Implementing support for laguna arch

src/llama-longhaul.cpp

@@ -113,6 +113,10 @@ void llama_longhaul_cache::release(int layer) {
     }
 }
 
+uint32_t llama_longhaul_cache::capacity() const {
+    return n_slots;
+}
+
 uint32_t llama_longhaul_cache::max_ubatch(uint32_t n_expert_used) const {
     return std::max<uint32_t>(1, n_slots / n_expert_used);
 }