_k_kda, n_embd_head_k_kda * n_head}, 0);
+
+             // o_norm (reusing SSM_NORM)
+             layer.ssm_o_norm = create_tensor(tn(LLM_TENSOR_SSM_NORM, "weight", i), {n_embd_head_k_kda}, 0); // FusedRMSNormGated
+
+             // o_proj
+             layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_v_kda * n_head, n_embd}, 0);
+
+        } else {
+             // MLA Layer - use MLA-specific head dimensions
+             const int64_t q_lora_rank  = hparams.n_lora_q;
+             const int64_t kv_lora_rank = hparams.n_lora_kv;
+             const int64_t n_embd_head_k_mla = hparams.n_embd_head_k_mla();
+             const int64_t n_embd_head_v_mla = hparams.n_embd_head_v_mla();
+
+             layer.attn_q_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, TENSOR_NOT_REQUIRED);
+             layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, 0);
+
+             if (layer.attn_q_a_norm) {
+                 layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, 0);
+                 layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k_mla}, 0);
+             } else {
+                 // Kimi MLA without Q compression: wq = [n_embd, n_head * n_embd_head_k_mla]
+                 layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_head * n_embd_head_k_mla}, 0);
+             }
+
+             // Kimi: qk_rope_head_dim = 64 (actual RoPE dimension for MLA)
+             // Note: hparams.n_rot may be 72 (from conversion) but actual is 64
+             const int64_t qk_rope_head_dim = hparams.n_rot();  // From config: qk_rope_head_dim
+             layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + qk_rope_head_dim}, 0);
+             // Support Legacy GGUFs that don't split wkv_b (MLA KV cache disabled)
+             layer.wkv_b = create_tensor(tn(LLM_TENSOR_ATTN_KV_B, "weight", i),
+                {kv_lora_rank, n_head * (n_embd_head_k_mla - qk_rope_head_dim + n_embd_head_v_mla)}, TENSOR_NOT_REQUIRED | TENSOR_SKIP_IF_VIRTUAL);
+             if (!layer.wkv_b) { // MLA KV cache enabled
+                 layer.wk_b = create_tensor(tn(LLM_TENSOR_ATTN_K_B, "weight", i), {n_embd_head_k_mla - qk_rope_head_dim, kv_lora_rank, n_head}, 0);
+                 layer.wv_b = create_tensor(tn(LLM_TENSOR_ATTN_V_B, "weight", i), {kv_lora_rank, n_embd_head_v_mla, n_head}, 0);
+             }
+             layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_embd_head_v_mla, n_embd}, 0);
+        }
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        // MoE intermediate size (different from dense FFN)
+        const int64_t n_ff_exp = hparams.n_ff_exp;
+
+        // Kimi uses n_layer_dense_lead to determine which layers use dense FFN vs MoE
+        // first_k_dense_replace = 1 means layer 0 uses dense FFN, layers 1+ use MoE
+        if (i < (int) hparams.n_layer_dense_lead) {
+            // Dense FFN layer - use normal n_ff
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+        } else {
+            // MoE layer - use n_ff_exp (1024) instead of n_ff (9216)
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+
+            // Shared experts use moe_intermediate_size * num_shared_experts
+            // Kimi: shared_expert_intermediate_size = 1024 * 1 = 1024
+            // Tensors are 2D: [n_embd, n_ff_shexp] or [n_ff_shexp, n_embd]
+            const int64_t n_ff_shexp_actual = n_ff_exp * (hparams.n_expert_shared > 0 ? hparams.n_expert_shared : 1);
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp_actual}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp_actual, n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp_actual}, TENSOR_NOT_REQUIRED);
+
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_kimi_linear::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+// Causal Conv1d function for Q,K,V
+// When qkv is 0, it is Q, 1 is K, 2 is V
+static ggml_tensor * causal_conv1d(ggml_cgraph * gf, ggml_context * ctx0, ggml_tensor * conv_states_all, ggml_tensor * conv_state_all, int64_t qkv, ggml_tensor * x, ggml_tensor * proj_w, ggml_tensor * conv_w, int64_t d_conv, int64_t head_dim, int64_t n_head, int64_t n_seq_tokens, int64_t n_seqs, int64_t n_tokens, int64_t kv_head) {
+    const int64_t d_inner = head_dim * n_head;
+    const int64_t conv_state_size = (d_conv - 1) * d_inner;
+    const int64_t n_embd_r_total = 3 * conv_state_size;  // Q + K + V
+
+    // conv_state_all is [n_embd_r_total, n_seqs], split into Q, K, V
+    // Each conv state is [(d_conv-1) * d_inner] per sequence, need to reshape to [d_conv-1, d_inner, n_seqs]
+    // Memory layout: for each seq, Q state is first conv_state_size elements, then K, then V
+    // conv_state_all has stride: nb[0] = element_size, nb[1] = n_embd_r_total * element_size
+    // View Q conv state: offset 0, size conv_state_size per seq
+    // conv_state_all is [n_embd_r_total, n_seqs] with memory layout:
+    //   state[i + seq * n_embd_r_total] where i = conv_step + channel * (d_conv-1) + {0, conv_state_size, 2*conv_state_size} for Q/K/V
+    // We want [d_conv-1, d_inner, n_seqs] view:
+    //   nb1 = (d_conv-1) * element_size (stride between channels)
+    //   nb2 = n_embd_r_total * element_size (stride between seqs)
+    ggml_tensor * conv_state_x = ggml_view_3d(ctx0, conv_state_all, d_conv - 1, d_inner, n_seqs,
+        (d_conv - 1) * ggml_element_size(conv_state_all),  // nb1: stride between channels
+        n_embd_r_total * ggml_element_size(conv_state_all),  // nb2: stride between seqs
+        qkv * conv_state_size * ggml_element_size(conv_state_all));
+
+// Causal Conv1d function for Q,K,V
+// When qkv is 0, it is Q, 1 is K, 2 is V
+    // Step 1: Q, K, V projections -> [d_inner, n_tokens]
+    ggml_tensor * x_proj = ggml_mul_mat(ctx0, proj_w, x);
+
+    // Reshape input: {d_inner, n_tokens} -> {d_inner, n_seq_tokens, n_seqs}
+    ggml_tensor * x_3d = ggml_reshape_3d(ctx0, x_proj, d_inner, n_seq_tokens, n_seqs);
+
+    // Concat Q conv state and current input: {d_conv-1 + n_seq_tokens, d_inner, n_seqs}
+    ggml_tensor * conv_x = ggml_concat(ctx0, conv_state_x, ggml_transpose(ctx0, x_3d), 0);
+
+    // Save last (d_conv-1) columns back to Q conv state
+    ggml_tensor * last_conv_x = ggml_view_3d(ctx0, conv_x, d_conv - 1, d_inner, n_seqs,
+        conv_x->nb[1], conv_x->nb[2], n_seq_tokens * conv_x->nb[0]);
+    ggml_build_forward_expand(gf,
+        ggml_cpy(ctx0, last_conv_x,
+            ggml_view_3d(ctx0, conv_states_all,
+                d_conv - 1, d_inner, n_seqs,
+                (d_conv - 1) * ggml_element_size(conv_states_all),           // nb1: contiguous within one channel's conv taps
+                n_embd_r_total * ggml_element_size(conv_states_all),         // nb2: stride between sequences (skip over K,V states)
+                (kv_head * n_embd_r_total + qkv * conv_state_size) * ggml_element_size(conv_states_all))));  // offset to first seq's Q/K/V state
+    // Reshape conv weight: GGUF [d_conv, 1, d_inner, 1] -> ggml_ssm_conv expects [d_conv, d_inner]
+    // GGUF stores as [d_conv, 1, d_inner, 1] with memory layout w[conv_step + channel * d_conv]
+    // vLLM stores as [d_inner, d_conv] with memory layout w[channel * d_conv + conv_step]
+    // ggml_ssm_conv computes: c[conv_step + channel * d_conv]
+    // GGUF layout: [d_conv, 1, d_inner] or [d_conv, 1, d_inner, 1] -> reshape to [d_conv, d_inner]
+    // Reshape conv weight from [d_conv, 1, d_inner, 1] to [d_conv, d_inner] for ggml_ssm_conv
+    ggml_tensor * conv_weight = ggml_reshape_2d(ctx0, conv_w, d_conv, d_inner);
+
+    // Apply conv1d
+    // ggml_ssm_conv output: {d_inner, n_seq_tokens, n_seqs}
+    ggml_tensor * Xcur = ggml_ssm_conv(ctx0, conv_x, conv_weight);
+    // Reshape to 2D for bias add: {d_inner, n_tokens}
+    Xcur = ggml_reshape_2d(ctx0, Xcur, d_inner, n_tokens);
+    Xcur = ggml_silu(ctx0, Xcur);
+
+    return ggml_reshape_4d(ctx0, Xcur, head_dim, n_head, n_seq_tokens, n_seqs);
+}
+
+llama_model_kimi_linear::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_build_delta_net_base(params), model(model) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+    cb(inpL, "model.embed_tokens", -1);
+
+    // Note: Kimi MLA does NOT use RoPE (rotary_emb=None in vLLM)
+    // So we don't need inp_pos
+
+    auto * inp_kv = !hparams.is_mla() ? build_inp_mem_hybrid() : nullptr;
+    auto * inp_k = hparams.is_mla() ? build_inp_mem_hybrid_k() : nullptr;
+    auto * inp_rs = hparams.is_mla() ? inp_k->get_recr() : inp_kv->get_recr();
+    auto * inp_attn_kv = !hparams.is_mla() ? inp_kv->get_attn() : nullptr;
+    auto * inp_attn_k = hparams.is_mla() ? inp_k->get_attn() : nullptr;
+
+    // Output ids for selecting which tokens to output
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    // Kimi dimension constants
+    const int64_t n_head = hparams.n_head();
+    const int64_t head_dim = hparams.n_embd_head_kda;
+    const int64_t d_conv = hparams.ssm_d_conv;
+    const int64_t d_inner = n_head * head_dim;  // 32 * 128 = 4096
+    const int64_t n_seqs = ubatch.n_seqs;
+    const int64_t n_seq_tokens = ubatch.n_seq_tokens;
+
+    // Verify batch consistency for recurrent layers
+    GGML_ASSERT(n_seqs != 0);
+    GGML_ASSERT(ubatch.equal_seqs());
+    GGML_ASSERT(ubatch.n_tokens == n_seq_tokens * n_seqs);
+
+    // MLA params
+    const int64_t n_embd_head_k_mla = hparams.n_embd_head_k_mla();
+    const int64_t n_embd_head_v_mla = hparams.n_embd_head_v_mla();
+    const int64_t kv_lora_rank = hparams.n_lora_kv;
+    // qk_rope_head_dim = 64 (from Kimi config) which is hparams.n_rot
+    // Confirmed from tensor shape: wkv_a_mqa [2304, 576] = [n_embd, kv_lora_rank + qk_rope_head_dim]
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();  // config.qk_rope_head_dim
+    const int64_t n_embd_head_qk_nope = n_embd_head_k_mla - n_embd_head_qk_rope;  // 192 - 64 = 128
+    // Attention scale for MLA
+    const float kq_scale_mla = 1.0f / sqrtf((float)n_embd_head_k_mla);
+
+    for (int il = 0; il < n_layer; ++il) {
+        const auto & layer = model.layers[il];
+        ggml_tensor * inpSA = inpL;
+
+        // Attention Norm
+        cur = build_norm(inpL, layer.attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        ggml_build_forward_expand(gf, cur);
+
+        if (hparams.is_recr(il)) {
+            // === KDA Layer (Kimi Delta Attention) with Recurrent State ===
+            // Reference: vLLM kda.py
+            const auto * mctx_cur = inp_rs->mctx;
+            const auto kv_head = mctx_cur->get_head();
+
+            // Get conv states from r_l tensor (Q, K, V each have separate state)
+            ggml_tensor * conv_states_all = mctx_cur->get_r_l(il);
+            cb(conv_states_all, "conv_states_all", il);
+            ggml_tensor * conv_state_all = build_rs(inp_rs, conv_states_all, hparams.n_embd_r(), n_seqs);
+            ggml_tensor * Qcur = causal_conv1d(gf, ctx0, conv_states_all, conv_state_all, 0, cur, layer.wq, layer.ssm_q_conv, d_conv, head_dim, n_head, n_seq_tokens, n_seqs, n_tokens, kv_head);
+            ggml_tensor * Kcur = causal_conv1d(gf, ctx0, conv_states_all, conv_state_all, 1, cur, layer.wk, layer.ssm_k_conv, d_conv, head_dim, n_head, n_seq_tokens, n_seqs, n_tokens, kv_head);
+            ggml_tensor * Vcur = causal_conv1d(gf, ctx0, conv_states_all, conv_state_all, 2, cur, layer.wv, layer.ssm_v_conv, d_conv, head_dim, n_head, n_seq_tokens, n_seqs, n_tokens, kv_head);
+
+            // g1 = -exp(A_log) * softplus(f_b(f_a(x)) + dt_bias)
+            ggml_tensor * f_a = ggml_mul_mat(ctx0, layer.ssm_f_a, cur);
+            ggml_tensor * g1 = ggml_mul_mat(ctx0, layer.ssm_f_b, f_a);
+            cb(g1, "g1 f_b(f_a(cur))", il);
+            g1 = ggml_add(ctx0, g1, layer.ssm_dt_b);
+            g1 = ggml_softplus(ctx0, g1);
+            g1 = ggml_reshape_3d(ctx0, g1, head_dim, n_head, n_tokens);
+
+            // A_log shape is [1, n_head] or [1, n_head, 1, 1], need to broadcast to [head_dim, n_head, n_tokens]. No need to -exp(a_log) because it was done in convert_hf_to_gguf.py
+            // Reshape to [1, n_head, 1] for broadcasting with g1 [head_dim, n_head, n_tokens]
+            ggml_tensor * A = ggml_reshape_3d(ctx0, layer.ssm_a, 1, n_head, 1);
+            g1 = ggml_mul(ctx0, g1, A);
+            cb(g1, "kda_g1", il);
+
+            g1 = ggml_reshape_4d(ctx0, g1, head_dim, n_head, n_seq_tokens, n_seqs);
+
+            // Compute beta (mixing coefficient)
+            ggml_tensor * beta = ggml_mul_mat(ctx0, layer.ssm_beta, cur);
+            beta = ggml_reshape_4d(ctx0, beta, 1, n_head, n_seq_tokens, n_seqs);
+            cb(beta, "kda_beta", il);
+
+            beta = ggml_sigmoid(ctx0, beta);
+
+            // Reshape for KDA recurrence
+            // {n_embd, n_tokens} -> {n_embd, n_seq_tokens, n_seqs}
+            cur = ggml_reshape_3d(ctx0, cur, cur->ne[0], n_seq_tokens, n_seqs);
+
+            // Get SSM state and compute KDA recurrence using ggml_kda_scan
+            ggml_tensor * ssm_states_all = mctx_cur->get_s_l(il);
+            ggml_tensor * state = build_rs(inp_rs, ssm_states_all, hparams.n_embd_s(), n_seqs);
+            state = ggml_reshape_4d(ctx0, state, head_dim, head_dim, n_head, n_seqs);
+
+            const float eps_norm = hparams.f_norm_rms_eps;
+
+            Qcur = ggml_l2_norm(ctx0, Qcur, eps_norm);
+            Kcur = ggml_l2_norm(ctx0, Kcur, eps_norm);
+
+            // Choose between build_delta_net_chunking and build_delta_net_recurrent based on n_tokens
+            auto attn_out = build_delta_net(Qcur, Kcur, Vcur, g1, beta, state, il);
+
+            ggml_tensor * output = ggml_cont(ctx0, attn_out.first);
+            ggml_tensor * new_state = attn_out.second;
+            cb(output, "attn_output", il);
+            cb(new_state, "new_state", il);
+
+            // Update the recurrent states
+            ggml_build_forward_expand(gf,
+                                     ggml_cpy(ctx0, new_state,
+                                              ggml_view_1d(ctx0, ssm_states_all, hparams.n_embd_s() * n_seqs,
+                                                           kv_head * hparams.n_embd_s() * ggml_element_size(ssm_states_all))));
+
+            // Output gating g2 = g_b(g_a(x))
+            ggml_tensor * cur_2d = ggml_reshape_2d(ctx0, cur, cur->ne[0], n_seq_tokens * n_seqs);
+            ggml_tensor * g_a = ggml_mul_mat(ctx0, layer.ssm_g_a, cur_2d);
+            ggml_tensor * g2 = ggml_mul_mat(ctx0, layer.ssm_g_b, g_a);
+            cb(g2, "g2 g_b(g_a(cur_2d))", il);
+            g2 = ggml_reshape_3d(ctx0, g2, head_dim, n_head, n_seq_tokens * n_seqs);
+
+            // Apply o_norm with sigmoid gating
+            // Note: Kimi model uses sigmoid gating, not SiLU (despite FusedRMSNormGated default being swish)
+            // Formula: output = RMSNorm(x) * sigmoid(g)
+            ggml_tensor * attn_out_final = ggml_reshape_3d(ctx0, output, head_dim, n_head,  n_seq_tokens * n_seqs);
+            ggml_tensor * normed = build_norm(attn_out_final, layer.ssm_o_norm, nullptr, LLM_NORM_RMS, il);
+            cb(normed, "kda_normed", il);
+            ggml_tensor * gate = ggml_sigmoid(ctx0, g2);
+            ggml_tensor * gated = ggml_mul(ctx0, normed, gate);
+
+            // Output projection
+            gated = ggml_cont_2d(ctx0, gated, d_inner, n_tokens);
+            cur = ggml_mul_mat(ctx0, layer.wo, gated);
+            cb(cur, "kda_out", il);
+
+        } else {
+            // === MLA Layer (Multi-head Latent Attention) without KV Cache ===
+            // Reference: vLLM mla.py
+            // Step 1: Q projection and reshape
+            // vLLM Kimi: q = q_proj(hidden_states), then view as [n_tokens, n_head, qk_head_dim]
+            // Note: Kimi MLA does NOT use RoPE (rotary_emb=None in vLLM)
+            ggml_tensor * Qcur = ggml_mul_mat(ctx0, layer.wq, cur);
+
+            // Step 2: KV compression
+            // kv_cmpr_pe = kv_a_proj_with_mqa(hidden_states) -> [kv_lora_rank + qk_rope_head_dim, n_tokens]
+            ggml_tensor * kv_cmpr_pe = ggml_mul_mat(ctx0, layer.wkv_a_mqa, cur);
+
+            // Split: kv_cmpr = kv_lora[:kv_lora_rank], k_pe = kv_lora[kv_lora_rank:]
+            ggml_tensor * kv_cmpr = ggml_view_2d(ctx0, kv_cmpr_pe, kv_lora_rank, n_tokens,
+                ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope), 0);
+            ggml_tensor * k_pe = ggml_view_3d(ctx0, kv_cmpr_pe, n_embd_head_qk_rope, 1, n_tokens,
+                ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+                ggml_row_size(kv_cmpr_pe->type, kv_lora_rank));
+            // Note: Kimi MLA does NOT apply RoPE (rotary_emb=None in vLLM)
+            // k_pe is used directly without RoPE
+            // Normalize kv_c
+            kv_cmpr = build_norm(kv_cmpr, layer.attn_kv_a_norm, nullptr, LLM_NORM_RMS, il);
+
+            if (layer.wk_b && layer.wv_b) { // MLA KV cache enabled
+                // extract q_nope
+                ggml_tensor * q_nope =
+                    ggml_view_3d(ctx0, Qcur, n_embd_head_qk_nope, n_head, n_tokens, ggml_row_size(Qcur->type, n_embd_head_k_mla),
+                                 ggml_row_size(Qcur->type, n_embd_head_k_mla) * n_head, 0);
+                cb(q_nope, "q_nope", il);
+
+                // and {n_embd_head_qk_rope, n_head, n_tokens}
+                ggml_tensor * q_pe = ggml_view_3d(
+                    ctx0, Qcur, n_embd_head_qk_rope, n_head, n_tokens, ggml_row_size(Qcur->type, n_embd_head_k_mla),
+                    ggml_row_size(Qcur->type, n_embd_head_k_mla) * n_head, ggml_row_size(Qcur->type, n_embd_head_qk_nope));
+                cb(q_pe, "q_pe", il);
+
+                // {n_embd_head_qk_nope, n_tokens, n_head}
+                q_nope = ggml_permute(ctx0, q_nope, 0, 2, 1, 3);
+                cb(q_nope, "q_nope_perm", il);
+
+                // {n_embd_head_qk_nope, kv_lora_rank, n_head} x {n_embd_head_qk_nope, n_tokens, n_head}
+                ggml_tensor * q_nope_absorbed = ggml_mul_mat(ctx0, layer.wk_b, q_nope);
+                cb(q_nope_absorbed, "q_nope_absorbed", il);
+
+                // {kv_lora_rank, n_head, n_tokens}
+                q_nope_absorbed = ggml_permute(ctx0, q_nope_absorbed, 0, 2, 1, 3);
+                cb(q_nope_absorbed, "q_nope_absorbed_perm", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, n_head, n_tokens}
+                // note: rope must go first for in-place context shifting in build_rope_shift()
+                Qcur = ggml_concat(ctx0, q_nope_absorbed, q_pe, 0);
+                cb(Qcur, "Qcur", il);
+
+                kv_cmpr = ggml_reshape_3d(ctx0, kv_cmpr, kv_lora_rank, 1, n_tokens);
+                cb(kv_cmpr, "kv_cmpr_reshape", il);
+
+                // {n_embd_head_qk_rope + kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Kcur = ggml_concat(ctx0, kv_cmpr, k_pe, 0);
+                cb(Kcur, "Kcur", il);
+
+                // {kv_lora_rank, 1, n_tokens}
+                ggml_tensor * Vcur = kv_cmpr;
+                cb(Vcur, "Vcur", il);
+
+                cur = build_attn(inp_attn_k, layer.wo, NULL, layer.wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, layer.wv_b, kq_scale_mla, il);
+                cb(cur, "mla_out", il);
+            } else { // MLA KV cache disabled. Fall back to MHA KV cache.
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head_k_mla, n_head, n_tokens);
+                cb(Qcur, "mla_Q", il);
+                // KV decompression: kv = kv_b_proj(kv_c_normed)
+                ggml_tensor * kv = ggml_mul_mat(ctx0, layer.wkv_b, kv_cmpr);
+                const int64_t kv_per_head = n_embd_head_qk_nope + n_embd_head_v_mla;
+
+                // Split kv into k_nope and v
+                ggml_tensor * k_nope = ggml_view_3d(ctx0, kv, n_embd_head_qk_nope, n_head, n_tokens,
+                    ggml_row_size(kv->type, kv_per_head),
+                    ggml_row_size(kv->type, kv_per_head * n_head), 0);
+                ggml_tensor * Vcur = ggml_view_3d(ctx0, kv, n_embd_head_v_mla, n_head, n_tokens,
+                    ggml_row_size(kv->type, kv_per_head),
+                    ggml_row_size(kv->type, kv_per_head * n_head),
+                    ggml_row_size(kv->type, n_embd_head_qk_nope));
+                Vcur = ggml_cont(ctx0, Vcur);
+                cb(Vcur, "mla_V", il);
+
+                // Concatenate k_nope + k_pe (broadcast k_pe to all heads)
+                // K = [k_nope, k_pe] where k_nope is [qk_nope_head_dim, n_head, n_tokens]
+                // and k_pe is [qk_rope_head_dim, 1, n_tokens] broadcast to all heads
+                // Need to broadcast k_pe from [qk_rope, 1, n_tokens] to [qk_rope, n_head, n_tokens]
+                ggml_tensor * k_pe_target = ggml_new_tensor_3d(ctx0, k_pe->type, n_embd_head_qk_rope, n_head, n_tokens);
+                ggml_tensor * k_pe_repeated = ggml_repeat(ctx0, k_pe, k_pe_target);
+                ggml_tensor * Kcur = ggml_concat(ctx0, k_pe_repeated, k_nope, 0);
+                cb(Kcur, "mla_K", il);
+
+                // Direct softmax attention (with MHA KV cache)
+                // Use build_attn with inp_attn for proper mask handling
+                cur = build_attn(inp_attn_kv, layer.wo, NULL, layer.wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale_mla, il);
+                cb(cur, "mla_out", il);
+            }
+        }
+
+        // On last layer, select only the output tokens
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur,   inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        // Residual
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // FFN Norm
+        cur = build_norm(ffn_inp, layer.ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            // Dense FFN layer
+            cur = build_ffn(cur,
+                layer.ffn_up, NULL, NULL,
+                layer.ffn_gate, NULL, NULL,
+                layer.ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE layer
+            // Kimi uses moe_renormalize=True and routed_scaling_factor (stored as expert_weights_scale) = 2.446
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                layer.ffn_gate_inp,
+                layer.ffn_up_exps,
+                layer.ffn_gate_exps,
+                layer.ffn_down_exps,
+                layer.ffn_exp_probs_b,
+                hparams.n_expert,
+                hparams.n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // Shared expert
+            {
+                ggml_tensor * ffn_shexp = build_ffn(cur,
+                        layer.ffn_up_shexp, NULL, NULL,
+                        layer.ffn_gate_shexp, NULL, NULL,
+                        layer.ffn_down_shexp, NULL, NULL,
+                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            }
+        }
+        // Residual
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    // Final Norm
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // Output
+    cur = ggml_mul_mat(ctx0, model.output, cur);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/laguna.cpp

