Merge pull request #200 from runpod-workers/release/0.10.0

chore(release): 0.10.0
Merge pull request #207 from JhennerTigreros/main
2025-08-11 15:33:51 -07:00 · 2025-08-09 09:08:19 -07:00 · 2025-08-09 09:06:53 -07:00 · 2025-08-09 09:01:52 -07:00 · 2025-08-07 16:40:06 -05:00 · 2025-08-07 15:59:43 -05:00
10 changed files with 1625 additions and 190 deletions
@@ -0,0 +1,32 @@
+{
+    "tests": [
+        {
+            "name": "basic_inference_test",
+            "input": {
+                "prompt": "Write a short poem about artificial intelligence."
+            },
+            "timeout": 30000
+        }
+    ],
+    "config": {
+        "gpuTypeId": "NVIDIA GeForce RTX 4090",
+        "gpuCount": 1,
+        "env": [
+            {
+                "key": "MODEL_NAME",
+                "value": "facebook/opt-350m"
+            }
+        ],
+        "allowedCudaVersions": [
+            "12.7",
+            "12.6",
+            "12.5",
+            "12.4",
+            "12.3",
+            "12.2",
+            "12.1",
+            "12.0",
+            "11.7"
+        ]
+    }
+}
@@ -12,7 +12,7 @@ RUN --mount=type=cache,target=/root/.cache/pip \
    python3 -m pip install --upgrade -r /requirements.txt

 # Install vLLM (switching back to pip installs since issues that required building fork are fixed and space optimization is not as important since caching) and FlashInfer 
-RUN python3 -m pip install vllm==0.8.2 && \
+RUN python3 -m pip install vllm==0.10.0 && \
    python3 -m pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3

 # Setup for Option 2: Building the Image with the Model included
@@ -47,4 +47,4 @@ RUN --mount=type=secret,id=HF_TOKEN,required=false \
    fi

 # Start the handler
-CMD ["python3", "/src/handler.py"]
+CMD ["python3", "/src/handler.py"]
@@ -18,9 +18,9 @@ Deploy OpenAI-Compatible Blazing-Fast LLM Endpoints powered by the [vLLM](https:
 ### 1. UI for Deploying vLLM Worker on RunPod console:
 ![Demo of Deploying vLLM Worker on RunPod console with new UI](media/ui_demo.gif)

-### 2. Worker vLLM `v2.2.0` with vLLM `0.8.2` now available under `stable` tags 
+### 2. Worker vLLM `v2.7.0` with vLLM `0.9.1` now available under `stable` tags 

-Update v2.2.0 is now available, use the image tag `runpod/worker-v1-vllm:v2.2.0stable-cuda12.1.0`.
+Update v2.7.0 is now available, use the image tag `runpod/worker-v1-vllm:v2.7.0stable-cuda12.1.0`.

 ### 3. OpenAI-Compatible [Embedding Worker](https://github.com/runpod-workers/worker-infinity-embedding) Released
 Deploy your own OpenAI-compatible Serverless Endpoint on RunPod with multiple embedding models and fast inference for RAG and more! 
@@ -38,9 +38,8 @@ Worker vLLM is now cached on all RunPod machines, resulting in near-instant depl
    - [Environment Variables](#environment-variables)
      - [LLM Settings](#llm-settings)
      - [Tokenizer Settings](#tokenizer-settings)
-      - [Tensor Parallelism (Multi-GPU) Settings](#tensor-parallelism-multi-gpu-settings)
-      - [System Settings](#system-settings)
-      - [Streaming Batch Size](#streaming-batch-size)
+      - [System and Parallelism Settings](#system-and-parallelism-settings)
+      - [Streaming Batch Size Settings](#streaming-batch-size-settings)
      - [OpenAI Settings](#openai-settings)
      - [Serverless Settings](#serverless-settings)
  - [Option 2: Build Docker Image with Model Inside](#option-2-build-docker-image-with-model-inside)
@@ -82,7 +81,7 @@ Below is a summary of the available RunPod Worker images, categorized by image s