@@ -0,0 +1,333 @@
+// Laguna (poolside): sigmoid-routed MoE with a score-correction bias, one shared
+// expert, a softplus attention output gate, QK-norm, and per-layer-type RoPE
+// (YaRN on full-attention layers, plain RoPE on sliding-window layers). XS.2 is
+// hybrid full/SWA with a per-head gate; M.1 is full-attention with a per-element
+// gate. Shares the MoE/gate structure with afmoe.
+
+#include "models.h"
+
+void llama_model_laguna::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+
+    // Laguna ships one shared expert and stores its size directly (routed and
+    // shared experts may differ), so read the size from expert_shared_feed_forward_length.
+    // The count is not in the config; default to 1 but read the key if present.
+    hparams.n_expert_shared = 1;
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,               hparams.n_expert_shared, false);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
+    if (hparams.n_ff_shexp == 0) {
+        // Weightless fixtures (test-llama-archs) omit this key; derive a nonzero
+        // size so the shared expert is still built. Real GGUFs always carry the
+        // exact value (routed and shared FF lengths may differ).
+        hparams.n_ff_shexp = hparams.n_ff_exp * hparams.n_expert_shared;
+    }
+
+    // Sliding-window attention is OPTIONAL. XS.2 is hybrid (full / SWA / SWA /
+    // SWA repeating, period 4 starting with full); M.1 has no sliding window
+    // (all layers full attention). When sliding_window is absent or zero we
+    // leave swa_type = NONE and skip the SWA-specific per-layer-type RoPE.
+    hparams.n_swa = 0;
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false);
+    if (hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+
+        uint32_t swa_period = 4;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period, /*dense_first=*/true);  // XS.2: FULL at il%4==0
+
+        // Per-layer-type RoPE: full layers use YaRN θ=500000 over 64 dims;
+        // SWA layers use default RoPE θ=10000 over 128 dims. Base load_hparams
+        // already reads ROPE_FREQ_BASE and ROPE_DIMENSION_COUNT into the
+        // non-SWA fields; we explicitly pull the SWA mirrors here.
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = 1.0f;  // SWA uses plain RoPE (no YaRN scaling); do NOT inherit full layers 1/factor
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+        ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT_SWA, hparams.n_rot_swa, false);
+    }
+
+    // Default the expert gating function to SIGMOID when the key is absent
+    // (matches the HF reference).
+    if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+    }
+
+    switch (hparams.n_layer()) {
+        case 40: type = LLM_TYPE_30B_A3B;   break;  // Laguna-XS.2
+        case 48: type = LLM_TYPE_118B_A8B;  break;  // Laguna-S.2
+        case 70: type = LLM_TYPE_230B_A10B; break;  // Laguna-M.1
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_laguna::load_arch_tensors(llama_model_loader & ml) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (output == NULL) {
+        // tied embeddings fallback
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    const int64_t n_ff_exp   = hparams.n_ff_exp;
+    const int64_t n_ff_shexp = hparams.n_ff_shexp;
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // Per-layer head count — Laguna varies n_head between full and SWA
+        // layers (48 vs 64 in XS.2). KV head count is uniform.
+        const int64_t n_head_il    = hparams.n_head(i);
+        const int64_t n_head_kv_il = hparams.n_head_kv(i);
+        const int64_t n_embd_q_il  = n_embd_head_k * n_head_il;
+        const int64_t n_embd_k_il  = n_embd_head_k * n_head_kv_il;
+        const int64_t n_embd_v_il  = n_embd_head_v * n_head_kv_il;
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_q_il, n_embd_k_il, n_embd_v_il, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_q_il, n_embd}, 0);
+
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        // Attention output gate. XS.2 is per-head (g_proj -> n_head, one scalar
+        // per head broadcast over head_dim at multiply time); M.1 is per-element
+        // (g_proj -> n_head*head_dim, like afmoe). Detect from the stored tensor
+        // shape so a single arch handles both; the graph mirrors this check.
+        // Gate width selects per-head vs per-element. Real GGUFs always carry the
+        // gate tensor, so read the width from it and require EXACTLY one of the two
+        // valid widths -- never guess between them. Weightless fixtures
+        // (test-llama-archs) have no gate tensor; fall back to the per-head layout so
+        // the per-head reshape path is still exercised.
+        const int64_t n_gate_per_head = n_head_il;
+        const int64_t n_gate_per_elem = n_embd_head_k * n_head_il;
+        const ggml_tensor * gate_meta = ml.get_tensor_meta(tn(LLM_TENSOR_ATTN_GATE, "weight", i).str().c_str());
+        int64_t n_gate_out;
+        if (gate_meta != nullptr) {
+            n_gate_out = gate_meta->ne[1];
+            if (n_gate_out != n_gate_per_head && n_gate_out != n_gate_per_elem) {
+                GGML_ABORT("Laguna: unexpected attention gate width %lld at layer %d "
+                           "(expected %lld per-head or %lld per-element)",
+                           (long long) n_gate_out, i, (long long) n_gate_per_head, (long long) n_gate_per_elem);
+            }
+        } else {
+            n_gate_out = n_gate_per_head;
+        }
+        layer.wqkv_gate = create_tensor(tn(LLM_TENSOR_ATTN_GATE, "weight", i), {n_embd, n_gate_out}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if ((uint32_t)i >= hparams.n_layer_dense_lead) {
+            // MoE layer
+            layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,    "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias",   i), {n_expert}, 0);
+
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff_exp, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd,   n_expert}, 0);
+
+            // Always-on shared expert.
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd,    n_ff_shexp}, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd,    n_ff_shexp}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd},    0);
+        } else {
+            // Dense layer (the leading n_layer_dense_lead layers)
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff,   n_embd}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_laguna::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_laguna::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+    // No MuP embedding scale (laguna omits this; afmoe scales by sqrt(hidden)).
+
+    ggml_tensor * inp_pos = build_inp_pos();
+    // XS.2 is hybrid SWA -> interleaved-SWA KV input; M.1 is all-full -> plain
+    // KV input. Pick the matching input (and build_attn overload) per swa_type.
+    const bool has_swa = hparams.swa_type != LLAMA_SWA_TYPE_NONE;
+    llm_graph_input_attn_kv      * inp_attn_kv   = has_swa ? nullptr : build_attn_inp_kv();
+    llm_graph_input_attn_kv_iswa * inp_attn_iswa = has_swa ? build_attn_inp_kv_iswa() : nullptr;
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    const float kq_scale = 1.0f / sqrtf(float(n_embd_head));
+
+    for (int il = 0; il < n_layer; ++il) {
+        const bool    is_swa_il   = hparams.is_swa(il);
+        const int64_t n_head_il   = hparams.n_head(il);
+        const int64_t n_head_kv_il = hparams.n_head_kv(il);
+
+        // Per-layer-type RoPE config. SWA layers run plain rope (no YaRN),
+        // achieved by zeroing the YaRN ext/beta params for those layers.
+        const int   n_rot_l       = is_swa_il ? hparams.n_rot_swa : n_rot;
+        const float freq_base_l   = is_swa_il ? hparams.rope_freq_base_train_swa : freq_base;
+        const float freq_scale_l  = is_swa_il ? hparams.rope_freq_scale_train_swa : freq_scale;
+        const float ext_factor_l  = is_swa_il ? 0.0f : ext_factor;
+        // YaRN magnitude scaling (mscale) is already handled by the framework:
+        // llama_context pre-divides cparams.yarn_attn_factor by (1 + 0.1*ln(factor))
+        // to cancel ggml rope_yarn's internal mscale *= 1 + 0.1*ln(1/freq_scale).
+        // Pass attn_factor straight through (like every other arch); SWA layers run
+        // plain RoPE (ext_factor 0, no mscale) so force 1.0 there.
+        const float attn_factor_l = is_swa_il ? 1.0f : attn_factor;
+        const float beta_fast_l   = is_swa_il ? 0.0f : beta_fast;
+        const float beta_slow_l   = is_swa_il ? 0.0f : beta_slow;
+        const int   n_ctx_orig_l  = is_swa_il ? hparams.n_ctx_train : n_ctx_orig;
+
+        ggml_tensor * inpSA = inpL;
+
+        // Pre-norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // Self-attention
+        {
+            ggml_tensor * attn_inp = cur;  // saved for the gate projection
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head_il, n_head_kv_il, il);
+
+            // g_proj on the *pre-attention* hidden state (matches HF
+            // reference: gate is computed from the same `hidden_states`
+            // input as q/k/v, not from the attn output).
+            ggml_tensor * gate = build_lora_mm(model.layers[il].wqkv_gate, attn_inp);
+            cb(gate, "attn_gate_proj", il);
+
+            // QK RMSNorm at head_dim level (Qwen3 style)
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr,
+                    n_rot_l, rope_type, n_ctx_orig_l, freq_base_l, freq_scale_l,
+                    ext_factor_l, attn_factor_l, beta_fast_l, beta_slow_l);
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr,
+                    n_rot_l, rope_type, n_ctx_orig_l, freq_base_l, freq_scale_l,
+                    ext_factor_l, attn_factor_l, beta_fast_l, beta_slow_l);
+            cb(Qcur, "Qcur_rope", il);
+            cb(Kcur, "Kcur_rope", il);
+
+            cur = has_swa
+                ? build_attn(inp_attn_iswa,
+                        NULL, NULL, NULL,    // o_proj deferred until after gating
+                        Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il)
+                : build_attn(inp_attn_kv,
+                        NULL, NULL, NULL,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+
+            // Softplus output gate (the unary kernel computes softplus in fp32
+            // and casts back). Two shapes, distinguished by the g_proj output
+            // dim (matching the load-time detection):
+            //   XS.2 per-head     : gate [n_head_il, n_tokens] -> reshape to
+            //                       [1, n_head_il, n_tokens] and broadcast over
+            //                       head_dim against cur [head_dim, n_head, T].
+            //   M.1  per-element  : gate [n_head_il*head_dim, n_tokens] spans the
+            //                       full attention output -> direct ggml_mul.
+            gate = ggml_softplus(ctx0, gate);
+            cb(gate, "attn_gate_softplus", il);
+
+            const int64_t n_tokens = cur->ne[1];
+            if (model.layers[il].wqkv_gate->ne[1] == n_head_il) {
+                cur  = ggml_reshape_3d(ctx0, cur,  n_embd_head, n_head_il, n_tokens);
+                gate = ggml_reshape_3d(ctx0, gate, 1,           n_head_il, n_tokens);
+                cur  = ggml_mul(ctx0, cur, gate);
+                cur  = ggml_reshape_2d(ctx0, cur, n_embd_head * n_head_il, n_tokens);
+            } else {
+                cur = ggml_mul(ctx0, cur, gate);
+            }
+            cb(cur, "attn_gated", il);
+
+            cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
+            cb(cur, "attn_o_proj", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // Pre-norm only (no post-attn norm)
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t)il >= hparams.n_layer_dense_lead) {
+            // MoE: sigmoid routing + score-correction bias + sum-norm +
+            // routed_scaling_factor (all handled by build_moe_ffn).
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU,
+                    hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // Always-on shared expert, summed in parallel.
+            ggml_tensor * ffn_shexp = build_ffn(cur,
+                    model.layers[il].ffn_up_shexp,   NULL, NULL,
+                    model.layers[il].ffn_gate_shexp, NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(ffn_shexp, "ffn_shexp", il);
+
+            cur = ggml_add(ctx0, moe_out, ffn_shexp);
+            cb(cur, "ffn_out", il);
+        } else {
+            // Dense FFN for the leading n_layer_dense_lead layers (XS.2: 1, M.1: 3)
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        // No post-ffn norm
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        inpL = cur;
+    }
+
+    cur = inpL;
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/lfm2.cpp