 | CUDA Version | Stable Image Tag                  | Development Image Tag             | Note                                                        |
 |--------------|-----------------------------------|-----------------------------------|----------------------------------------------------------------------|
-| 12.1.0       | `runpod/worker-v1-vllm:v2.2.0stable-cuda12.1.0` | `runpod/worker-v1-vllm:v2.2.0dev-cuda12.1.0` | When creating an Endpoint, select CUDA Version 12.3, 12.2 and 12.1 in the filter. |
+| 12.1.0       | `runpod/worker-v1-vllm:v2.7.0stable-cuda12.1.0` | `runpod/worker-v1-vllm:v2.7.0dev-cuda12.1.0` | When creating an Endpoint, select CUDA Version 12.3, 12.2 and 12.1 in the filter. |



@@ -91,9 +90,10 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 #### Prerequisites
 - RunPod Account

-#### Environment Variables/Settings
+#### Environment Variables
 > Note:  `0` is equivalent to `False` and `1` is equivalent to `True` for boolean as int values.

+#### LLM Settings
 | `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
 |-------------------------------------------|-----------------------|--------------------------------------------|---------------|
 | `MODEL_NAME`                                   | 'facebook/opt-125m'   | `str`                                      | Name or path of the Hugging Face model to use. |
@@ -125,7 +125,7 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 | `MAX_NUM_SEQS`                            | 256                   | `int`                                      | Maximum number of sequences per iteration. |
 | `MAX_LOGPROBS`                            | 20                    | `int`                                      | Max number of log probs to return when logprobs is specified in SamplingParams. |
 | `DISABLE_LOG_STATS`                       | False                 | `bool`                                     | Disable logging statistics. |
-| `QUANTIZATION`                            | None                  | ['awq', 'squeezellm', 'gptq']              | Method used to quantize the weights. |
+| `QUANTIZATION`                            | None                  | ['awq', 'squeezellm', 'gptq', 'bitsandbytes']              | Method used to quantize the weights. |
 | `ROPE_SCALING`                            | None                  | `dict`                                     | RoPE scaling configuration in JSON format. |
 | `ROPE_THETA`                              | None                  | `float`                                    | RoPE theta. Use with rope_scaling. |
 | `TOKENIZER_POOL_SIZE`                     | 0                     | `int`                                      | Size of tokenizer pool to use for asynchronous tokenization. |
@@ -139,6 +139,7 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 | `LONG_LORA_SCALING_FACTORS`               | None                  | `tuple`                                    | Specify multiple scaling factors for LoRA adapters. |
 | `MAX_CPU_LORAS`                           | None                  | `int`                                      | Maximum number of LoRAs to store in CPU memory. |
 | `FULLY_SHARDED_LORAS`                     | False                 | `bool`                                     | Enable fully sharded LoRA layers. |
+| `LORA_MODULES`| `[]`| `list[dict]`| Add lora adapters from Hugging Face `[{"name": "xx", "path": "xxx/xxxx", "base_model_name": "xxx/xxxx"}`|
 | `SCHEDULER_DELAY_FACTOR`                  | 0.0                   | `float`                                    | Apply a delay before scheduling next prompt. |
 | `ENABLE_CHUNKED_PREFILL`                  | False                 | `bool`                                     | Enable chunked prefill requests. |
 | `SPECULATIVE_MODEL`                       | None                  | `str`                                      | The name of the draft model to be used in speculative decoding. |
@@ -157,11 +158,20 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 | `PREEMPTION_CPU_CAPACITY`                 | 2                     | `float`                                    | The percentage of CPU memory used for the saved activations. |
 | `DISABLE_LOGGING_REQUEST`                 | False                 | `bool`                                     | Disable logging requests. |
 | `MAX_LOG_LEN`                             | None                  | `int`                                      | Max number of prompt characters or prompt ID numbers being printed in log. |
-**Tokenizer Settings**
+
+
+#### Tokenizer Settings
+
+| `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
+|-------------------------------------------|-----------------------|--------------------------------------------|---------------|
 | `TOKENIZER_NAME`                    | `None`               | `str`                                         |Tokenizer repository to use a different tokenizer than the model's default. |
 | `TOKENIZER_REVISION`                | `None`               | `str`                                         |Tokenizer revision to load. |
 | `CUSTOM_CHAT_TEMPLATE`              | `None`               | `str` of single-line jinja template                                         |Custom chat jinja template. [More Info](https://huggingface.co/docs/transformers/chat_templating) |
-**System, GPU, and Tensor Parallelism(Multi-GPU) Settings**
+
+#### System and Parallelism Settings
+
+| `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