@@ -0,0 +1,290 @@
+#include "models.h"
+#include "../llama-memory-hybrid-iswa.h"
+#include "../llama-memory-hybrid.h"
+
+void llama_model_lfm2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_SHORTCONV_L_CACHE,           hparams.n_shortconv_l_cache);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    for (uint32_t il = 0; il < hparams.n_layer(); ++il) {
+        hparams.is_recr_impl[il] = hparams.n_head_kv(il) == 0;
+    }
+
+    hparams.n_layer_dense_lead = hparams.n_layer();
+
+    switch (hparams.n_ff()) {
+        case  2560: type = LLM_TYPE_230M; break;
+        case  4608: type = LLM_TYPE_350M; break;
+        case  6912: type = LLM_TYPE_700M; break;
+        case  8192: type = LLM_TYPE_1_2B; break;
+        case 10752: type = LLM_TYPE_2_6B; break;
+        default:    type = LLM_TYPE_UNKNOWN;
+    }
+
+    if (const auto is_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); is_swa && hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+        for (uint32_t il = 0; il < hparams.n_layer(); ++il) {
+            hparams.is_swa_impl[il] = !hparams.is_recr_impl[il];
+        }
+    }
+}
+
+void llama_model_lfm2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM_LFM2, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,           "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        const bool is_moe_layer = i >= static_cast<int>(hparams.n_layer_dense_lead);
+
+        // ffn/moe is same for transformer and conv layers
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        if (is_moe_layer) {
+            GGML_ASSERT(n_expert && n_expert_used);
+            layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i),  {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps   = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, hparams.n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {hparams.n_ff_exp,   n_embd, n_expert}, 0);
+            layer.ffn_up_exps     = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i),   {n_embd, hparams.n_ff_exp, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0);
+        } else {  // dense
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        }
+
+        // for operator_norm
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (!hparams.is_recr(i)) {
+            layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+            layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+            GGML_ASSERT(n_embd_v_gqa == n_embd_k_gqa);
+
+            create_tensor_qkv(layer, i, n_embd, n_embd, hparams.n_embd_k_gqa(i), hparams.n_embd_v_gqa(i), 0);
+
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        } else {
+            layer.shortconv.conv     = create_tensor(tn(LLM_TENSOR_SHORTCONV_CONV,    "weight", i), {hparams.n_shortconv_l_cache, n_embd}, 0);
+            layer.shortconv.in_proj  = create_tensor(tn(LLM_TENSOR_SHORTCONV_INPROJ,  "weight", i), {n_embd, 3 * n_embd}, 0);
+            layer.shortconv.out_proj = create_tensor(tn(LLM_TENSOR_SHORTCONV_OUTPROJ, "weight", i), {n_embd, n_embd}, 0);
+        }
+    }
+
+    // for LFM2-ColBert-350M
+    dense_2_out_layers   = create_tensor(tn(LLM_TENSOR_DENSE_2_OUT, "weight"), {n_embd, hparams.n_embd_out()}, TENSOR_NOT_REQUIRED);
+    dense_2_out_layers_b = create_tensor(tn(LLM_TENSOR_DENSE_2_OUT, "bias"),   {hparams.n_embd_out()        }, TENSOR_NOT_REQUIRED);
+}
+
+std::unique_ptr<llm_graph_context> llama_model_lfm2::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_STANDARD) {
+        return std::make_unique<graph<true>>(*this, params);
+    } else {
+        return std::make_unique<graph<false>>(*this, params);
+    }
+}
+
+template <bool iswa>
+llama_model_lfm2::graph<iswa>::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+    using inp_hybrid_type = std::conditional_t<iswa, llm_graph_input_mem_hybrid_iswa,  llm_graph_input_mem_hybrid>;
+    using inp_attn_type   = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa,     llm_graph_input_attn_kv>;
+    using mem_hybrid_ctx  = std::conditional_t<iswa, llama_memory_hybrid_iswa_context, llama_memory_hybrid_context>;
+
+    // lambda helpers for readability
+    auto build_dense_feed_forward = [&model, this](ggml_tensor * cur, int il) -> ggml_tensor * {
+        GGML_ASSERT(!model.layers[il].ffn_up_b);
+        GGML_ASSERT(!model.layers[il].ffn_gate_b);
+        GGML_ASSERT(!model.layers[il].ffn_down_b);
+        return build_ffn(cur,
+            model.layers[il].ffn_up, NULL, NULL,
+            model.layers[il].ffn_gate, NULL, NULL,
+            model.layers[il].ffn_down, NULL, NULL,
+            NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+    };
+    auto build_moe_feed_forward = [&model, this](ggml_tensor * cur, int il) -> ggml_tensor * {
+        return build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                static_cast<llama_expert_gating_func_type>(hparams.expert_gating_func),
+                il);
+    };
+    auto build_attn_block = [&model, this](ggml_tensor *   cur,
+                                           ggml_tensor *   inp_pos,
+                                           inp_attn_type * inp_attn,
+                                           int             il) -> ggml_tensor * {
+        GGML_ASSERT(hparams.n_embd_v_gqa(il) == hparams.n_embd_k_gqa(il));
+        const auto n_embd_head = hparams.n_embd_head_v();
+        const auto n_head_kv   = hparams.n_head_kv(il);
+
+        auto [q, k, v] = build_qkv(model.layers[il], cur,
+                n_embd_head, n_head, n_head_kv, il);
+
+        // qk norm
+        q = build_norm(q, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+        cb(q, "model.layers.{}.self_attn.q_layernorm", il);
+        k = build_norm(k, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+        cb(k, "model.layers.{}.self_attn.k_layernorm", il);
+
+        // RoPE
+        q = ggml_rope_ext(ctx0, q, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, ext_factor,
+                          attn_factor, beta_fast, beta_slow);
+        k = ggml_rope_ext(ctx0, k, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, ext_factor,
+                          attn_factor, beta_fast, beta_slow);
+
+        cur = build_attn(inp_attn,
+                model.layers[il].wo, NULL, model.layers[il].wo_s,
+                q, k, v, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+
+        cb(cur, "model.layers.{}.self_attn.out_proj", il);
+
+        return cur;
+    };
+    auto build_shortconv_block = [&model, this](ggml_tensor *        cur,
+                                                llm_graph_input_rs * inp_recr,
+                                                int                  il) -> ggml_tensor * {
+        const auto * mctx_cur = static_cast<const mem_hybrid_ctx *>(mctx)->get_recr();
+        const uint32_t kv_head      = mctx_cur->get_head();
+        const int64_t  n_seq_tokens = ubatch.n_seq_tokens;
+        const int64_t  n_seqs       = ubatch.n_seqs;
+        GGML_ASSERT(n_seqs != 0);
+        GGML_ASSERT(ubatch.equal_seqs());
+        GGML_ASSERT(ubatch.n_tokens == n_seq_tokens * n_seqs);
+
+        GGML_ASSERT(hparams.n_shortconv_l_cache > 1);
+        const uint32_t d_conv = hparams.n_shortconv_l_cache - 1;
+
+        // {n_embd, n_tokens} => {n_embd, n_seq_tokens, n_seqs}
+        cur = ggml_reshape_3d(ctx0, cur, cur->ne[0], n_seq_tokens, n_seqs);
+
+        auto * bcx = build_lora_mm(model.layers[il].shortconv.in_proj, cur);
+        cb(bcx, "model.layers.{}.conv.in_proj", il);
+
+        constexpr auto n_chunks = 3;
+        GGML_ASSERT(bcx->ne[0] % n_chunks == 0);
+        const auto chunk_size = bcx->ne[0] / n_chunks;
+        auto *     b          = ggml_view_3d(ctx0, bcx, chunk_size, bcx->ne[1], bcx->ne[2], bcx->nb[1], bcx->nb[2],
+                                             0 * chunk_size * ggml_element_size(bcx));
+        auto *     c          = ggml_view_3d(ctx0, bcx, chunk_size, bcx->ne[1], bcx->ne[2], bcx->nb[1], bcx->nb[2],
+                                             1 * chunk_size * ggml_element_size(bcx));
+        auto *     x          = ggml_view_3d(ctx0, bcx, chunk_size, bcx->ne[1], bcx->ne[2], bcx->nb[1], bcx->nb[2],
+                                             2 * chunk_size * ggml_element_size(bcx));
+
+        auto * bx = ggml_transpose(ctx0, ggml_mul(ctx0, b, x));
+
+        // read conv state
+        auto * conv_state = mctx_cur->get_r_l(il);
+        auto * conv_rs    = build_rs(inp_recr, conv_state, hparams.n_embd_r(), n_seqs);
+        auto * conv       = ggml_reshape_3d(ctx0, conv_rs, d_conv, hparams.n_embd, n_seqs);
+
+        // causal prepends the state, non-causal pads symmetrically for a centered window
+        if (hparams.causal_attn) {
+            bx = ggml_concat(ctx0, conv, bx, 0);
+        } else {
+            const int64_t pad = (hparams.n_shortconv_l_cache - 1) / 2;
+            auto * left = ggml_cont(ctx0,
+                ggml_view_3d(ctx0, conv, pad, hparams.n_embd, n_seqs, conv->nb[1], conv->nb[2], (d_conv - pad) * conv->nb[0]));
+            bx = ggml_pad_ext(ctx0, ggml_concat(ctx0, left, bx, 0), 0, pad, 0, 0, 0, 0, 0, 0);
+        }
+        GGML_ASSERT(bx->ne[0] > conv->ne[0]);
+
+        // last d_conv columns is a new conv state
+        auto * new_conv = ggml_view_3d(ctx0, bx, conv->ne[0], bx->ne[1], bx->ne[2], bx->nb[1], bx->nb[2],
+                                       (bx->ne[0] - conv->ne[0]) * ggml_element_size(bx));
+        GGML_ASSERT(ggml_are_same_shape(conv, new_conv));
+
+        // write new conv conv state
+        ggml_build_forward_expand(gf, ggml_cpy(ctx0, new_conv,
+                                               ggml_view_1d(ctx0, conv_state, ggml_nelements(new_conv),
+                                                            kv_head * d_conv * n_embd * ggml_element_size(new_conv))));
+
+        auto * conv_kernel = model.layers[il].shortconv.conv;
+        auto * conv_out    = ggml_ssm_conv(ctx0, bx, conv_kernel);
+        cb(conv_out, "model.layers.{}.conv.conv", il);
+
+        auto * y = ggml_mul(ctx0, c, conv_out);
+        y        = build_lora_mm(model.layers[il].shortconv.out_proj, y);
+        cb(y, "model.layers.{}.conv.out_proj", il);
+        // {n_embd, n_seq_tokens, n_seqs} => {n_embd, n_tokens}
+        y = ggml_reshape_2d(ctx0, y, y->ne[0], n_seq_tokens * n_seqs);
+
+        return y;
+    };
+
+    // actual graph construction starts here
+    ggml_tensor * cur = build_inp_embd(model.tok_embd);
+    cb(cur, "model.embed_tokens", -1);
+
+    ggml_build_forward_expand(gf, cur);
+
+    inp_hybrid_type * inp_hybrid = nullptr;
+    if constexpr (iswa) {
+        inp_hybrid = build_inp_mem_hybrid_iswa();
+    } else {
+        inp_hybrid = build_inp_mem_hybrid();
+    }
+
+    ggml_tensor * inp_pos     = build_inp_pos();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const bool is_moe_layer = il >= static_cast<int>(hparams.n_layer_dense_lead);
+
+        auto * prev_cur = cur;
+        cur             = build_norm(cur, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "model.layers.{}.operator_norm", il);
+
+        cur = hparams.is_recr(il) ? build_shortconv_block(cur, inp_hybrid->get_recr(), il) :
+                                    build_attn_block(cur, inp_pos, inp_hybrid->get_attn(), il);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur      = ggml_get_rows(ctx0, cur, inp_out_ids);
+            prev_cur = ggml_get_rows(ctx0, prev_cur, inp_out_ids);
+        }
+
+        cur = ggml_add(ctx0, prev_cur, cur);
+
+        auto * ffn_norm_out = build_norm(cur, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(ffn_norm_out, "model.layers.{}.ffn_norm", il);
+
+        ggml_tensor * ffn_out =
+            is_moe_layer ? build_moe_feed_forward(ffn_norm_out, il) : build_dense_feed_forward(ffn_norm_out, il);
+        cb(ffn_norm_out, "model.layers.{}.ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_out);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+    }
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    if (!cparams.embeddings) {
+        cur = build_lora_mm(model.output, cur, model.output_s);
+        cb(cur, "result_output", -1);
+
+        res->t_logits = cur;
+    }
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// Explicit template instantiations
+template struct llama_model_lfm2::graph<true>;
+template struct llama_model_lfm2::graph<false>;

src/models/lfm2moe.cpp

@@ -0,0 +1,85 @@
+#include "models.h"
+#include "../llama-memory-hybrid-iswa.h"
+#include "../llama-memory-hybrid.h"
+
+void llama_model_lfm2moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_SHORTCONV_L_CACHE,           hparams.n_shortconv_l_cache);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func);
+
+    for (uint32_t il = 0; il < hparams.n_layer(); ++il) {
+        hparams.is_recr_impl[il] = hparams.n_head_kv(il) == 0;
+    }
+
+    switch (hparams.n_layer()) {
+        case 24: type = LLM_TYPE_8B_A1B;  break;
+        case 40: type = LLM_TYPE_24B_A2B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_lfm2moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM_LFM2, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,           "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        const bool is_moe_layer = i >= static_cast<int>(hparams.n_layer_dense_lead);
+
+        // ffn/moe is same for transformer and conv layers
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        if (is_moe_layer) {
+            GGML_ASSERT(n_expert && n_expert_used);
+            layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i),  {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps   = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, hparams.n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {hparams.n_ff_exp,   n_embd, n_expert}, 0);
+            layer.ffn_up_exps     = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i),   {n_embd, hparams.n_ff_exp, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0);
+        } else {  // dense
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        }
+
+        // for operator_norm
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (!hparams.is_recr(i)) {
+            layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+            layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+            GGML_ASSERT(n_embd_v_gqa == n_embd_k_gqa);
+
+            create_tensor_qkv(layer, i, n_embd, n_embd, hparams.n_embd_k_gqa(i), hparams.n_embd_v_gqa(i), 0);
+
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        } else {
+            layer.shortconv.conv     = create_tensor(tn(LLM_TENSOR_SHORTCONV_CONV,    "weight", i), {hparams.n_shortconv_l_cache, n_embd}, 0);
+            layer.shortconv.in_proj  = create_tensor(tn(LLM_TENSOR_SHORTCONV_INPROJ,  "weight", i), {n_embd, 3 * n_embd}, 0);
+            layer.shortconv.out_proj = create_tensor(tn(LLM_TENSOR_SHORTCONV_OUTPROJ, "weight", i), {n_embd, n_embd}, 0);
+        }
+    }
+
+    // for LFM2-ColBert-350M
+    dense_2_out_layers   = create_tensor(tn(LLM_TENSOR_DENSE_2_OUT, "weight"), {n_embd, hparams.n_embd_out()}, TENSOR_NOT_REQUIRED);
+    dense_2_out_layers_b = create_tensor(tn(LLM_TENSOR_DENSE_2_OUT, "bias"),   {hparams.n_embd_out()        }, TENSOR_NOT_REQUIRED);
+}
+
+std::unique_ptr<llm_graph_context> llama_model_lfm2moe::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_STANDARD) {
+        return std::make_unique<graph<true>>(*this, params);
+    } else {
+        return std::make_unique<graph<false>>(*this, params);
+    }
+}
+

src/models/llada-moe.cpp

@@ -0,0 +1,163 @@
+#include "models.h"
+
+void llama_model_llada_moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false);
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    // diffusion language model uses non-causal attention
+    hparams.causal_attn = false;
+
+    switch (hparams.n_layer()) {
+        case 16: type = LLM_TYPE_A1_7B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_llada_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    GGML_ASSERT(n_expert > 0 && "n_expert must be > 0 for llada-moe");
+    GGML_ASSERT(n_expert_used > 0 && "n_expert_used must be > 0 for llada-moe");
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+        const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_llada_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_llada_moe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // MoE branch
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, false,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(cur, "ffn_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/llada.cpp

@@ -0,0 +1,158 @@
+#include "models.h"
+
+void llama_model_llada::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    // LLaDA-8B has 32 layers, similar to LLaMA but for diffusion
+    switch (hparams.n_layer()) {
+        case 32:
+            type = LLM_TYPE_8B;
+            break;
+        default:
+            type = LLM_TYPE_UNKNOWN;
+    }
+
+    // Set non-causal attention for diffusion models
+    hparams.causal_attn = false;
+}
+
+void llama_model_llada::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output =
+            create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+
+        // Use separate Q, K, V projections without bias, matching LLaDALlamaBlock
+        layer.wq =
+            create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), { n_embd, n_embd_head_k * n_head }, 0);
+        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), { n_embd, n_embd_k_gqa }, 0);
+        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), { n_embd, n_embd_v_gqa }, 0);
+        // No bias for QKV projections as per config: include_bias=false, include_qkv_bias=false
+        layer.wo =
+            create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), { n_embd }, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), { n_embd }, 0);
+
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), { n_rot / 2 },
+                                         TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, n_ff }, 0);
+
+        // optional MLP bias
+        layer.ffn_gate_b =
+            create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), { n_ff }, TENSOR_NOT_REQUIRED);
+        layer.ffn_down_b =
+            create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), { n_embd }, TENSOR_NOT_REQUIRED);
+        layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), { n_ff }, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_llada::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_llada::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    // LLaDA is similar to LLaMA but uses non-causal attention for diffusion
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // Non-causal attention for diffusion
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute separate Q, K, V projections without bias, matching LLaDALlamaBlock
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up, NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/llama-embed.cpp

@@ -0,0 +1,6 @@
+#include "models.h"
+
+std::unique_ptr<llm_graph_context> llama_model_llama_embed::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph<true>>(*this, params);
+}
+

src/models/llama.cpp

@@ -0,0 +1,250 @@
+#include "models.h"
+
+void llama_model_llama::load_arch_hparams(llama_model_loader & ml) {
+    uint32_t n_vocab = 0;
+    ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false);
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    if (hparams.n_expert == 8) {
+        switch (hparams.n_layer()) {
+            case 32: type = LLM_TYPE_8x7B; break;
+            case 56: type = LLM_TYPE_8x22B; break;
+            default: type = LLM_TYPE_UNKNOWN;
+        }
+    } else {
+        switch (hparams.n_layer()) {
+            case 16: type = LLM_TYPE_1B; break; // Llama 3.2 1B
+            case 22: type = LLM_TYPE_1B; break;
+            case 26: type = LLM_TYPE_3B; break;
+            case 28: type = LLM_TYPE_3B; break; // Llama 3.2 3B
+            case 30: type = LLM_TYPE_256M; break; // smoldocling 256M
+            // granite uses a vocab with len 49152
+            case 32: type = n_vocab == 49152 ? LLM_TYPE_3B : (n_vocab < 40000 ? LLM_TYPE_7B : LLM_TYPE_8B); break;
+            case 36: type = LLM_TYPE_8B; break; // granite
+            case 40: type = LLM_TYPE_13B; break;
+            case 48: type = LLM_TYPE_34B; break;
+            case 60: type = LLM_TYPE_30B; break;
+            case 80: type = hparams.n_head() == hparams.n_head_kv() ? LLM_TYPE_65B : LLM_TYPE_70B; break;
+            default: type = LLM_TYPE_UNKNOWN;
+        }
+    }
+}
+
+void llama_model_llama::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+        else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        if (n_expert == 0) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+            // optional MLP bias
+            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        } else {
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+            // For Granite MoE Shared
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd}, 0);
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_llama::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph<false>>(*this, params);
+}
+
+template <bool embed>
+llama_model_llama::graph<embed>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    using inp_attn_type = std::conditional_t<embed, llm_graph_input_attn_no_cache, llm_graph_input_attn_kv>;
+
+    inp_attn_type * inp_attn = nullptr;
+    if constexpr (embed) {
+        inp_attn = build_attn_inp_no_cache();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+
+    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        res->t_layer_inp[il] = inpL;
+
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            if (hparams.use_kq_norm) {
+                // Llama4TextL2Norm
+                Qcur = ggml_rms_norm(ctx0, Qcur, hparams.f_norm_rms_eps);
+                Kcur = ggml_rms_norm(ctx0, Kcur, hparams.f_norm_rms_eps);
+                cb(Qcur, "Qcur_normed", il);
+                cb(Kcur, "Kcur_normed", il);
+            }
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network (non-MoE)
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   model.layers[il].ffn_up_s,
+                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, model.layers[il].ffn_gate_s,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, model.layers[il].ffn_down_s,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, true,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il,
+                    nullptr, nullptr,
+                    model.layers[il].ffn_up_exps_s,
+                    model.layers[il].ffn_gate_exps_s,
+                    model.layers[il].ffn_down_exps_s);
+            cb(cur, "ffn_moe_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    if constexpr (!embed) {
+        // lm_head
+        cur = build_lora_mm(model.output, cur, model.output_s);
+
+        cb(cur, "result_output", -1);
+        res->t_logits = cur;
+    }
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+template struct llama_model_llama::graph<false>;
+template struct llama_model_llama::graph<true>;

src/models/llama4.cpp

@@ -0,0 +1,274 @@
+#include "models.h"
+
+void llama_model_llama4::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_INTERLEAVE_MOE_LAYER_STEP,   hparams.n_moe_layer_step);
+
+    const bool found_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false);
+    if (found_swa && hparams.n_swa == 0) {
+        hparams.swa_type             = LLAMA_SWA_TYPE_NONE;
+        hparams.n_no_rope_layer_step = hparams.n_layer(); // always use rope
+    } else {
+        hparams.swa_type                = LLAMA_SWA_TYPE_CHUNKED;
+        hparams.n_swa                   = 8192;
+        hparams.n_attn_temp_floor_scale = 8192;
+        hparams.f_attn_temp_scale       = 0.1f;
+        hparams.f_attn_temp_offset      = 1.0f;
+
+        uint32_t swa_period = 4; // pattern: 3 chunked - 1 full
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period);
+
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    }
+
+    switch (hparams.n_expert) {
+        case 0: {
+            // MobileLLM (no MoE)
+            switch (hparams.n_embd) {
+                case 2048: type = LLM_TYPE_140M; break;
+                case 4096: type = LLM_TYPE_360M; break;
+                case 6144: type = LLM_TYPE_950M; break;
+                default:   type = LLM_TYPE_UNKNOWN;
+            }
+        } break;
+        case 16:  type = LLM_TYPE_17B_16E; break;
+        case 128: type = LLM_TYPE_17B_128E; break;
+        default:  type = LLM_TYPE_UNKNOWN;
+    }
+
+    hparams.use_kq_norm = type != LLM_TYPE_17B_128E;
+}
+
+void llama_model_llama4::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_expert == 0) {
+        throw std::runtime_error(arch_name() + " model cannot have zero experts");
+    }
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        const bool is_moe_layer = hparams.n_moe_layer_step > 0 && (i + 1) % hparams.n_moe_layer_step == 0;
+
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+
+        if (is_moe_layer) {
+            const int64_t n_ff_exp = hparams.n_ff_exp;
+
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff_exp, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff_exp, n_expert}, 0);
+
+            // Shared expert
+            const int64_t n_ff_shexp = n_ff_exp;
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {    n_embd, n_ff_shexp}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd    }, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {    n_embd, n_ff_shexp}, 0);
+        } else {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_llama4::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_NONE) {
+        return std::make_unique<graph<false>>(*this, params);
+    } else {
+        return std::make_unique<graph<true>>(*this, params);
+    }
+}
+
+template <bool iswa>
+llama_model_llama4::graph<iswa>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // temperature tuning
+    ggml_tensor * inp_attn_scale = nullptr;
+    inp_attn_scale = build_inp_attn_scale();
+
+    using inp_attn_type = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa, llm_graph_input_attn_kv>;
+    inp_attn_type * inp_attn = nullptr;
+
+    if constexpr (iswa) {
+        inp_attn = build_attn_inp_kv_iswa();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+
+    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const float freq_base_l  = model.get_rope_freq_base (cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        ggml_tensor * inpSA = inpL;
+
+        // This overlaps with SWA layers in current models, so get_rope_freq_base/scale may be superfluous
+        const bool use_rope = hparams.n_no_rope_layer_step > 0 &&
+                              (il + 1) % hparams.n_no_rope_layer_step != 0;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            if (use_rope) {
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, rope_factors,
+                        n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, rope_factors,
+                        n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+            } else if (inp_attn_scale) {
+                Qcur = ggml_mul(ctx0, Qcur, inp_attn_scale);
+            }
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            if (use_rope && hparams.use_kq_norm) {
+                // Llama4TextL2Norm
+                Qcur = ggml_rms_norm(ctx0, Qcur, hparams.f_norm_rms_eps);
+                Kcur = ggml_rms_norm(ctx0, Kcur, hparams.f_norm_rms_eps);
+                cb(Qcur, "Qcur_normed", il);
+                cb(Kcur, "Kcur_normed", il);
+            }
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network (non-MoE)
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            ggml_tensor * ffn_inp_normed = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            ggml_tensor * moe_out = build_moe_ffn(ffn_inp_normed,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, false,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID,
+                    il);
+
+            // Shared experts
+            ggml_tensor * shexp_out = build_ffn(ffn_inp_normed,
+                model.layers[il].ffn_up_shexp,   NULL, NULL,
+                model.layers[il].ffn_gate_shexp, NULL, NULL,
+                model.layers[il].ffn_down_shexp, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(shexp_out, "ffn_moe_shexp", il);
+
+            cur = ggml_add(ctx0, moe_out, shexp_out);
+            cb(cur, "ffn_moe_out_merged", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// Explicit template instantiations
+template struct llama_model_llama4::graph<false>;
+template struct llama_model_llama4::graph<true>;

src/models/maincoder.cpp

@@ -0,0 +1,151 @@
+#include "models.h"
+
+void llama_model_maincoder::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_1B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_maincoder::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_maincoder::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_maincoder::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/mamba-base.cpp

@@ -0,0 +1,288 @@
+#include "models.h"
+
+#include "llama-memory-recurrent.h"
+
+llm_build_mamba_base::llm_build_mamba_base(const llm_graph_params & params) : llm_graph_context(params) {}
+
+ggml_tensor * llm_build_mamba_base::build_mamba_layer(llm_graph_input_rs * inp,
+                                                         ggml_tensor *        cur,
+                                                         const llama_model &  model,
+                                                         const llama_ubatch & ubatch,
+                                                         int                  il) {
+    const auto * mctx_cur = inp->mctx;
+
+    const auto kv_head = mctx_cur->get_head();
+
+    const auto & layer = model.layers[il];
+
+    const int64_t d_conv         = hparams.ssm_d_conv;
+    const int64_t d_inner        = hparams.ssm_d_inner;
+    const int64_t d_state        = hparams.ssm_d_state;
+    const int64_t dt_rank        = hparams.ssm_dt_rank;
+    const int64_t n_head         = d_inner;
+    const int64_t head_dim       = 1;
+    const int64_t n_seqs         = ubatch.n_seqs;
+    // Some variants of Mamba arch (e.g. FalconMamba do apply layer norm on B and Dt layers)
+    const bool    ssm_dt_b_c_rms = hparams.ssm_dt_b_c_rms;
+
+    const int64_t n_seq_tokens = ubatch.n_seq_tokens;
+
+    GGML_ASSERT(n_seqs != 0);
+    GGML_ASSERT(ubatch.equal_seqs());
+    GGML_ASSERT(ubatch.n_tokens == n_seq_tokens * n_seqs);
+    GGML_ASSERT(d_inner % n_head == 0);
+
+    ggml_tensor * conv_states_all = mctx_cur->get_r_l(il);
+    ggml_tensor * ssm_states_all  = mctx_cur->get_s_l(il);
+
+    ggml_tensor * conv = build_rs(inp, conv_states_all, hparams.n_embd_r(), n_seqs);
+    conv               = ggml_reshape_3d(ctx0, conv, d_conv - 1, d_inner, n_seqs);
+
+    // {n_embd, n_tokens} => {n_embd, n_seq_tokens, n_seqs}
+    cur = ggml_reshape_3d(ctx0, cur, cur->ne[0], n_seq_tokens, n_seqs);
+
+    // {n_embd, 2*d_inner} @ {n_embd, n_seq_tokens, n_seqs} => {2*d_inner, n_seq_tokens, n_seqs}
+    ggml_tensor * xz = build_lora_mm(layer.ssm_in, cur, layer.ssm_in_s);
+    // split the above in two
+    // => {d_inner, n_seq_tokens, n_seqs}
+    ggml_tensor * x  = ggml_view_3d(ctx0, xz, d_inner, xz->ne[1], xz->ne[2], xz->nb[1], xz->nb[2], 0);
+    ggml_tensor * z =
+        ggml_view_3d(ctx0, xz, d_inner, xz->ne[1], xz->ne[2], xz->nb[1], xz->nb[2], d_inner * ggml_element_size(xz));
+
+    // conv
+    {
+        // => {d_conv - 1 + n_seq_tokens, d_inner, n_seqs}
+        ggml_tensor * conv_x = ggml_concat(ctx0, conv, ggml_transpose(ctx0, x), 0);
+
+        // copy last (d_conv - 1) columns back into the state cache
+        ggml_tensor * last_conv = ggml_view_3d(ctx0, conv_x, d_conv - 1, d_inner, n_seqs, conv_x->nb[1], conv_x->nb[2],
+                                               n_seq_tokens * (conv_x->nb[0]));
+
+        ggml_build_forward_expand(
+            gf, ggml_cpy(ctx0, last_conv,
+                         ggml_view_1d(ctx0, conv_states_all, (d_conv - 1) * (d_inner) * (n_seqs),
+                                      kv_head * (d_conv - 1) * (d_inner) *ggml_element_size(conv_states_all))));
+
+        // 1D convolution
+        // The equivalent is to make a self-overlapping view of conv_x
+        // over d_conv columns at each stride in the 3rd dimension,
+        // then element-wise multiply that with the conv1d weight,
+        // then sum the elements of each row,
+        // (the last two steps are a dot product over rows (also doable with mul_mat))
+        // then permute away the ne[0] dimension,
+        // and then you're left with the resulting x tensor.
+        // For simultaneous sequences, all sequences need to have the same length.
+        x = ggml_ssm_conv(ctx0, conv_x, layer.ssm_conv1d);
+
+        // bias
+        x = ggml_add(ctx0, x, layer.ssm_conv1d_b);
+
+        x = ggml_silu(ctx0, x);
+    }
+
+    // ssm
+    {
+        // {d_inner, dt_rank + 2*d_state} @ {d_inner, n_seq_tokens, n_seqs} => {dt_rank + 2*d_state, n_seq_tokens, n_seqs}
+        ggml_tensor * x_db = build_lora_mm(layer.ssm_x, x);
+        // split
+        ggml_tensor * dt   = ggml_view_3d(ctx0, x_db, dt_rank, n_seq_tokens, n_seqs, x_db->nb[1], x_db->nb[2], 0);
+        ggml_tensor * B =
+            ggml_view_4d(ctx0, x_db, d_state, /* n_group */ 1, n_seq_tokens, n_seqs, d_state * x_db->nb[0], x_db->nb[1],
+                         x_db->nb[2], ggml_element_size(x_db) * dt_rank);
+        ggml_tensor * C =
+            ggml_view_4d(ctx0, x_db, d_state, /* n_group */ 1, n_seq_tokens, n_seqs, d_state * x_db->nb[0], x_db->nb[1],
+                         x_db->nb[2], ggml_element_size(x_db) * (dt_rank + d_state));
+
+        // Some Mamba variants (e.g. FalconMamba, Jamba) apply RMS norm in B, C & Dt layers
+        if (ssm_dt_b_c_rms || (layer.ssm_dt_norm && layer.ssm_b_norm && layer.ssm_c_norm)) {
+            dt = build_norm(dt, layer.ssm_dt_norm, NULL, LLM_NORM_RMS, il);
+            B  = build_norm(B, layer.ssm_b_norm, NULL, LLM_NORM_RMS, il);
+            C  = build_norm(C, layer.ssm_c_norm, NULL, LLM_NORM_RMS, il);
+        }
+
+        // {dt_rank, d_inner} @ {dt_rank, n_seq_tokens, n_seqs} => {d_inner, n_seq_tokens, n_seqs}
+        dt = build_lora_mm(layer.ssm_dt, dt);
+        dt = ggml_add(ctx0, dt, layer.ssm_dt_b);
+
+        cur = x;
+        x   = ggml_reshape_4d(ctx0, x, head_dim, n_head, n_seq_tokens, n_seqs);
+
+        ggml_tensor * A = layer.ssm_a;
+
+        // use the states and the indices provided by build_recurrent_state
+        // (this is necessary in order to properly use the states before they are overwritten,
+        //  while avoiding to make unnecessary copies of the states)
+        auto get_ssm_rows = [&](ggml_context * ctx, ggml_tensor * states, ggml_tensor * ids) {
+            ggml_tensor * ssm = ggml_reshape_4d(ctx, states, d_state, head_dim, n_head, mctx_cur->get_size());
+
+            // Custom operator to optimize the parallel associative scan
+            // as described in the Annex D of the Mamba paper.
+            // => {d_inner, n_seq_tokens, n_seqs} and {d_state, d_inner, n_seqs}
+            return ggml_ssm_scan(ctx, ssm, x, dt, A, B, C, ids);
+        };
+
+        ggml_tensor * y_ssm = build_rs(inp, ssm_states_all, hparams.n_embd_s(), ubatch.n_seqs, get_ssm_rows);
+
+        // store last states
+        ggml_build_forward_expand(
+            gf, ggml_cpy(ctx0, ggml_view_1d(ctx0, y_ssm, d_state * d_inner * n_seqs, x->nb[3] * x->ne[3]),
+                         ggml_view_1d(ctx0, ssm_states_all, d_state * d_inner * n_seqs,
+                                      kv_head * d_state * d_inner * ggml_element_size(ssm_states_all))));
+
+        ggml_tensor * y = ggml_view_3d(ctx0, y_ssm, d_inner, n_seq_tokens, n_seqs, x->nb[2], x->nb[3], 0);
+
+        // TODO: skip computing output earlier for unused tokens
+
+        y = ggml_add(ctx0, y, ggml_mul(ctx0, cur, layer.ssm_d));
+        y = ggml_swiglu_split(ctx0, ggml_cont(ctx0, z), y);
+
+        // {d_inner, n_embd} @ {d_inner, n_seq_tokens, n_seqs} => {n_embd, n_seq_tokens, n_seqs}
+        cur = build_lora_mm(layer.ssm_out, y, layer.ssm_out_s);
+    }
+
+    // {n_embd, n_seq_tokens, n_seqs} => {n_embd, n_tokens}
+    cur = ggml_reshape_2d(ctx0, cur, cur->ne[0], n_seq_tokens * n_seqs);
+
+    return cur;
+}
+
+ggml_tensor * llm_build_mamba_base::build_mamba2_layer(llm_graph_input_rs * inp,
+                                                          ggml_tensor *        cur,
+                                                          const llama_model &  model,
+                                                          const llama_ubatch & ubatch,
+                                                          int                  il) const {
+    const auto * mctx_cur = inp->mctx;
+
+    const auto kv_head = mctx_cur->get_head();
+
+    const int64_t d_conv   = hparams.ssm_d_conv;
+    const int64_t d_inner  = hparams.ssm_d_inner;
+    const int64_t d_state  = hparams.ssm_d_state;
+    const int64_t n_head   = hparams.ssm_dt_rank;
+    const int64_t head_dim = d_inner / n_head;
+    const int64_t n_group  = hparams.ssm_n_group;
+    const int64_t n_seqs   = ubatch.n_seqs;
+
+    const int64_t n_seq_tokens = ubatch.n_seq_tokens;
+
+    GGML_ASSERT(n_seqs != 0);
+    GGML_ASSERT(ubatch.equal_seqs());
+    GGML_ASSERT(ubatch.n_tokens == n_seq_tokens * n_seqs);
+    GGML_ASSERT(d_inner % n_head  == 0);
+    GGML_ASSERT(d_inner % n_group == 0);
+
+    ggml_tensor * conv_states_all = mctx_cur->get_r_l(il);
+    ggml_tensor * ssm_states_all  = mctx_cur->get_s_l(il);
+
+    ggml_tensor * conv = build_rs(inp, conv_states_all, hparams.n_embd_r(), n_seqs);
+    conv               = ggml_reshape_3d(ctx0, conv, d_conv - 1, d_inner + 2 * n_group * d_state, n_seqs);
+
+    // {n_embd, n_tokens} => {n_embd, n_seq_tokens, n_seqs}
+    cur = ggml_reshape_3d(ctx0, cur, cur->ne[0], n_seq_tokens, n_seqs);
+
+    // d_in_proj = 2 * self.d_inner + 2 * self.ngroups * self.d_state + self.nheads
+
+    // {n_embd, d_in_proj} @ {n_embd, n_seq_tokens, n_seqs} => {d_in_proj, n_seq_tokens, n_seqs}
+    ggml_tensor * zxBCdt = build_lora_mm(model.layers[il].ssm_in, cur, model.layers[il].ssm_in_s);
+
+    // split the above in three
+    ggml_tensor * z   = ggml_view_4d(ctx0, zxBCdt, head_dim, n_head, n_seq_tokens, n_seqs, head_dim * zxBCdt->nb[0],
+                                     zxBCdt->nb[1], zxBCdt->nb[2], 0);
+    ggml_tensor * xBC = ggml_view_3d(ctx0, zxBCdt, d_inner + 2 * n_group * d_state, n_seq_tokens, n_seqs, zxBCdt->nb[1],
+                                     zxBCdt->nb[2], d_inner * ggml_element_size(zxBCdt));
+    ggml_tensor * dt  = ggml_view_3d(ctx0, zxBCdt, n_head, n_seq_tokens, n_seqs, zxBCdt->nb[1], zxBCdt->nb[2],
+                                     (2 * d_inner + 2 * n_group * d_state) * ggml_element_size(zxBCdt));
+
+    // conv
+    {
+        // => {d_conv - 1 + n_seq_tokens, d_inner + 2*n_group*d_state, n_seqs}
+        ggml_tensor * conv_x = ggml_concat(ctx0, conv, ggml_transpose(ctx0, xBC), 0);
+
+        // copy last (d_conv - 1) columns back into the state cache
+        ggml_tensor * last_conv = ggml_view_3d(ctx0, conv_x, d_conv - 1, d_inner + 2 * n_group * d_state, n_seqs,
+                                               conv_x->nb[1], conv_x->nb[2], n_seq_tokens * (conv_x->nb[0]));
+
+        ggml_build_forward_expand(gf, ggml_cpy(ctx0, last_conv,
+                                               ggml_view_1d(ctx0, conv_states_all,
+                                                            (d_conv - 1) * (d_inner + 2 * n_group * d_state) * (n_seqs),
+                                                            kv_head * (d_conv - 1) * (d_inner + 2 * n_group * d_state) *
+                                                                ggml_element_size(conv_states_all))));
+
+        // 1D convolution
+        // The equivalent is to make a self-overlapping view of conv_x
+        // over d_conv columns at each stride in the 3rd dimension,
+        // then element-wise multiply that with the conv1d weight,
+        // then sum the elements of each row,
+        // (the last two steps are a dot product over rows (also doable with mul_mat))
+        // then permute away the ne[0] dimension,
+        // and then you're left with the resulting x tensor.
+        // For simultaneous sequences, all sequences need to have the same length.
+        xBC = ggml_ssm_conv(ctx0, conv_x, model.layers[il].ssm_conv1d);
+
+        // bias
+        xBC = ggml_add(ctx0, xBC, model.layers[il].ssm_conv1d_b);
+
+        xBC = ggml_silu(ctx0, xBC);
+    }
+
+    // ssm
+    {
+        // These correspond to V K Q in SSM/attention duality
+        ggml_tensor * x = ggml_view_4d(ctx0, xBC, head_dim, n_head, n_seq_tokens, n_seqs, head_dim * xBC->nb[0],
+                                       xBC->nb[1], xBC->nb[2], 0);
+        ggml_tensor * B = ggml_view_4d(ctx0, xBC, d_state, n_group, n_seq_tokens, n_seqs, d_state * xBC->nb[0],
+                                       xBC->nb[1], xBC->nb[2], d_inner * ggml_element_size(xBC));
+        ggml_tensor * C = ggml_view_4d(ctx0, xBC, d_state, n_group, n_seq_tokens, n_seqs, d_state * xBC->nb[0],
+                                       xBC->nb[1], xBC->nb[2], (d_inner + n_group * d_state) * ggml_element_size(xBC));
+
+        // {n_head, n_seq_tokens, n_seqs}
+        dt = ggml_add(ctx0, ggml_cont(ctx0, dt), model.layers[il].ssm_dt_b);
+
+        ggml_tensor * A = model.layers[il].ssm_a;
+
+        // use the states and the indices provided by build_recurrent_state
+        // (this is necessary in order to properly use the states before they are overwritten,
+        //  while avoiding to make unnecessary copies of the states)
+        auto get_ssm_rows = [&](ggml_context * ctx, ggml_tensor * states, ggml_tensor * ids) {
+            ggml_tensor * ssm = ggml_reshape_4d(ctx, states, d_state, head_dim, n_head, mctx_cur->get_size());
+
+            // TODO: use semistructured matrices to implement state-space duality
+            // => {d_inner, n_seq_tokens, n_seqs} and {d_state, d_inner, n_seqs}
+            return ggml_ssm_scan(ctx, ssm, x, dt, A, B, C, ids);
+        };
+
+        ggml_tensor * y_ssm = build_rs(inp, ssm_states_all, hparams.n_embd_s(), ubatch.n_seqs, get_ssm_rows);
+
+        // store last states
+        ggml_build_forward_expand(
+            gf, ggml_cpy(ctx0, ggml_view_1d(ctx0, y_ssm, d_state * d_inner * n_seqs, ggml_nelements(x) * x->nb[0]),
+                         ggml_view_1d(ctx0, ssm_states_all, d_state * d_inner * n_seqs,
+                                      kv_head * d_state * d_inner * ggml_element_size(ssm_states_all))));
+
+        ggml_tensor * y = ggml_view_4d(ctx0, y_ssm, head_dim, n_head, n_seq_tokens, n_seqs, x->nb[1], n_head * x->nb[1],
+                                       n_seq_tokens * n_head * x->nb[1], 0);
+
+        // TODO: skip computing output earlier for unused tokens
+
+        y = ggml_add(ctx0, y, ggml_mul(ctx0, x, model.layers[il].ssm_d));
+        cb(y, "mamba2_y_add_d", il);
+        y = ggml_swiglu_split(ctx0, ggml_cont(ctx0, z), y);
+
+        // grouped RMS norm
+        if (model.layers[il].ssm_norm) {
+            y = ggml_reshape_4d(ctx0, y, d_inner / n_group, n_group, n_seq_tokens, n_seqs);
+            y = build_norm(y, model.layers[il].ssm_norm, NULL, LLM_NORM_RMS, il);
+        }
+
+        y = ggml_reshape_3d(ctx0, y, d_inner, n_seq_tokens, n_seqs);
+
+        // {d_inner, n_embd} @ {d_inner, n_seq_tokens, n_seqs} => {n_embd, n_seq_tokens, n_seqs}
+        cur = build_lora_mm(model.layers[il].ssm_out, y, model.layers[il].ssm_out_s);
+    }
+
+    // {n_embd, n_seq_tokens, n_seqs} => {n_embd, n_tokens}
+    cur = ggml_reshape_2d(ctx0, cur, cur->ne[0], n_seq_tokens * n_seqs);
+    cb(cur, "mamba_out", il);
+
+    return cur;
+}

src/models/mamba.cpp

@@ -0,0 +1,137 @@
+#include "models.h"
+
+void llama_model_mamba::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
+    ml.get_key(LLM_KV_SSM_STATE_SIZE,     hparams.ssm_d_state);
+    ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
+    ml.get_key(LLM_KV_SSM_DT_B_C_RMS,     hparams.ssm_dt_b_c_rms, false);
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 24:
+            switch (hparams.n_embd) {
+                case 768: type = LLM_TYPE_SMALL; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 48:
+            switch (hparams.n_embd) {
+                case 1024: type = LLM_TYPE_MEDIUM; break;
+                case 1536: type = LLM_TYPE_LARGE; break;
+                case 2048: type = LLM_TYPE_XL; break;
+                default:   type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 64:
+            switch (hparams.n_embd) {
+                case 2560: type = LLM_TYPE_3B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_mamba::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t d_conv  = hparams.ssm_d_conv;
+    const int64_t d_inner = hparams.ssm_d_inner;
+    const int64_t d_state = hparams.ssm_d_state;
+    const int64_t dt_rank = hparams.ssm_dt_rank;
+
+    // only an expansion factor of 2 is supported for now
+    if (2 * n_embd != d_inner) {
+        throw std::runtime_error("only an expansion factor of 2 is supported for now");
+    }
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed, duplicated to allow offloading
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // norm
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, 2*d_inner}, 0);
+
+        layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner}, 0);
+        layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner}, 0);
+
+        layer.ssm_x = create_tensor(tn(LLM_TENSOR_SSM_X, "weight", i), {d_inner, dt_rank + 2*d_state}, 0);
+
+        layer.ssm_dt = create_tensor(tn(LLM_TENSOR_SSM_DT, "weight", i), {dt_rank, d_inner}, 0);
+        layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {d_inner}, 0);
+
+        // no "weight" suffix for these
+        layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {d_state, d_inner}, 0);
+        layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {d_inner}, 0);
+
+        // out_proj
+        layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_mamba::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_mamba::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_build_mamba_base(params) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    // {n_embd, n_tokens}
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * rs_inp = build_rs_inp();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        if (model.arch == LLM_ARCH_MAMBA2) {
+            cur = build_mamba2_layer(rs_inp, cur, model, ubatch, il);
+        } else {
+            cur = build_mamba_layer(rs_inp, cur, model, ubatch, il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // residual
+        cur = ggml_add(ctx0, cur, inpL);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    // final rmsnorm
+    cur = build_norm(inpL, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/mamba2.cpp