+|-------------------------------------------|-----------------------|--------------------------------------------|---------------|
 | `GPU_MEMORY_UTILIZATION`            | `0.95`               | `float`                                         |Sets GPU VRAM utilization. |
 | `MAX_PARALLEL_LOADING_WORKERS`      | `None`               | `int`                                         |Load model sequentially in multiple batches, to avoid RAM OOM when using tensor parallel and large models. |
 | `BLOCK_SIZE`                        | `16`                 | `8`, `16`, `32`                           |Token block size for contiguous chunks of tokens. |
@@ -169,16 +179,31 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 | `ENFORCE_EAGER`                     | False                  | `bool`                                         |Always use eager-mode PyTorch. If False(`0`), will use eager mode and CUDA graph in hybrid for maximal performance and flexibility. |
 | `MAX_SEQ_LEN_TO_CAPTURE`        | `8192`               | `int`                                     |Maximum context length covered by CUDA graphs. When a sequence has context length larger than this, we fall back to eager mode.|
 | `DISABLE_CUSTOM_ALL_REDUCE`         | `0`                  | `int`                                         |Enables or disables custom all reduce. |
-**Streaming Batch Size Settings**:  
+
+
+#### Streaming Batch Size Settings
+
+The way this works is that the first request will have a batch size of `DEFAULT_MIN_BATCH_SIZE`, and each subsequent request will have a batch size of `previous_batch_size * DEFAULT_BATCH_SIZE_GROWTH_FACTOR`. This will continue until the batch size reaches `DEFAULT_BATCH_SIZE`. E.g. for the default values, the batch sizes will be `1, 3, 9, 27, 50, 50, 50, ...`. You can also specify this per request, with inputs `max_batch_size`, `min_batch_size`, and `batch_size_growth_factor`. This has nothing to do with vLLM's internal batching, but rather the number of tokens sent in each HTTP request from the worker
+
+
+| `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
+|-------------------------------------------|-----------------------|--------------------------------------------|---------------|
 | `DEFAULT_BATCH_SIZE`                | `50`                 | `int`                                         |Default and Maximum batch size for token streaming to reduce HTTP calls. |
 | `DEFAULT_MIN_BATCH_SIZE`            | `1`                  | `int`                                         |Batch size for the first request, which will be multiplied by the growth factor every subsequent request. |
 | `DEFAULT_BATCH_SIZE_GROWTH_FACTOR`  | `3`                  | `float`                                         |Growth factor for dynamic batch size. |
-The way this works is that the first request will have a batch size of `DEFAULT_MIN_BATCH_SIZE`, and each subsequent request will have a batch size of `previous_batch_size * DEFAULT_BATCH_SIZE_GROWTH_FACTOR`. This will continue until the batch size reaches `DEFAULT_BATCH_SIZE`. E.g. for the default values, the batch sizes will be `1, 3, 9, 27, 50, 50, 50, ...`. You can also specify this per request, with inputs `max_batch_size`, `min_batch_size`, and `batch_size_growth_factor`. This has nothing to do with vLLM's internal batching, but rather the number of tokens sent in each HTTP request from the worker |
-**OpenAI Settings**
+
+#### OpenAI Settings
+
+| `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
+|-------------------------------------------|-----------------------|--------------------------------------------|---------------|
 | `RAW_OPENAI_OUTPUT`                 | `1`                  | boolean as `int`                                         |Enables raw OpenAI SSE format string output when streaming.  **Required** to be enabled (which it is by default) for OpenAI compatibility. |
 | `OPENAI_SERVED_MODEL_NAME_OVERRIDE` | `None`               | `str`                                         |Overrides the name of the served model from model repo/path to specified name, which you will then be able to use the value for the `model` parameter when making OpenAI requests |
 | `OPENAI_RESPONSE_ROLE`              | `assistant`          | `str`                       |Role of the LLM's Response in OpenAI Chat Completions. |
-**Serverless Settings**
+
+#### Serverless Settings
+
+| `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
+|-------------------------------------------|-----------------------|--------------------------------------------|---------------|
 | `MAX_CONCURRENCY`                   | `300`                | `int`                                         |Max concurrent requests per worker. vLLM has an internal queue, so you don't have to worry about limiting by VRAM, this is for improving scaling/load balancing efficiency |
 | `DISABLE_LOG_STATS`                 | False                  | `bool`                                         |Enables or disables vLLM stats logging. |
 | `DISABLE_LOG_REQUESTS`              | False                  | `bool`                                         |Enables or disables vLLM request logging. |
@@ -4,8 +4,11 @@ pyarrow
 runpod~=1.7.7
 huggingface-hub
 packaging
-typing-extensions==4.7.1
+typing-extensions>=4.8.0
 pydantic
 pydantic-settings
 hf-transfer
-transformers
+transformers>=4.55.0
+bitsandbytes>=0.45.0
+kernels
+torch==2.6.0
@@ -7,7 +7,7 @@ variable "REPOSITORY" {
 }