@@ -0,0 +1,88 @@
+#include "models.h"
+
+void llama_model_mamba2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
+    ml.get_key(LLM_KV_SSM_STATE_SIZE,     hparams.ssm_d_state);
+    ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
+    ml.get_key(LLM_KV_SSM_GROUP_COUNT,    hparams.ssm_n_group);
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 24:
+            switch (hparams.n_embd) {
+                case 768: type = LLM_TYPE_SMALL; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 48:
+            switch (hparams.n_embd) {
+                case 1024: type = LLM_TYPE_MEDIUM; break;
+                case 1536: type = LLM_TYPE_LARGE; break;
+                case 2048: type = LLM_TYPE_XL; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        case 64:
+            switch (hparams.n_embd) {
+                case 2560: type = LLM_TYPE_3B; break;
+                case 4096: type = LLM_TYPE_7B; break;
+                default: type = LLM_TYPE_UNKNOWN;
+            } break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_mamba2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t d_conv  = hparams.ssm_d_conv;
+    const int64_t d_inner = hparams.ssm_d_inner;
+    const int64_t d_state = hparams.ssm_d_state;
+    const int64_t n_group = hparams.ssm_n_group;
+    const int64_t dt_rank  = hparams.ssm_dt_rank;
+
+    const int64_t conv_dim = d_inner + 2 * n_group * d_state;
+    const int64_t d_in_proj = d_inner + conv_dim + dt_rank;
+
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    {
+        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+        output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+        // if output is NULL, init from the input tok embed, duplicated to allow offloading
+        if (output == NULL) {
+            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+        }
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // norm
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, d_in_proj}, 0);
+
+        layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner + 2*n_group*d_state}, 0);
+        layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner + 2*n_group*d_state}, 0);
+
+        layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {dt_rank}, 0);
+
+        // no "weight" suffix for these
+        layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {1, dt_rank}, 0);
+        layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {1, dt_rank}, 0);
+
+        layer.ssm_norm = create_tensor(tn(LLM_TENSOR_SSM_NORM, "weight", i), {d_inner / n_group, n_group}, 0);
+
+        // out_proj
+        layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_mamba2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/mellum.cpp

@@ -0,0 +1,225 @@
+#include "models.h"
+
+void llama_model_mellum::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa, false);
+
+    if (hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+
+        uint32_t swa_period = 4;
+        const auto res = ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        if (res) {
+            hparams.set_swa_pattern(swa_period);
+        } else {
+            ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+        }
+
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    } else {
+        hparams.swa_type = LLAMA_SWA_TYPE_NONE;
+    }
+
+    switch (hparams.n_layer()) {
+        case 28: type = LLM_TYPE_12B_A2_5B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_mellum::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+        if (n_expert == 0) {
+            throw std::runtime_error("n_expert must be > 0 for Mellum");
+        }
+        if (n_expert_used == 0) {
+            throw std::runtime_error("n_expert_used must be > 0 for Mellum");
+        }
+
+        const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_mellum::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_STANDARD) {
+        return std::make_unique<graph<true>>(*this, params);
+    }
+    return std::make_unique<graph<false>>(*this, params);
+}
+
+template <bool iswa>
+llama_model_mellum::graph<iswa>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    using inp_attn_type = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa, llm_graph_input_attn_kv>;
+    inp_attn_type * inp_attn = nullptr;
+
+    if constexpr (iswa) {
+        inp_attn = build_attn_inp_kv_iswa();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, nullptr,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, nullptr, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, nullptr, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            const bool is_swa = hparams.is_swa(il);
+
+            if (is_swa) {
+                // For sliding window layers, use regular rope with no yarn rope scaling.
+                // This is achieved here by setting freq_scale and attn_factor to 1.
+                // We also set ext_factor to 0 to avoid a few unnecessary computations.
+                Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, 1.0,
+                    0.0, 1.0, beta_fast, beta_slow
+                    );
+
+                Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, 1.0,
+                    0.0, 1.0, beta_fast, beta_slow
+                    );
+            } else {
+                Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+                Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // MoE
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, nullptr,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        ggml_tensor * moe_out =
+            build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, true,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il,
+                    nullptr, nullptr,
+                    model.layers[il].ffn_up_exps_s,
+                    model.layers[il].ffn_gate_exps_s,
+                    model.layers[il].ffn_down_exps_s);
+        cb(moe_out, "ffn_moe_out", il);
+        cur = moe_out;
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, nullptr,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+template struct llama_model_mellum::graph<false>;
+template struct llama_model_mellum::graph<true>;

src/models/mimo2.cpp

@@ -0,0 +1,235 @@
+#include "models.h"
+
+void llama_model_mimo2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,   hparams.n_swa);
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA,         hparams.rope_freq_base_train_swa, false);
+
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+
+    float value_scale = 0.0f;
+    if (ml.get_key(LLM_KV_ATTENTION_VALUE_SCALE, value_scale, false) && value_scale != 1.0f) {
+        hparams.f_attn_value_scale = value_scale;
+    }
+
+    ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
+    GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl");
+
+    switch (hparams.n_layer()) {
+        case 48: type = LLM_TYPE_310B_A15B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_mimo2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer_all; ++i) {
+        auto & layer = layers[i];
+        uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(i);
+        uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(i);
+        uint32_t n_head = hparams.n_head(i);
+
+        // NextN/MTP layers (the last n_nextn blocks) are preserved but disabled pending support
+        const bool is_nextn = i >= n_layer;
+        const int  skip     = is_nextn ? TENSOR_SKIP : 0;
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, skip);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_v * n_head, n_embd }, skip);
+
+        layer.attn_norm  = create_tensor(tn(LLM_TENSOR_ATTN_NORM,  "weight", i), {n_embd}, skip);
+        layer.attn_sinks = create_tensor(tn(LLM_TENSOR_ATTN_SINKS, "weight", i), {n_head}, TENSOR_NOT_REQUIRED | skip);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, skip);
+
+        // non-MoE branch
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, TENSOR_NOT_REQUIRED | skip);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, TENSOR_NOT_REQUIRED | skip);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, TENSOR_NOT_REQUIRED | skip);
+
+        // MoE branch
+        int64_t n_ff_exp = hparams.n_ff_exp;
+        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, TENSOR_NOT_REQUIRED | skip);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff_exp,   n_expert}, TENSOR_NOT_REQUIRED | skip);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, TENSOR_NOT_REQUIRED | skip);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff_exp,   n_expert}, TENSOR_NOT_REQUIRED | skip);
+        layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED | skip);
+
+        if (is_nextn) {
+            layer.nextn.eh_proj  = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), {2 * n_embd, n_embd}, skip);
+            layer.nextn.enorm    = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM,   "weight", i), {n_embd}, skip);
+            layer.nextn.hnorm    = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM,   "weight", i), {n_embd}, skip);
+            layer.layer_out_norm = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, skip);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_mimo2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_mimo2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    ggml_tensor * inp_pos = build_inp_pos();
+    auto * inp_attn = build_attn_inp_kv_iswa();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    const float v_scale = hparams.f_attn_value_scale;
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        uint32_t n_head_l    = hparams.n_head(il);
+        uint32_t n_head_kv_l = hparams.n_head_kv(il);
+        const float freq_base_l  = model.get_rope_freq_base(cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        cur = inpL;
+
+        // self_attention
+        {
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+
+            ggml_tensor * Qcur;
+            ggml_tensor * Kcur;
+            ggml_tensor * Vcur;
+
+            if (model.layers[il].wqkv) {
+                // Fused qkv_proj - Q/K share head_dim_k, V uses head_dim_v
+                ggml_tensor * qkv = build_lora_mm(model.layers[il].wqkv, cur);
+                cb(qkv, "wqkv", il);
+
+                const size_t row_k    = ggml_row_size(qkv->type, n_embd_head_k);
+                const size_t row_v    = ggml_row_size(qkv->type, n_embd_head_v);
+                const size_t row_full = qkv->nb[1];
+                const size_t k_off    = row_k * n_head_l;
+                const size_t v_off    = k_off + row_k * n_head_kv_l;
+
+                Qcur = ggml_view_3d(ctx0, qkv, n_embd_head_k, n_head_l,    n_tokens, row_k, row_full, 0);
+                Kcur = ggml_view_3d(ctx0, qkv, n_embd_head_k, n_head_kv_l, n_tokens, row_k, row_full, k_off);
+                Vcur = ggml_view_3d(ctx0, qkv, n_embd_head_v, n_head_kv_l, n_tokens, row_v, row_full, v_off);
+            } else {
+                // Split path
+                Qcur = build_lora_mm(model.layers[il].wq, cur);
+                cb(Qcur, "Qcur", il);
+
+                Kcur = build_lora_mm(model.layers[il].wk, cur);
+                cb(Kcur, "Kcur", il);
+
+                Vcur = build_lora_mm(model.layers[il].wv, cur);
+                cb(Vcur, "Vcur", il);
+
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head_k, n_head_l,    n_tokens);
+                Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head_k, n_head_kv_l, n_tokens);
+                Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head_v, n_head_kv_l, n_tokens);
+            }
+
+            Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+            Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            ggml_tensor * sinks = model.layers[il].attn_sinks;
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, sinks, nullptr, 1.0f/sqrtf(float(n_embd_head_k)), il);
+            cb(cur, "attn_out", il);
+
+            if (v_scale) {
+                cur = ggml_scale(ctx0, cur, v_scale);
+                cb(cur, "attn_out_scaled", il);
+            }
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+            // dense branch
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            cur = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, true,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID,
+                    il);
+            cb(cur, "ffn_moe_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/minicpm.cpp

@@ -0,0 +1,89 @@
+#include "models.h"
+
+void llama_model_minicpm::load_arch_hparams(llama_model_loader & ml) {
+    // Backward-compatible defaults for older MiniCPM GGUFs
+    hparams.f_embedding_scale = 12.0f;
+    hparams.f_residual_scale  = 1.4f / sqrtf(float(hparams.n_layer()));
+    hparams.f_logit_scale     = hparams.n_embd ? (256.0f / float(hparams.n_embd)) : 1.0f;
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    // Optional KV reads, override defaults if present in newer GGUF exports
+    ml.get_key(LLM_KV_EMBEDDING_SCALE, hparams.f_embedding_scale, /*required=*/false);
+    ml.get_key(LLM_KV_RESIDUAL_SCALE, hparams.f_residual_scale, /*required=*/false);
+    ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale, /*required=*/false);
+
+    // MiniCPM uses rope by default, unlike Granite which uses it as a switch
+    hparams.rope_finetuned = true;
+
+    switch (hparams.n_layer()) {
+        case 52: type = LLM_TYPE_1B; break;
+        case 40: type = LLM_TYPE_2B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_minicpm::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+        else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        if (n_expert == 0) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+            // optional MLP bias
+            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        } else {
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+            // For Granite MoE Shared
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd}, 0);
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_minicpm::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/minicpm3.cpp

@@ -0,0 +1,260 @@
+#include "models.h"
+
+void llama_model_minicpm3::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK,       hparams.n_lora_q);
+    ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK,      hparams.n_lora_kv);
+
+    switch (hparams.n_layer()) {
+        case 62: type = LLM_TYPE_4B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_minicpm3::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_embd_head_qk_rope = hparams.n_rot();
+    const int64_t n_embd_head_qk_nope = hparams.n_embd_head_k() - hparams.n_rot();
+
+    const int64_t q_lora_rank  = hparams.n_lora_q;
+    const int64_t kv_lora_rank = hparams.n_lora_kv;
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_q_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, 0);
+
+        layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, 0);
+
+        layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, 0);
+        layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k}, 0);
+
+        layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + (n_embd_head_qk_rope)}, 0);
+        layer.wkv_b     = create_tensor(tn(LLM_TENSOR_ATTN_KV_B,     "weight", i), {kv_lora_rank, n_head * (n_embd_head_qk_nope + n_embd_head_v)}, 0);
+        layer.wo        = create_tensor(tn(LLM_TENSOR_ATTN_OUT,      "weight", i), {              n_head * (                      n_embd_head_v), n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+        layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), { n_embd_head_qk_rope/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), { n_embd_head_qk_rope/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_minicpm3::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_minicpm3::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    //TODO: if the model varies, these parameters need to be read from the model
+    const int64_t n_embd_base = 256;
+    const float scale_embd  = 12.0f;
+    const float scale_depth = 1.4f;
+    const float kq_scale = 1.0f / sqrtf(float(hparams.n_embd_head_k()));
+
+    const uint32_t n_embd_head_qk_rope = hparams.n_rot();
+    const uint32_t n_embd_head_qk_nope = hparams.n_embd_head_k() - hparams.n_rot();
+
+    const uint32_t kv_lora_rank = hparams.n_lora_kv;
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // scale the input embeddings
+    inpL = ggml_scale(ctx0, inpL, scale_embd);
+    cb(inpL, "inp_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            ggml_tensor * q = NULL;
+            // {n_embd, q_lora_rank} * {n_embd, n_tokens} -> {q_lora_rank, n_tokens}
+            q = ggml_mul_mat(ctx0, model.layers[il].wq_a, cur);
+            cb(q, "q", il);
+
+            q = build_norm(q,
+                    model.layers[il].attn_q_a_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(q, "q", il);
+
+            // {q_lora_rank, n_head * hparams.n_embd_head_k()} * {q_lora_rank, n_tokens} -> {n_head * hparams.n_embd_head_k(), n_tokens}
+            q = ggml_mul_mat(ctx0, model.layers[il].wq_b, q);
+            cb(q, "q", il);
+
+            // split into {n_head * n_embd_head_qk_nope, n_tokens}
+            ggml_tensor * q_nope = ggml_view_3d(ctx0, q, n_embd_head_qk_nope, n_head, n_tokens,
+                    ggml_row_size(q->type, hparams.n_embd_head_k()),
+                    ggml_row_size(q->type, hparams.n_embd_head_k() * n_head),
+                    0);
+            cb(q_nope, "q_nope", il);
+
+            // and {n_head * n_embd_head_qk_rope, n_tokens}
+            ggml_tensor * q_pe = ggml_view_3d(ctx0, q, n_embd_head_qk_rope, n_head, n_tokens,
+                    ggml_row_size(q->type, hparams.n_embd_head_k()),
+                    ggml_row_size(q->type, hparams.n_embd_head_k() * n_head),
+                    ggml_row_size(q->type, n_embd_head_qk_nope));
+            cb(q_pe, "q_pe", il);
+
+            // {n_embd, kv_lora_rank + n_embd_head_qk_rope} * {n_embd, n_tokens} -> {kv_lora_rank + n_embd_head_qk_rope, n_tokens}
+            ggml_tensor * kv_pe_compresseed = ggml_mul_mat(ctx0, model.layers[il].wkv_a_mqa, cur);
+            cb(kv_pe_compresseed, "kv_pe_compresseed", il);
+
+            // split into {kv_lora_rank, n_tokens}
+            ggml_tensor * kv_compressed = ggml_view_2d(ctx0, kv_pe_compresseed, kv_lora_rank, n_tokens,
+                    kv_pe_compresseed->nb[1],
+                    0);
+            cb(kv_compressed, "kv_compressed", il);
+
+            // and {n_embd_head_qk_rope, n_tokens}
+            ggml_tensor * k_pe = ggml_view_3d(ctx0, kv_pe_compresseed, n_embd_head_qk_rope, 1, n_tokens,
+                    kv_pe_compresseed->nb[1],
+                    kv_pe_compresseed->nb[1],
+                    ggml_row_size(kv_pe_compresseed->type, kv_lora_rank));
+            cb(k_pe, "k_pe", il);
+
+            kv_compressed = build_norm(kv_compressed,
+                    model.layers[il].attn_kv_a_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(kv_compressed, "kv_compressed", il);
+
+            // {kv_lora_rank, n_head * (n_embd_head_qk_nope + n_embd_head_v)} * {kv_lora_rank, n_tokens} -> {n_head * (n_embd_head_qk_nope + n_embd_head_v), n_tokens}
+            ggml_tensor * kv = ggml_mul_mat(ctx0, model.layers[il].wkv_b, kv_compressed);
+            cb(kv, "kv", il);
+
+            // split into {n_head * n_embd_head_qk_nope, n_tokens}
+            ggml_tensor * k_nope = ggml_view_3d(ctx0, kv, n_embd_head_qk_nope, n_head, n_tokens,
+                    ggml_row_size(kv->type, n_embd_head_qk_nope + hparams.n_embd_head_v()),
+                    ggml_row_size(kv->type, n_head * (n_embd_head_qk_nope + hparams.n_embd_head_v())),
+                    0);
+            cb(k_nope, "k_nope", il);
+
+            // and {n_head * n_embd_head_v, n_tokens}
+            ggml_tensor * v_states = ggml_view_3d(ctx0, kv, hparams.n_embd_head_v(), n_head, n_tokens,
+                    ggml_row_size(kv->type, (n_embd_head_qk_nope + hparams.n_embd_head_v())),
+                    ggml_row_size(kv->type, (n_embd_head_qk_nope + hparams.n_embd_head_v())*n_head),
+                    ggml_row_size(kv->type, (n_embd_head_qk_nope)));
+            cb(v_states, "v_states", il);
+
+            v_states = ggml_cont(ctx0, v_states);
+            cb(v_states, "v_states", il);
+
+            q_pe = ggml_rope_ext(
+                    ctx0, q_pe, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+            cb(q_pe, "q_pe", il);
+
+            // shared RoPE key
+            k_pe = ggml_rope_ext(
+                    ctx0, k_pe, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+            cb(k_pe, "k_pe", il);
+
+            ggml_tensor * q_states = ggml_concat(ctx0, q_nope, q_pe, 0);
+            cb(q_states, "q_states", il);
+
+            ggml_tensor * k_states = ggml_concat(ctx0, k_nope, ggml_repeat(ctx0, k_pe, q_pe), 0);
+            cb(k_states, "k_states", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    q_states, k_states, v_states, nullptr, nullptr, nullptr, kq_scale, il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        // scale_res - scale the hidden states for residual connection
+        const float scale_res = scale_depth/sqrtf(float(n_layer)); // TODO: is this correct?
+        cur = ggml_scale(ctx0, cur, scale_res);
+        cb(cur, "hidden_scaled", il);
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        {
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        // scale the hidden states for residual connection
+        cur = ggml_scale(ctx0, cur, scale_res);
+        cb(cur, "hidden_scaled_ffn", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head scaling
+    const float scale_lmhead = float(n_embd_base)/float(n_embd);
+    cur = ggml_scale(ctx0, cur, scale_lmhead);
+    cb(cur, "lmhead_scaling", -1);
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/minimax-m2.cpp

@@ -0,0 +1,169 @@
+#include "models.h"
+
+void llama_model_minimax_m2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,  hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,   hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,           hparams.expert_gating_func, false);
+
+    switch (hparams.n_layer()) {
+        case 62: type = LLM_TYPE_230B_A10B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_minimax_m2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k * n_head}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_k_gqa}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff,   n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff,   n_expert}, 0);
+        layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_minimax_m2::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_minimax_m2::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    // GGML_ASSERT(n_embd_head == n_rot); this is wrong in case of minimax, head_dim = 128, n_rot = 64
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    ggml_tensor * inp_pos = build_inp_pos();
+    auto inp_attn = build_attn_inp_kv();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        res->t_layer_inp[il] = inpL;
+
+        ggml_tensor * inpSA = inpL;
+
+        cur = inpL;
+
+        // self_attention
+        {
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+
+            // compute Q and K and RoPE them
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            cb(Qcur, "Qcur", il);
+
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+            cb(Kcur, "Kcur", il);
+
+            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+            cb(Vcur, "Vcur", il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+            Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // MoE branch
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                model.layers[il].ffn_exp_probs_b,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, true,
+                hparams.expert_weights_scale,
+                (llama_expert_gating_func_type) hparams.expert_gating_func,
+                il);
+        cb(cur, "ffn_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/minimax-m3.cpp

@@ -0,0 +1,562 @@
+#include "models.h"
+#include "llama-kv-cache.h"
+#include <cmath>
+#include <vector>
+#include <algorithm>
+#include <cstdint>
+
+// MiniMax-M3: MiniMax-M2 style GQA (per-head QK-norm, partial rotary) with
+// DeepSeek-V3 leading-dense + routed/shared experts (sigmoid gating, routed scaling),
+// swigluoai activation, and MiniMax Sparse Attention (MSA). MTP is not in released model weights.
+// Notes: Blocks are anchored to absolute KV cache slots.
+
+void llama_model_minimax_m3::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_HEAD_COUNT,    hparams.indexer_n_head);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_KEY_LENGTH,    hparams.indexer_head_size);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_TOP_K,         hparams.indexer_top_k);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_BLOCK_SIZE,    hparams.indexer_block_size);
+    ml.get_key(LLM_KV_ATTENTION_INDEXER_LOCAL_BLOCKS,  hparams.indexer_local_blocks);
+    msa_p = { (int) hparams.indexer_block_size, (int) hparams.indexer_top_k, (int) hparams.indexer_local_blocks };
+    hparams.indexer_kv = true;
+
+    switch (hparams.n_layer()) {
+        case 60: type = LLM_TYPE_428B_A23B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_minimax_m3::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+    const int64_t n_expert_shared = hparams.n_expert_shared;
+    const int64_t n_ff_exp        = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        // per-head QK-norm: a single head_dim vector applied to every head
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (i < (int) hparams.n_layer_dense_lead) {
+            // leading dense layers
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        } else {
+            // routed experts
+            layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,    "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias",   i), {n_expert}, 0);
+            layer.ffn_gate_exps   = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS,   "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+            layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS,   "weight", i), {n_ff_exp, n_embd, n_expert}, 0);
+            layer.ffn_up_exps     = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,     "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+
+            // shared expert
+            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+
+            // indexer
+            layer.index_q_proj = create_tensor(tn(LLM_TENSOR_INDEXER_Q_PROJ, "weight", i), {n_embd, hparams.indexer_n_head * hparams.indexer_head_size}, 0);
+            layer.index_k_proj = create_tensor(tn(LLM_TENSOR_INDEXER_K_PROJ, "weight", i), {n_embd, hparams.indexer_head_size}, 0);
+            layer.index_q_norm = create_tensor(tn(LLM_TENSOR_INDEXER_Q_NORM, "weight", i), {hparams.indexer_head_size}, 0);
+            layer.index_k_norm = create_tensor(tn(LLM_TENSOR_INDEXER_K_NORM, "weight", i), {hparams.indexer_head_size}, 0);
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_minimax_m3::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+// per-query local-force bias for MSA selection
+// local window always wins a slot
+class llm_graph_input_msa_local : public llm_graph_input_i {
+public:
+    llm_graph_input_msa_local(int blk, int local, int64_t nblk) : blk(blk), local(local), nblk(nblk) {}
+
+    void set_input(const llama_ubatch * ubatch) override {
+        if (!bias || !ubatch->pos) {
+            return;
+        }
+        const int64_t n_tokens = ubatch->n_tokens;
+        std::vector<float> data((size_t) nblk * n_tokens, 0.0f);
+        for (int64_t i = 0; i < n_tokens; ++i) {
+            const int64_t L = ubatch->pos[i] / blk;
+            for (int l = 0; l < local && L - l >= 0; ++l) {
+                if (L - l < nblk) {
+                    data[(size_t) i * nblk + (L - l)] = 1e30f;
+                }
+            }
+        }
+        ggml_backend_tensor_set(bias, data.data(), 0, data.size() * sizeof(float));
+    }
+
+    // valid as long as the bias tensor dims still match the new ubatch/cache window
+    bool can_reuse(const llm_graph_params & params) override {
+        const auto * mctx = static_cast<const llama_kv_cache_context *>(params.mctx);
+
+        bool res = true;
+        res &= bias->ne[1] == params.ubatch.n_tokens;
+        res &= bias->ne[0] * blk == (int64_t) mctx->get_n_kv();
+        return res;
+    }
+
+    ggml_tensor * bias = nullptr;
+    int     blk;
+    int     local;
+    int64_t nblk;
+};
+
+// pooled score of a block with no visible token: -inf from the mask, or -FLT_MAX from the
+// max-pool identity when every element of the block is -inf
+static inline bool msa_score_masked(float x) { return x <= -1e30f; }
+
+// MSA block selection (batch regime)
+// CPU custom op, the token-level expansion and the combination with the causal mask happen on the GPU.
+static void msa_block_mask_op(struct ggml_tensor * dst, int ith, int nth, void * userdata) {
+    const struct ggml_tensor * bs   = dst->src[0];
+    const struct ggml_tensor * bias = dst->src[1];
+    const msa_params * p = (const msa_params *) userdata;
+
+    const int nblk = (int) bs->ne[0];
+    const int Hd   = (int) bs->ne[1];
+    const int S    = (int) bs->ne[2];
+
+    GGML_ASSERT(bs->type   == GGML_TYPE_F32 && ggml_is_contiguous(bs));
+    GGML_ASSERT(bias->type == GGML_TYPE_F32 && ggml_is_contiguous(bias));
+    GGML_ASSERT(dst->type  == GGML_TYPE_F16 && ggml_is_contiguous(dst));
+    GGML_ASSERT(dst->ne[0] == nblk && dst->ne[1] == S && dst->ne[2] == Hd);
+    GGML_ASSERT(bias->ne[0] == nblk && bias->ne[1] == S);
+
+    const int topk = p->topk_blocks < nblk ? p->topk_blocks : nblk;
+
+    const ggml_fp16_t f16_zero = ggml_fp32_to_fp16(0.0f);
+    const ggml_fp16_t f16_ninf = ggml_fp32_to_fp16(-INFINITY);
+
+    std::vector<float> rank(nblk);
+    std::vector<char>  valid(nblk);
+    std::vector<int>   ord(nblk);
+
+    ggml_fp16_t * out = (ggml_fp16_t *) dst->data;
+
+    for (int i = ith; i < S; i += nth) {
+        const float * bias_col = (const float *) bias->data + (size_t) i * nblk;
+        for (int h = 0; h < Hd; ++h) {
+            const float * bs_col = (const float *) bs->data + ((size_t) i * Hd + h) * nblk;
+
+            for (int bk = 0; bk < nblk; ++bk) {
+                // a block is selectable if it has a visible token or is locally forced
+                valid[bk] = !msa_score_masked(bs_col[bk]) || bias_col[bk] > 0.0f;
+                rank [bk] = bias_col[bk] > 0.0f ? bias_col[bk] : bs_col[bk];
+                ord  [bk] = bk;
+            }
+
+            std::partial_sort(ord.begin(), ord.begin() + topk, ord.end(),
+                              [&](int a, int b) { return rank[a] > rank[b]; });
+
+            ggml_fp16_t * dst_col = out + ((size_t) h * S + i) * nblk;
+            for (int bk = 0; bk < nblk; ++bk) {
+                dst_col[bk] = f16_ninf;
+            }
+            for (int t = 0; t < topk; ++t) {
+                const int bk = ord[t];
+                if (!valid[bk]) {
+                    break;   // sorted desc: first invalid -> fewer than topk selectable blocks
+                }
+                dst_col[bk] = f16_zero;
+            }
+        }
+    }
+}
+
+// One FA call for all GQA groups (and at multi-stream decode, all streams) by mapping them onto the FA sequence dim (ne[3])
+ggml_tensor * llama_model_minimax_m3::graph::build_attn_msa_fa(
+        ggml_tensor * q_cur,   // [D, HQ, T]
+        ggml_tensor * k,       // [D, n_keys, 1, C]
+        ggml_tensor * v,       // [D, n_keys, 1, C]
+        ggml_tensor * mask,    // [n_keys, R, 1, C] f16, contiguous
+        int64_t Gp, float kq_scale, int il) const {
+
+    const int64_t D  = q_cur->ne[0];
+    const int64_t HQ = q_cur->ne[1];
+    const int64_t T  = q_cur->ne[2];
+    const int64_t C  = k->ne[3];
+    const int64_t R  = HQ*T/(Gp*C);
+    GGML_ASSERT(Gp*C*R == HQ*T);
+    GGML_ASSERT(mask->type == GGML_TYPE_F16);
+
+    // [D, HQ, T] -> [D, Gp, C, R] -> [D, R, Gp, C]
+    // batch  (C=HKV,   R=T): channel = group
+    // decode (C=HKV*ns, R=1): channel = (group, stream), group innermost
+    ggml_tensor * q = ggml_reshape_4d(ctx0, q_cur, D, Gp, C, R);
+    q = ggml_permute(ctx0, q, 0, 2, 3, 1);
+
+    ggml_tensor * o = ggml_flash_attn_ext(ctx0, q, k, v, mask, kq_scale,
+                                          hparams.f_max_alibi_bias, 0.0f);
+    ggml_flash_attn_ext_set_prec(o, GGML_PREC_F32);
+    cb(o, "msa_fattn", il);
+
+    // [D, Gp, R, C] -> [D, Gp, C, R] -> [n_embd, T]
+    o = ggml_permute(ctx0, o, 0, 1, 3, 2);
+    if (!ggml_is_contiguous(o)) {
+        o = ggml_cont(ctx0, o);   // no-op layout at decode (R == 1), copy at batch
+    }
+    return ggml_reshape_2d(ctx0, o, D*HQ, T);
+}
+
+llama_model_minimax_m3::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    const auto & mm = static_cast<const llama_model_minimax_m3 &>(model);
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    // partial rotary: head_dim != n_rot, so don't assert n_embd_head == n_rot
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    ggml_tensor * inp_pos = build_inp_pos();
+    auto inp_attn = build_attn_inp_kv();
+
+    // MSA calls ggml_flash_attn_ext directly and assumes the non-transposed V layout that
+    // llama.cpp only provides when flash attention is enabled. Block selection is anchored
+    // to absolute KV cache slots, which equal positions only for append-only per-stream
+    // caches either a single sequence, or multiple sequences with kv_unified == false (each
+    // stream then has its own slot space). A unified cache with multiple sequences
+    // interleaves slots and would silently break block anchoring so it falls back to dense.
+    const bool fa_on       = cparams.flash_attn;
+    const bool streams_ok  = cparams.n_seq_max == 1 || !cparams.kv_unified;
+    const bool msa_enabled = fa_on && streams_ok;
+
+    static bool warned_no_fa = false;
+    if (!fa_on && !warned_no_fa) {
+        LLAMA_LOG_WARN("%s: flash attention disabled; MSA requires it -> running DENSE attention "
+                       "(output may be degraded). Enable flash attention for MSA.\n", __func__);
+        warned_no_fa = true;
+    }
+    static bool warned_unified = false;
+    if (fa_on && !streams_ok && !warned_unified) {
+        LLAMA_LOG_WARN("%s: unified KV cache with n_seq_max > 1; MSA needs per-sequence streams "
+                       "-> running DENSE attention. Output may be degraded. Drop --kv-unified to enable MSA.\n", __func__);
+        warned_unified = true;
+    }
+
+    // hoisted per-graph MSA state (shared by every sparse layer)
+    llm_graph_input_msa_local * msa_loc = nullptr;
+    ggml_tensor * msa_kqm = nullptr;
+    ggml_tensor * msa_mf  = nullptr;
+    int64_t n_kv = 0, nblk = 0, ns = 1, n_tps = 0;
+    bool msa_decode = false;           // gather (1 token per stream) vs mask
+    const int     blk = mm.msa_p.blk;
+    const int64_t Hd  = hparams.indexer_n_head;   // one indexer head per GQA group
+
+    if (msa_enabled) {
+        msa_kqm = inp_attn->get_kq_mask();
+        n_kv  = msa_kqm->ne[0];
+        n_tps = msa_kqm->ne[1];        // tokens per stream
+        ns    = msa_kqm->ne[3];        // streams in this ubatch
+        GGML_ASSERT(msa_kqm->type == GGML_TYPE_F16 && "MSA requires the FA (f16) mask");
+        GGML_ASSERT(n_tps*ns == n_tokens);
+        GGML_ASSERT(n_kv % blk == 0 &&
+            "MSA: KV/mask n_kv must be a multiple of indexer.block_size (128); "
+            "the flash-attention KV padding must be a multiple of the block size. "
+            "A non-multiple would silently drop the partial tail block.");
+        nblk = n_kv / blk;
+        msa_decode = n_tps == 1;
+
+        msa_mf = ggml_cast(ctx0, msa_kqm, GGML_TYPE_F32);
+
+        auto loc = std::make_unique<llm_graph_input_msa_local>(blk, mm.msa_p.local, nblk);
+        loc->bias = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, nblk, n_tokens);  // stream-grouped tokens
+        ggml_set_input(loc->bias);
+        msa_loc = (llm_graph_input_msa_local *) res->add_input(std::move(loc));
+    }
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // self-attention
+        {
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // per-head QK RMSNorm (weights already include Gemma's +1)
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            // partial rotary: only the first n_rot dims are rotated
+            Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow);
+            Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            const bool is_sparse = msa_enabled && il >= (int) hparams.n_layer_dense_lead;
+
+            if (!is_sparse) {
+                cur = build_attn(inp_attn, model.layers[il].wo, NULL, model.layers[il].wo_s,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, nullptr,
+                        1.0f/sqrtf(float(n_embd_head)), il);
+            } else {
+                const int64_t n_idx_dim = hparams.indexer_head_size;   // 128
+
+                GGML_ASSERT(!inp_attn->self_k_rot && !inp_attn->self_v_rot && "MSA: attn-rot not supported");
+
+                // Index Branch, project, norm, partial RoPE, cache
+                ggml_tensor * iq = build_lora_mm(model.layers[il].index_q_proj, cur);
+                ggml_tensor * ik = build_lora_mm(model.layers[il].index_k_proj, cur);
+                iq = ggml_reshape_3d(ctx0, iq, n_idx_dim, Hd, n_tokens);
+                ik = ggml_reshape_3d(ctx0, ik, n_idx_dim, 1,  n_tokens);
+                iq = build_norm(iq, model.layers[il].index_q_norm, NULL, LLM_NORM_RMS, il);  // +1 baked
+                ik = build_norm(ik, model.layers[il].index_k_norm, NULL, LLM_NORM_RMS, il);
+                iq = ggml_rope_ext(ctx0, iq, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig,
+                                   freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
+                ik = ggml_rope_ext(ctx0, ik, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig,
+                                   freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
+
+                const auto * mctx_cur = inp_attn->mctx;
+                ggml_build_forward_expand(gf, mctx_cur->cpy_k_idx(ctx0, ik, inp_attn->get_k_idxs(), il));
+                ggml_tensor * ik_kv = mctx_cur->get_k_idx(ctx0, il);
+
+                // Main branch: store K/V, take cache views
+                ggml_build_forward_expand(gf, Qcur);
+                ggml_build_forward_expand(gf, Kcur);
+                ggml_build_forward_expand(gf, Vcur);
+                ggml_build_forward_expand(gf, mctx_cur->cpy_k(ctx0, Kcur, inp_attn->get_k_idxs(), il));
+                ggml_build_forward_expand(gf, mctx_cur->cpy_v(ctx0, Vcur, inp_attn->get_v_idxs(), il));
+                ggml_tensor * k = mctx_cur->get_k(ctx0, il);
+                ggml_tensor * v = mctx_cur->get_v(ctx0, il);
+                GGML_ASSERT(!(v->nb[1] > v->nb[2]) && "MSA assumes v_trans=false (FA on)");
+
+                const int64_t D   = k->ne[0];
+                const int64_t HKV = k->ne[1];
+                const int64_t Gp  = n_head/HKV;
+                GGML_ASSERT(HKV == Hd && "MSA: one indexer head per GQA group");
+                GGML_ASSERT(k->ne[3] == ns);
+                const int K = mm.msa_p.topk_blocks < (int) nblk ? mm.msa_p.topk_blocks : (int) nblk;
+
+                const float kq_scale = 1.0f/sqrtf(float(n_embd_head));
+
+                if (msa_decode) {
+                    // decode: batched over streams top-k + gather, one grouped FA
+                    // scores: per-stream batched matmul over the stream dim (ne[3]).
+                    // the cache views are not contiguous across streams (stride = kv_size, not n_kv)
+                    ggml_tensor * ikv4 = ggml_view_4d(ctx0, ik_kv, n_idx_dim, n_kv, 1, ns,
+                            ik_kv->nb[2], ik_kv->nb[3], ik_kv->nb[3], 0);
+                    ggml_tensor * iq4 = ggml_reshape_4d(ctx0, iq, n_idx_dim, Hd, 1, ns);
+                    ggml_tensor * sc  = ggml_mul_mat(ctx0, ikv4, iq4);
+                    ggml_mul_mat_set_prec(sc, GGML_PREC_F32);
+                    sc = ggml_add_inplace(ctx0, sc, msa_mf);
+                    ggml_tensor * bs = ggml_pool_2d(ctx0, sc, GGML_OP_POOL_MAX, blk, 1, blk, 1, 0, 0);
+                    cb(bs, "msa_bs", il);
+
+                    ggml_tensor * bsf = ggml_add(ctx0, bs,
+                            ggml_reshape_4d(ctx0, msa_loc->bias, nblk, 1, 1, ns));
+                    ggml_tensor * idx = ggml_top_k(ctx0, bsf, K);
+
+                    // token idx: tj[t,k,h,s] = blk*idx[k,h,s] + t   (for the mask gather)
+                    // row   idx: tr[t,k,h,s] = tj*HKV + h           (for the per-stream K/V gather)
+                    ggml_tensor * a = ggml_scale(ctx0, ggml_cast(ctx0, idx, GGML_TYPE_F32), (float) blk);
+                    a = ggml_reshape_4d(ctx0, a, 1, K, Hd, ns);
+                    ggml_tensor * tj = ggml_add(ctx0,
+                            ggml_repeat_4d(ctx0, a, blk, K, Hd, ns),
+                            ggml_reshape_3d(ctx0, ggml_arange(ctx0, 0.0f, (float) blk, 1.0f), blk, 1, 1));
+                    ggml_tensor * tr = ggml_add(ctx0,
+                            ggml_scale(ctx0, tj, (float) HKV),
+                            ggml_reshape_3d(ctx0, ggml_arange(ctx0, 0.0f, (float) HKV, 1.0f), 1, 1, Hd));
+
+                    ggml_tensor * tokj = ggml_cast(ctx0, ggml_reshape_2d(ctx0, tj, (int64_t) blk*K*Hd, ns), GGML_TYPE_I32);
+                    ggml_tensor * tokr = ggml_cast(ctx0, ggml_reshape_2d(ctx0, tr, (int64_t) blk*K*Hd, ns), GGML_TYPE_I32);
+
+                    ggml_tensor * k3 = ggml_view_3d(ctx0, k, D, HKV*n_kv, ns, k->nb[1], k->nb[3], 0);
+                    ggml_tensor * v3 = ggml_view_3d(ctx0, v, D, HKV*n_kv, ns, v->nb[1], v->nb[3], 0);
+                    ggml_tensor * m3 = ggml_reshape_3d(ctx0, msa_kqm, 1, n_kv, ns);
+
+                    ggml_tensor * kg = ggml_get_rows(ctx0, k3, tokr);
+                    ggml_tensor * vg = ggml_get_rows(ctx0, v3, tokr);
+                    ggml_tensor * mg = ggml_get_rows(ctx0, m3, tokj);
+
+                    // fold (group, stream) onto the FA channel dim
+                    const ggml_type kt = ggml_is_quantized(k->type) ? GGML_TYPE_F16 : k->type;
+                    const ggml_type vt = ggml_is_quantized(v->type) ? GGML_TYPE_F16 : v->type;
+                    ggml_tensor * kfa = ggml_reshape_4d(ctx0, kg, D, (int64_t) blk*K, 1, Hd*ns);
+                    ggml_tensor * vfa = ggml_reshape_4d(ctx0, vg, D, (int64_t) blk*K, 1, Hd*ns);
+                    if (kfa->type != kt) { kfa = ggml_cast(ctx0, kfa, kt); }
+                    if (vfa->type != vt) { vfa = ggml_cast(ctx0, vfa, vt); }
+                    // the FA mask must be F16
+                    ggml_tensor * mfa = ggml_cast(ctx0, ggml_reshape_4d(ctx0, mg, (int64_t) blk*K, 1, 1, Hd*ns), GGML_TYPE_F16);
+
+                    cur = build_attn_msa_fa(Qcur, kfa, vfa, mfa, Gp, kq_scale, il);
+                } else {
+                    // batch: per-stream loop
+                    std::vector<ggml_tensor *> outs(ns);
+                    for (int64_t st = 0; st < ns; ++st) {
+                        ggml_tensor * iq_s = ggml_view_3d(ctx0, iq, n_idx_dim, Hd, n_tps,
+                                iq->nb[1], iq->nb[2], st*n_tps*iq->nb[2]);
+                        ggml_tensor * ik_s = ggml_view_2d(ctx0, ik_kv, n_idx_dim, n_kv,
+                                ik_kv->nb[2], st*ik_kv->nb[3]);
+                        ggml_tensor * mf_s = ggml_view_3d(ctx0, msa_mf, n_kv, 1, n_tps,
+                                msa_mf->nb[1], msa_mf->nb[1], st*msa_mf->nb[3]);
+                        ggml_tensor * km_s = ggml_view_3d(ctx0, msa_kqm, n_kv, n_tps, 1,
+                                msa_kqm->nb[1], msa_kqm->nb[3], st*msa_kqm->nb[3]);
+                        ggml_tensor * bias_s = ggml_view_2d(ctx0, msa_loc->bias, nblk, n_tps,
+                                msa_loc->bias->nb[1], st*n_tps*msa_loc->bias->nb[1]);
+                        ggml_tensor * q_s = ggml_view_3d(ctx0, Qcur, D, n_head, n_tps,
+                                Qcur->nb[1], Qcur->nb[2], st*n_tps*Qcur->nb[2]);
+                        ggml_tensor * k_s = ggml_view_4d(ctx0, k, D, HKV, n_kv, 1,
+                                k->nb[1], k->nb[2], k->nb[3], st*k->nb[3]);
+                        ggml_tensor * v_s = ggml_view_4d(ctx0, v, D, HKV, n_kv, 1,
+                                v->nb[1], v->nb[2], v->nb[3], st*v->nb[3]);
+
+                        // block scores: bs = maxpool_blk(idx_q * idx_k^T + causal mask)
+                        // scores are unscaled, only the top-k ordering matters
+                        ggml_tensor * sc = ggml_mul_mat(ctx0, ik_s,
+                                ggml_reshape_2d(ctx0, iq_s, n_idx_dim, Hd*n_tps));
+                        // indexer scores run in F32
+                        ggml_mul_mat_set_prec(sc, GGML_PREC_F32);
+                        sc = ggml_reshape_3d(ctx0, sc, n_kv, Hd, n_tps);
+                        sc = ggml_add_inplace(ctx0, sc, mf_s);
+                        ggml_tensor * bs = ggml_pool_2d(ctx0, sc, GGML_OP_POOL_MAX, blk, 1, blk, 1, 0, 0);
+                        cb(bs, "msa_bs", il);
+
+                        // block-level 0/-inf keep mask on the CPU, tiny transfer
+                        ggml_tensor * srcs[2] = { bs, bias_s };
+                        ggml_tensor * bm = ggml_custom_4d(ctx0, GGML_TYPE_F16,
+                                nblk, n_tps, Hd, 1,
+                                srcs, 2, msa_block_mask_op, GGML_N_TASKS_MAX,
+                                const_cast<msa_params *>(&mm.msa_p));
+                        cb(bm, "msa_block_mask", il);
+
+                        // expand block -> token granularity on the GPU (j = bk*blk + t),
+                        // then combine with the causal mask in place
+                        ggml_tensor * bmx = ggml_repeat_4d(ctx0,
+                                ggml_reshape_3d(ctx0, bm, 1, nblk, n_tps*Hd),
+                                blk, nblk, n_tps*Hd, 1);
+                        bmx = ggml_reshape_3d(ctx0, bmx, n_kv, n_tps, Hd);
+                        ggml_tensor * mask4 = ggml_add_inplace(ctx0, bmx, km_s);
+                        mask4 = ggml_reshape_4d(ctx0, mask4, n_kv, n_tps, 1, Hd);
+                        cb(mask4, "msa_mask4", il);
+
+                        // cache views with groups on ne[3];
+                        ggml_tensor * kfa = ggml_permute(ctx0, k_s, 0, 3, 1, 2);
+                        ggml_tensor * vfa = ggml_permute(ctx0, v_s, 0, 3, 1, 2);
+
+                        outs[st] = build_attn_msa_fa(q_s, kfa, vfa, mask4, Gp, kq_scale, il);
+                    }
+                    cur = outs[0];
+                    for (int64_t st = 1; st < ns; ++st) {
+                        cur = ggml_concat(ctx0, cur, outs[st], 1);
+                    }
+                }
+
+                cb(cur, "kqv_out", il);
+                if (model.layers[il].wo) {
+                    cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
+                }
+            }
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        if ((uint32_t) il < hparams.n_layer_dense_lead) {
+            // leading dense FFN (swigluoai)
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SWIGLU_OAI_MOE, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // routed experts (swigluoai MoE)
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SWIGLU_OAI_MOE, hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // shared expert (swigluoai)
+            ggml_tensor * ffn_shexp = build_ffn(cur,
+                    model.layers[il].ffn_up_shexp,   NULL, NULL,
+                    model.layers[il].ffn_gate_shexp, NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SWIGLU_OAI_MOE, LLM_FFN_PAR, il);
+            cb(ffn_shexp, "ffn_shexp", il);
+
+            cur = ggml_add(ctx0, moe_out, ffn_shexp);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/mistral3.cpp

@@ -0,0 +1,235 @@
+#include "models.h"
+
+void llama_model_mistral3::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_ATTENTION_TEMPERATURE_SCALE, hparams.f_attn_temp_scale, false);
+
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_FAST, hparams.yarn_beta_fast,    false);
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_SLOW, hparams.yarn_beta_slow,    false);
+    ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL,   hparams.rope_yarn_log_mul, false);
+
+    hparams.f_attn_temp_offset = 0.0f;
+
+    // TODO: maybe add n_attn_temp_floor_scale as a separate KV?
+    if (hparams.f_attn_temp_scale != 0.0f) {
+        hparams.n_attn_temp_floor_scale = hparams.n_ctx_orig_yarn;
+        if (hparams.n_attn_temp_floor_scale == 0) {
+            throw std::runtime_error("invalid n_ctx_orig_yarn for attention temperature scaling");
+        }
+    }
+
+    switch (hparams.n_layer()) {
+        case 26: type = LLM_TYPE_3B; break;
+        case 34: type = LLM_TYPE_8B; break;
+        case 40: type = LLM_TYPE_14B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_mistral3::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+        else {
+            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+        }
+
+        if (n_expert == 0) {
+            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+            // optional MLP bias
+            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+        } else {
+            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+
+            // For Granite MoE Shared
+            if (hparams.n_ff_shexp > 0) {
+                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, hparams.n_ff_shexp}, 0);
+                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {hparams.n_ff_shexp, n_embd}, 0);
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_mistral3::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_mistral3::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // (optional) temperature tuning
+    ggml_tensor * inp_attn_scale = nullptr;
+    if (hparams.f_attn_temp_scale != 0.0f) {
+        inp_attn_scale = build_inp_attn_scale();
+    }
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // rope freq factors for llama3; may return nullptr for llama2 and other models
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            if (inp_attn_scale) {
+                // apply llama 4 temperature scaling
+                Qcur = ggml_mul(ctx0, Qcur, inp_attn_scale);
+                cb(Qcur, "Qcur_attn_temp_scaled", il);
+            }
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network (non-MoE)
+        if (model.layers[il].ffn_gate_inp == nullptr) {
+
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   model.layers[il].ffn_up_s,
+                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, model.layers[il].ffn_gate_s,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, model.layers[il].ffn_down_s,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        } else {
+            // MoE branch
+            cur = build_norm(ffn_inp,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU, true,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                    il,
+                    nullptr, nullptr,
+                    model.layers[il].ffn_up_exps_s,
+                    model.layers[il].ffn_gate_exps_s,
+                    model.layers[il].ffn_down_exps_s);
+            cb(cur, "ffn_moe_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/mistral4.cpp

@@ -0,0 +1,6 @@
+#include "models.h"
+
+std::unique_ptr<llm_graph_context> llama_model_mistral4::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/modern-bert.cpp

@@ -0,0 +1,174 @@
+#include "models.h"
+
+void llama_model_modern_bert::load_arch_hparams(llama_model_loader & ml) {
+    const bool found_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false);
+    if (found_swa && hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_SYMMETRIC;
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+        uint32_t swa_period = 3;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period, true);
+    } else {
+        hparams.swa_type = LLAMA_SWA_TYPE_NONE;
+    }
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    // Some ModernBert derivatives (e.g. IBM Granite Embedding 97m R2) use
+    // SiLU/SwiGLU in the FFN instead of the default GELU/GeGLU.
+    hparams.llm_ffn_op = LLM_FFN_GEGLU;
+    std::string hidden_act;
+    if (ml.get_key(LLM_KV_HIDDEN_ACT, hidden_act, false)) {
+        hparams.llm_ffn_op = llm_ffn_op_type_from_string(hidden_act, LLM_FFN_GEGLU);
+    }
+
+    switch (hparams.n_layer()) {
+        case 12:
+            type = LLM_TYPE_47M; break; // granite-embedding-small
+        case 22:
+            type = LLM_TYPE_149M; break; // modern-bert-base
+        case 28:
+            type = LLM_TYPE_395M; break; // modern-bert-large
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_modern_bert::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+    tok_norm = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    for(int i = 0; i < n_layer; ++i) {
+        auto& layer = layers[i];
+
+        if ( i != 0 ) {
+            layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        } else{
+            // layer 0 uses identity
+            layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        }
+
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, 3 * n_embd }, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT,   "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, 2 * n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+    }
+
+    cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT,  "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+    cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT,  "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+    cls       = create_tensor(tn(LLM_TENSOR_CLS,      "weight"), {n_embd, n_embd},            TENSOR_NOT_REQUIRED);
+    cls_norm  = create_tensor(tn(LLM_TENSOR_CLS_NORM, "weight"), {n_embd},                    TENSOR_NOT_REQUIRED);
+
+}
+
+std::unique_ptr<llm_graph_context> llama_model_modern_bert::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_modern_bert::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // construct input embeddings (token, type, position)
+    inpL = build_inp_embd(model.tok_embd);
+    cb(inpL, "inp_embd", -1);
+
+    // embed layer norm
+    inpL = build_norm(inpL, model.tok_norm, nullptr, LLM_NORM, 0);
+    cb(inpL, "inp_norm", 0);
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    for (int il = 0; il < n_layer; ++il) {
+        const float freq_base_l  = model.get_rope_freq_base(cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        cur = inpL;
+
+        // attention layer norm
+        if (model.layers[il].attn_norm) {
+            cur = build_norm(inpL,
+                    model.layers[il].attn_norm, NULL,
+                    LLM_NORM, il);
+            cb(cur, "attn_norm", il);
+        }
+
+        // self attention
+        auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                n_embd_head, n_head, n_head_kv, il);
+
+        // RoPE
+        Qcur = ggml_rope_ext(
+                ctx0, Qcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+        Kcur = ggml_rope_ext(
+                ctx0, Kcur, inp_pos, nullptr,
+                n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                ext_factor, attn_factor, beta_fast, beta_slow
+                );
+
+        cb(Qcur, "Qcur", il);
+        cb(Kcur, "Kcur", il);
+        cb(Vcur, "Vcur", il);
+
+        cur = build_attn(inp_attn,
+                    model.layers[il].wo, nullptr, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        cb(cur, "kqv_out", il);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // re-add the layer input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // attention layer norm
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                NULL,                      NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                hparams.llm_ffn_op,
+                LLM_FFN_SEQ, il);
+
+        // attentions bypass the intermediate layer
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM, -1);
+    cb(cur, "final_norm_out", -1);
+
+    res->t_embd = cur;
+    ggml_build_forward_expand(gf, cur);
+}

src/models/mpt.cpp

@@ -0,0 +1,170 @@
+#include "models.h"
+
+void llama_model_mpt::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,  hparams.f_norm_eps);
+    ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV,      hparams.f_clamp_kqv, false);
+    ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias, false);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_7B; break;
+        case 48: type = LLM_TYPE_30B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_mpt::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+    pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,   "weight"), {n_embd, n_ctx_train}, TENSOR_NOT_REQUIRED);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (!output) {
+        output    = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // needs to be on GPU
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wqkv_b = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, TENSOR_NOT_REQUIRED);
+
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, TENSOR_NOT_REQUIRED);
+
+        // FIXME test-llama-archs crashes if q_norm is created
+        layer.attn_q_norm   = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED | TENSOR_SKIP_IF_VIRTUAL);
+        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED | TENSOR_SKIP_IF_VIRTUAL);
+
+        layer.attn_k_norm   = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        // AWQ ScaleActivation layer
+        layer.ffn_act = create_tensor(tn(LLM_TENSOR_FFN_ACT, "scales", i), {n_ff}, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_mpt::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_mpt::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * pos;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    if (model.pos_embd) {
+        // inp_pos - contains the positions
+        ggml_tensor * inp_pos = build_inp_pos();
+        pos                   = ggml_get_rows(ctx0, model.pos_embd, inp_pos);
+        cb(pos, "pos_embd", -1);
+
+        inpL = ggml_add(ctx0, inpL, pos);
+        cb(inpL, "inpL", -1);
+    }
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * attn_norm;
+
+        attn_norm = build_norm(inpL, model.layers[il].attn_norm, model.layers[il].attn_norm_b, LLM_NORM, il);
+        cb(attn_norm, "attn_norm", il);
+
+        // self-attention
+        {
+            cur = attn_norm;
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // Q/K Layernorm
+            if (model.layers[il].attn_q_norm) {
+                Qcur = ggml_reshape_2d(ctx0, Qcur, n_embd_head * n_head, n_tokens);
+                Kcur = ggml_reshape_2d(ctx0, Kcur, n_embd_head * n_head_kv, n_tokens);
+
+                Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, model.layers[il].attn_q_norm_b, LLM_NORM, il);
+
+                Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, model.layers[il].attn_k_norm_b, LLM_NORM, il);
+
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+                Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            }
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+
+        // Add the input
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed forward
+        {
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, model.layers[il].ffn_norm_b, LLM_NORM, il);
+            cb(cur, "ffn_norm", il);
+            cur = build_ffn(cur,
+                model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
+                NULL, NULL, NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                model.layers[il].ffn_act, LLM_FFN_GELU, LLM_FFN_SEQ, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, model.output_norm_b, LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/nanbeige.cpp

@@ -0,0 +1,184 @@
+#include "models.h"
+
+void llama_model_nanbeige::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    uint32_t n_loops_u = 1;
+    ml.get_key(LLM_KV_NUM_LOOPS, n_loops_u, false);
+    GGML_ASSERT(n_loops_u >= 1);
+
+    skip_loop_final_norm = false;
+    ml.get_key(LLM_KV_SKIP_LOOP_FINAL_NORM, skip_loop_final_norm, false);
+
+    n_layer_phys = (int) hparams.n_layer();
+
+    // Bound-check before casting: signed int mul can overflow and bypass the guard.
+    GGML_ASSERT((size_t) n_layer_phys * (size_t) n_loops_u <= (size_t) LLAMA_MAX_LAYERS);
+    n_loops = (int) n_loops_u;
+
+    // Expand logical layer count before load_tensors() allocates layers / KV.
+    if (n_loops > 1) {
+        for (int j = 1; j < n_loops; ++j) {
+            for (int i = 0; i < n_layer_phys; ++i) {
+                const int dst = i + j * n_layer_phys;
+                hparams.n_head_arr[dst]    = hparams.n_head_arr[i];
+                hparams.n_head_kv_arr[dst] = hparams.n_head_kv_arr[i];
+                hparams.n_ff_arr[dst]      = hparams.n_ff_arr[i];
+                hparams.is_swa_impl[dst]   = hparams.is_swa_impl[i];
+                hparams.is_recr_impl[dst]  = hparams.is_recr_impl[i];
+            }
+        }
+        hparams.n_layer_all = (uint32_t) ((size_t) n_layer_phys * (size_t) n_loops);
+    }
+
+    type = LLM_TYPE_UNKNOWN;
+}
+
+void llama_model_nanbeige::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    const int n_phys = n_layer_phys > 0 ? n_layer_phys : n_layer;
+    for (int i = 0; i < n_phys; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2},
+                TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+
+    // Share physical weights across loops; each slot still has its own KV index.
+    if (n_loops > 1) {
+        for (int j = 1; j < n_loops; ++j) {
+            for (int i = 0; i < n_phys; ++i) {
+                layers[i + j * n_phys] = layers[i];
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_nanbeige::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_nanbeige::graph::graph(const llama_model & model, const llm_graph_params & params) :
+        llm_graph_context(params) {
+    const auto & nb = static_cast<const llama_model_nanbeige &>(model);
+
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    const int n_phys  = nb.n_layer_phys > 0 ? nb.n_layer_phys : (int) n_layer;
+    const int n_loops = nb.n_loops > 0 ? nb.n_loops : 1;
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    const float kq_scale = hparams.f_attention_scale == 0.0f
+        ? 1.0f / sqrtf(float(n_embd_head))
+        : hparams.f_attention_scale;
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        {
+            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
+
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow);
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur,   inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   model.layers[il].ffn_up_s,
+                model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, model.layers[il].ffn_gate_s,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, model.layers[il].ffn_down_s,
+                NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        inpL = cur;
+
+        if (n_loops > 1 &&
+            ((il + 1) % n_phys) == 0 &&
+            (il + 1) < n_layer &&
+            !nb.skip_loop_final_norm) {
+            cur = build_norm(inpL, model.output_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "loop_norm", il);
+            inpL = cur;
+        }
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/nemotron-h-moe.cpp

@@ -0,0 +1,6 @@
+#include "models.h"
+
+std::unique_ptr<llm_graph_context> llama_model_nemotron_h_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/nemotron-h.cpp

@@ -0,0 +1,258 @@
+#include "models.h"
+
+void llama_model_nemotron_h::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
+    ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
+    ml.get_key(LLM_KV_SSM_STATE_SIZE,     hparams.ssm_d_state);
+    ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
+    ml.get_key(LLM_KV_SSM_GROUP_COUNT,    hparams.ssm_n_group);
+
+    // A layer is recurrent IFF the n_head_kv value is set to 0 and
+    // the n_ff value is set to 0
+    for (uint32_t i = 0; i < hparams.n_layer(); ++i) {
+        hparams.is_recr_impl[i] = (hparams.n_head_kv(i) == 0 && hparams.n_ff(i) == 0);
+    }
+
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp,        false);
+    ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp,      false);
+    ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,               hparams.n_expert_shared, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,               hparams.expert_weights_norm, false);
+    ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,              hparams.expert_weights_scale, false);
+    ml.get_key(LLM_KV_MOE_LATENT_SIZE,                   hparams.moe_latent_size, false);
+
+    switch (hparams.n_layer()) {
+        case 52: type = LLM_TYPE_31B_A3_5B; break; // Nemotron-H_MOE 31B
+        case 56: type = LLM_TYPE_9B; break;
+        case 88: type = LLM_TYPE_120B_A12B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_nemotron_h::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    // mamba2 Mixer SSM params
+    // NOTE: int64_t for tensor dimensions
+    const int64_t d_conv     = hparams.ssm_d_conv;
+    const int64_t d_inner    = hparams.ssm_d_inner;
+    const int64_t d_state    = hparams.ssm_d_state;
+    const int64_t n_ssm_head = hparams.ssm_dt_rank;
+    const int64_t n_group    = hparams.ssm_n_group;
+    const int64_t d_in_proj  = 2*d_inner + 2*n_group*d_state + n_ssm_head;
+    const int64_t moe_n_embd = hparams.moe_latent_size > 0 ? hparams.moe_latent_size : n_embd;
+
+    // embeddings
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    {
+        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+        output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+        // if output is NULL, init from the input tok embed, duplicated to allow offloading
+        if (output == NULL) {
+            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+        }
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        // all blocks use the attn norm
+        layer.attn_norm  = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        if (hparams.is_recr(i)) {
+            // ssm layers
+            layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, d_in_proj}, 0);
+
+            layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner + 2*n_group*d_state}, 0);
+            layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner + 2*n_group*d_state}, TENSOR_NOT_REQUIRED);
+
+            layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {n_ssm_head}, 0);
+
+            // no "weight" suffix for these
+            layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {1, n_ssm_head}, 0);
+            layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {1, n_ssm_head}, 0);
+
+            layer.ssm_norm = create_tensor(tn(LLM_TENSOR_SSM_NORM, "weight", i), {d_inner / n_group, n_group}, 0);
+
+            // out_proj
+            layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd}, 0);
+        } else if (hparams.n_ff(i) == 0) {
+            // attention layers (with optional bias)
+            const int64_t n_head_i = hparams.n_head(i);
+            const int64_t n_embd_k_gqa_i = hparams.n_embd_k_gqa(i);
+            const int64_t n_embd_v_gqa_i = hparams.n_embd_v_gqa(i);
+            create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head_i, n_embd_k_gqa_i, n_embd_v_gqa_i, 0);
+            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head_i, n_embd}, 0);
+            layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        }  else {
+            if (n_expert != 0) {
+                const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+                const int64_t n_ff_shexp = hparams.n_ff_shexp;
+
+                layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), { n_embd, n_expert}, 0);
+                layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert         }, 0);
+
+                // MoE branch
+                layer.ffn_latent_down = create_tensor(tn(LLM_TENSOR_FFN_LATENT_DOWN, "weight", i), {n_embd, moe_n_embd}, TENSOR_NOT_REQUIRED);
+                layer.ffn_latent_up   = create_tensor(tn(LLM_TENSOR_FFN_LATENT_UP,   "weight", i), {moe_n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+
+                layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   moe_n_embd, n_expert}, 0);
+                layer.ffn_up_exps     = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {moe_n_embd, n_ff_exp, n_expert}, 0);
+
+                // Shared expert branch
+                layer.ffn_down_shexp  = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0);
+                layer.ffn_up_shexp    = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, 0);
+
+            } else {
+                // mlp layers
+                layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  hparams.n_ff(i), n_embd}, 0);
+                layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   hparams.n_ff(i)}, 0);
+                layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+                layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias",   i), {hparams.n_ff(i)}, TENSOR_NOT_REQUIRED);
+            }
+        }
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_nemotron_h::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_nemotron_h::graph::graph(const llama_model & model, const llm_graph_params & params) :
+    llm_build_mamba_base(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+    ggml_build_forward_expand(gf, inpL);
+
+    auto * inp = build_inp_mem_hybrid();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        struct ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        if (hparams.is_recr(il)) {
+            // ssm layer //
+            cur = build_mamba2_layer(inp->get_recr(), cur, model, ubatch, il);
+        } else if (hparams.n_ff(il) == 0) {
+            // attention layer //
+            cur = build_attention_layer(cur, inp->get_attn(), model, n_embd_head, il);
+        } else {
+            cur = build_ffn_layer(cur, model, il);
+        }
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        // add residual
+        cur = ggml_add(ctx0, cur, inpSA);
+        cb(cur, "nemotron_h_block_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+ggml_tensor * llama_model_nemotron_h::graph::build_attention_layer(ggml_tensor *             cur,
+                                                          llm_graph_input_attn_kv * inp_attn,
+                                                          const llama_model &       model,
+                                                                int64_t             n_embd_head,
+                                                                int                 il) {
+    auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur, n_embd_head, hparams.n_head(il), hparams.n_head_kv(il), il);
+
+    const float kq_scale =
+        hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale;
+    cur = build_attn(inp_attn,
+            model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+            Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+    cb(cur, "attn_out", il);
+    return cur;
+}
+
+ggml_tensor * llama_model_nemotron_h::graph::build_ffn_layer(ggml_tensor * cur, const llama_model & model, int il) {
+    if (model.layers[il].ffn_gate_inp == nullptr) {
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   model.layers[il].ffn_up_s,
+                NULL,                      NULL,                        NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, model.layers[il].ffn_down_s,
+                NULL,
+                LLM_FFN_RELU_SQR, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+    } else {
+        ggml_tensor * inp_emb    = cur;
+        ggml_tensor * inp_latent = cur;
+
+        if (model.layers[il].ffn_latent_down) {
+            inp_latent = ggml_mul_mat(ctx0, model.layers[il].ffn_latent_down, cur);
+        }
+
+        ggml_tensor * router_logits = build_lora_mm(model.layers[il].ffn_gate_inp, cur);
+        cb(router_logits, "ffn_moe_logits", il);
+
+        ggml_tensor * moe_out =
+            build_moe_ffn(inp_latent,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    nullptr, // no gate
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_RELU_SQR, hparams.expert_weights_norm,
+                    hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID,
+                    il,
+                    router_logits, nullptr,
+                    model.layers[il].ffn_up_exps_s,
+                    nullptr, // no gate
+                    model.layers[il].ffn_down_exps_s);
+        cb(moe_out, "ffn_moe_out", il);
+
+        if (model.layers[il].ffn_latent_up) {
+            moe_out = ggml_mul_mat(ctx0, model.layers[il].ffn_latent_up, moe_out);
+        }
+
+        ggml_tensor * ffn_shexp = build_ffn(inp_emb,
+                    model.layers[il].ffn_up_shexp,   NULL, model.layers[il].ffn_up_shexp_s,
+                    NULL /* no gate */           ,   NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, model.layers[il].ffn_down_shexp_s,
+                    NULL,
+                    LLM_FFN_RELU_SQR, LLM_FFN_PAR, il);
+        cb(ffn_shexp, "ffn_shexp", il);
+
+        cur = ggml_add(ctx0, moe_out, ffn_shexp);
+        cb(cur, "ffn_out", il);
+    }
+
+    cur = build_cvec(cur, il);
+    cb(cur, "l_out", il);
+
+    return cur;
+}

src/models/nemotron.cpp

@@ -0,0 +1,150 @@
+#include "models.h"
+
+void llama_model_nemotron::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+
+    switch (hparams.n_layer()) {
+        case 32: type = LLM_TYPE_4B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_nemotron::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, 0);
+    output        = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        // optional bias tensors
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}, 0);
+
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+        // optional MLP bias
+        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_nemotron::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_nemotron::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    //GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm,
+                model.layers[il].attn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm,
+                model.layers[il].ffn_norm_b,
+                LLM_NORM, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                NULL,                      NULL,                        NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL,
+                LLM_FFN_RELU_SQR, LLM_FFN_SEQ, il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, model.output_norm_b,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/neo-bert.cpp

@@ -0,0 +1,134 @@
+#include "models.h"
+
+void llama_model_neo_bert::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    if (hparams.n_layer() == 28) {
+        type = LLM_TYPE_250M;
+    }
+}
+
+void llama_model_neo_bert::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0);
+
+    cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+    cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {n_embd},         TENSOR_NOT_REQUIRED);
+
+    cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+    cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+
+    output_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff*2}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_neo_bert::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_neo_bert::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    // construct input embeddings (token, type, position)
+    inpL = build_inp_embd(model.tok_embd);
+    cb(inpL, "inp_embd", -1);
+
+    auto * inp_attn = build_attn_inp_no_cache();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * cur = inpL;
+
+        // pre-norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+
+        {
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            // RoPE
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, nullptr, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+            cb(cur, "kqv_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+            inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+        }
+        // re-add the layer input
+        cur = ggml_add(ctx0, cur, inpL);
+
+        ggml_tensor * ffn_inp = cur;
+        cb(ffn_inp, "ffn_inp", il);
+
+        // pre-norm
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // feed-forward network
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,
+                NULL, NULL, NULL, NULL, NULL,
+                model.layers[il].ffn_down,
+                NULL, NULL, NULL,
+                LLM_FFN_SWIGLU, LLM_FFN_SEQ, il);
+
+        // attentions bypass the intermediate layer
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm_enc, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_embd", -1);
+    res->t_embd = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/nomic-bert-moe.cpp

@@ -0,0 +1,72 @@
+#include "models.h"
+
+void llama_model_nomic_bert_moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,    hparams.f_norm_eps);
+    ml.get_key(LLM_KV_MOE_EVERY_N_LAYERS,         hparams.moe_every_n_layers, 0);
+
+    if (hparams.n_layer() == 12 && hparams.n_embd == 768) {
+        if (arch == LLM_ARCH_NOMIC_BERT) {
+            type = LLM_TYPE_137M;
+        } else if (arch == LLM_ARCH_NOMIC_BERT_MOE && hparams.moe_every_n_layers == 2) {
+            type = LLM_TYPE_475M;
+        }
+    }
+}
+
+void llama_model_nomic_bert_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_token_types == 0) {
+        throw std::runtime_error(arch_name() + " model needs to define token type count");
+    }
+    tok_embd     = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0);
+    type_embd    = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, TENSOR_NOT_REQUIRED);
+
+    if (arch == LLM_ARCH_BERT) {
+        pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,    "weight"), {n_embd, n_ctx_train}, 0);
+
+        cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+        cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {n_embd},         TENSOR_NOT_REQUIRED);
+
+        cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+        cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+    }
+
+    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0);
+    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias",   0), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i),   {n_embd}, 0);
+
+        if (hparams.moe_every_n_layers > 0 && i % hparams.moe_every_n_layers == 1) {
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff,   n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff,   n_embd, n_expert}, 0);
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,   "weight", i), {n_embd, n_expert}, 0);
+        } else {
+            layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+            if (arch == LLM_ARCH_NOMIC_BERT) {
+                layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            }
+        }
+
+        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i),   {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_nomic_bert_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/nomic-bert.cpp

@@ -0,0 +1,72 @@
+#include "models.h"
+
+void llama_model_nomic_bert::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS,    hparams.f_norm_eps);
+    ml.get_key(LLM_KV_MOE_EVERY_N_LAYERS,         hparams.moe_every_n_layers, 0);
+
+    if (hparams.n_layer() == 12 && hparams.n_embd == 768) {
+        if (arch == LLM_ARCH_NOMIC_BERT) {
+            type = LLM_TYPE_137M;
+        } else if (arch == LLM_ARCH_NOMIC_BERT_MOE && hparams.moe_every_n_layers == 2) {
+            type = LLM_TYPE_475M;
+        }
+    }
+}
+
+void llama_model_nomic_bert::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    if (n_token_types == 0) {
+        throw std::runtime_error(arch_name() + " model needs to define token type count");
+    }
+    tok_embd     = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0);
+    type_embd    = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, TENSOR_NOT_REQUIRED);
+
+    if (arch == LLM_ARCH_BERT) {
+        pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,    "weight"), {n_embd, n_ctx_train}, 0);
+
+        cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+        cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {n_embd},         TENSOR_NOT_REQUIRED);
+
+        cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+        cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+    }
+
+    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0);
+    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias",   0), {n_embd}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i),   {n_embd}, 0);
+
+        if (hparams.moe_every_n_layers > 0 && i % hparams.moe_every_n_layers == 1) {
+            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff,   n_expert}, 0);
+            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff,   n_embd, n_expert}, 0);
+            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,   "weight", i), {n_embd, n_expert}, 0);
+        } else {
+            layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, TENSOR_NOT_REQUIRED);
+            layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+            if (arch == LLM_ARCH_NOMIC_BERT) {
+                layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+            }
+        }
+
+        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i),   {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_nomic_bert::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+

src/models/olmo.cpp

@@ -0,0 +1,142 @@
+#include "models.h"
+
+void llama_model_olmo::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+    ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV,     hparams.f_clamp_kqv, false);
+
+    switch (hparams.n_layer()) {
+        case 22: type = LLM_TYPE_1B; break;
+        case 32: type = LLM_TYPE_7B; break;
+        case 80: type = LLM_TYPE_70B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_olmo::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+    // if output is NULL, init from the input tok embed
+    if (output == NULL) {
+        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+    }
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_olmo::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_olmo::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                NULL, NULL,
+                LLM_NORM, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_embd_head, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, nullptr, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_norm(ffn_inp,
+                NULL, NULL,
+                LLM_NORM, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            NULL, NULL,
+            LLM_NORM, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/olmo2.cpp

@@ -0,0 +1,211 @@
+#include "models.h"
+
+void llama_model_olmo2::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    const bool found_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false);
+    if (found_swa && hparams.n_swa > 0) {
+        hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+        uint32_t swa_period = 4;
+        ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+        hparams.set_swa_pattern(swa_period);
+
+        hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+        hparams.rope_freq_scale_train_swa = 1.0; // See olmo2.cpp
+        ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+    } else {
+        hparams.swa_type = LLAMA_SWA_TYPE_NONE;
+    }
+
+    switch (hparams.n_layer()) {
+        case 16: type = LLM_TYPE_1B; break;
+        case 32: type = LLM_TYPE_7B; break;
+        case 40: type = LLM_TYPE_13B; break;
+        case 64: type = LLM_TYPE_32B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_olmo2::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_embd_head = n_embd / n_head;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_head_kv * n_embd_head}, 0);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_olmo2::build_arch_graph(const llm_graph_params & params) const {
+    if (hparams.swa_type == LLAMA_SWA_TYPE_STANDARD) {
+        return std::make_unique<graph<true>>(*this, params);
+    } else {
+        return std::make_unique<graph<false>>(*this, params);
+    }
+}
+
+template <bool iswa>
+llama_model_olmo2::graph<iswa>::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    using inp_attn_type = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa, llm_graph_input_attn_kv>;
+    inp_attn_type * inp_attn = nullptr;
+
+    if constexpr (iswa) {
+        inp_attn = build_attn_inp_kv_iswa();
+    } else {
+        inp_attn = build_attn_inp_kv();
+    }
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        cur = inpL;
+
+        // self_attention
+        {
+            // compute Q and K and RoPE them
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            cb(Qcur, "Qcur", il);
+
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+            cb(Kcur, "Kcur", il);
+
+            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+            cb(Vcur, "Vcur", il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            const bool is_swa = hparams.is_swa(il);
+
+            if (is_swa) {
+                // For sliding window layers, Olmo3 use regular rope with no yarn rope scaling.
+                // This is achieved here by setting freq_scale and attn_factor to 1.
+                // We also set ext_factor to 0 to avoid a few unnecessary computations.
+                Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, 1.0,
+                    0.0, 1.0, beta_fast, beta_slow
+                    );
+
+                Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, 1.0,
+                    0.0, 1.0, beta_fast, beta_slow
+                    );
+            } else {
+                Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+                Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+            }
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        cur = build_ffn(ffn_inp,
+                model.layers[il].ffn_up,   NULL, NULL,
+                model.layers[il].ffn_gate, NULL, NULL,
+                model.layers[il].ffn_down, NULL, NULL,
+                NULL,
+                LLM_FFN_SILU, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, NULL,
+                LLM_NORM_RMS, -1);
+        cb(cur, "ffn_post_norm", -1);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cb(cur, "ffn_out", il);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}
+
+// Explicit template instantiations
+template struct llama_model_olmo2::graph<false>;
+template struct llama_model_olmo2::graph<true>;

src/models/olmoe.cpp

@@ -0,0 +1,174 @@
+#include "models.h"
+
+void llama_model_olmoe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+    switch (hparams.n_layer()) {
+        case 16: type = LLM_TYPE_A1_7B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_olmoe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_embd, n_embd_gqa, n_embd_gqa, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, 0);
+        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+        if (n_expert == 0) {
+            throw std::runtime_error("n_expert must be > 0");
+        }
+        if (n_expert_used == 0) {
+            throw std::runtime_error("n_expert_used must be > 0");
+        }
+
+        // MoE branch
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff,   n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff,   n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_olmoe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_olmoe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self_attention
+        {
+            // compute Q and K and RoPE them
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            cb(Qcur, "Qcur", il);
+
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+            cb(Kcur, "Kcur", il);
+
+            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+            cb(Vcur, "Vcur", il);
+
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL,
+                    LLM_NORM_RMS, il);
+            cb(Kcur, "Kcur_normed", il);
+
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, NULL, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // MoE branch
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        cur = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,
+                model.layers[il].ffn_up_exps,
+                model.layers[il].ffn_gate_exps,
+                model.layers[il].ffn_down_exps,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SILU, false,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
+                il);
+        cb(cur, "ffn_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

src/models/openai-moe.cpp

@@ -0,0 +1,177 @@
+#include "models.h"
+
+void llama_model_openai_moe::load_arch_hparams(llama_model_loader & ml) {
+    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+    ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa);
+
+    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+    uint32_t swa_period = 2;
+    ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false);
+    hparams.set_swa_pattern(swa_period);
+
+    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+    ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false);
+
+    switch (hparams.n_layer()) {
+        case 24: type = LLM_TYPE_20B; break;
+        case 36: type = LLM_TYPE_120B; break;
+        default: type = LLM_TYPE_UNKNOWN;
+    }
+}
+
+void llama_model_openai_moe::load_arch_tensors(llama_model_loader &) {
+    LLAMA_LOAD_LOCALS;
+
+    const int64_t n_ff_exp = hparams.n_ff_exp;
+
+    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+    // output
+    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+    for (int i = 0; i < n_layer; ++i) {
+        auto & layer = layers[i];
+
+        layer.attn_norm      = create_tensor(tn(LLM_TENSOR_ATTN_NORM,      "weight", i), {n_embd}, 0);
+        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+        create_tensor_qkv(layer, i, n_embd, n_head * n_rot, n_head_kv * n_rot, n_head_kv * n_rot, 0);
+        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_rot, n_embd}, 0);
+
+        layer.attn_sinks = create_tensor(tn(LLM_TENSOR_ATTN_SINKS, "weight", i), {n_head}, 0);
+
+        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {  n_embd, n_expert}, 0);
+        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
+
+        layer.ffn_gate_inp_b  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "bias", i), {n_expert}, 0);
+        layer.ffn_gate_exps_b = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "bias", i), {n_ff_exp, n_expert}, 0);
+        layer.ffn_down_exps_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "bias", i), {  n_embd, n_expert}, 0);
+        layer.ffn_up_exps_b   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "bias", i), {n_ff_exp, n_expert}, 0);
+    }
+}
+
+std::unique_ptr<llm_graph_context> llama_model_openai_moe::build_arch_graph(const llm_graph_params & params) const {
+    return std::make_unique<graph>(*this, params);
+}
+
+llama_model_openai_moe::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv_iswa();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        res->t_layer_inp[il] = inpL;
+
+        const float freq_base_l  = model.get_rope_freq_base (cparams, il);
+        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
+
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, nullptr,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            // compute Q and K and RoPE them
+            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
+                    n_rot, n_head, n_head_kv, il);
+
+            Qcur = ggml_rope_ext(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_ext(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            cb(Qcur, "Qcur", il);
+            cb(Kcur, "Kcur", il);
+            cb(Vcur, "Vcur", il);
+
+            cur = build_attn(inp_attn,
+                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,
+                    Qcur, Kcur, Vcur, nullptr, model.layers[il].attn_sinks, nullptr, 1.0f/sqrtf(float(n_rot)), il);
+
+            cb(cur, "attn_out", il);
+        }
+        if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) {
+            // skip computing output for unused tokens
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        cur = ffn_inp;
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, nullptr,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        // MoE branch
+        cur = build_moe_ffn(cur,
+                model.layers[il].ffn_gate_inp,  model.layers[il].ffn_gate_inp_b,
+                model.layers[il].ffn_up_exps,   model.layers[il].ffn_up_exps_b,
+                model.layers[il].ffn_gate_exps, model.layers[il].ffn_gate_exps_b,
+                model.layers[il].ffn_down_exps, model.layers[il].ffn_down_exps_b,
+                nullptr,
+                n_expert, n_expert_used,
+                LLM_FFN_SWIGLU_OAI_MOE, false,
+                hparams.expert_weights_scale,
+                LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX_WEIGHT,
+                il);
+        cb(cur, "ffn_moe_out", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    res->t_h_nextn = cur;
+
+    if (!cparams.embeddings_nextn_masked && inp_out_ids) {
+        cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+    }
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur, model.output_s);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}