 variable "BASE_IMAGE_VERSION" {
-  default = "v2.0.0stable"
+  default = "v2.8.0stable"
 }

 group "all" {
@@ -29,4 +29,4 @@ target "worker-1210" {
    WORKER_CUDA_VERSION = "12.1.0"
  }
  output = ["type=docker,push=${PUSH}"]
-}
+}
@@ -25,15 +25,68 @@ class vLLMEngine:
        load_dotenv() # For local development
        self.engine_args = get_engine_args()
        logging.info(f"Engine args: {self.engine_args}")
-        self.tokenizer = TokenizerWrapper(self.engine_args.tokenizer or self.engine_args.model, 
-                                          self.engine_args.tokenizer_revision, 
-                                          self.engine_args.trust_remote_code)
+        
+        # Initialize vLLM engine first
        self.llm = self._initialize_llm() if engine is None else engine.llm
+        
+        # Only create custom tokenizer wrapper if not using mistral tokenizer mode
+        # For mistral models, let vLLM handle tokenizer initialization
+        if self.engine_args.tokenizer_mode != 'mistral':
+            self.tokenizer = TokenizerWrapper(self.engine_args.tokenizer or self.engine_args.model, 
+                                              self.engine_args.tokenizer_revision, 
+                                              self.engine_args.trust_remote_code)
+        else:
+            # For mistral models, we'll get the tokenizer from vLLM later
+            self.tokenizer = None
+            
        self.max_concurrency = int(os.getenv("MAX_CONCURRENCY", DEFAULT_MAX_CONCURRENCY))
        self.default_batch_size = int(os.getenv("DEFAULT_BATCH_SIZE", DEFAULT_BATCH_SIZE))
        self.batch_size_growth_factor = int(os.getenv("BATCH_SIZE_GROWTH_FACTOR", DEFAULT_BATCH_SIZE_GROWTH_FACTOR))
        self.min_batch_size = int(os.getenv("MIN_BATCH_SIZE", DEFAULT_MIN_BATCH_SIZE))

+    def _get_tokenizer_for_chat_template(self):
+        """Get tokenizer for chat template application"""
+        if self.tokenizer is not None:
+            return self.tokenizer
+        else:
+            # For mistral models, get tokenizer from vLLM engine
+            # This is a fallback - ideally chat templates should be handled by vLLM directly
+            try:
+                from transformers import AutoTokenizer
+                tokenizer = AutoTokenizer.from_pretrained(
+                    self.engine_args.tokenizer or self.engine_args.model,
+                    revision=self.engine_args.tokenizer_revision or "main",
+                    trust_remote_code=self.engine_args.trust_remote_code
+                )
+                # Create a minimal wrapper
+                class MinimalTokenizerWrapper:
+                    def __init__(self, tokenizer):
+                        self.tokenizer = tokenizer
+                        self.custom_chat_template = os.getenv("CUSTOM_CHAT_TEMPLATE")
+                        self.has_chat_template = bool(self.tokenizer.chat_template) or bool(self.custom_chat_template)
+                        if self.custom_chat_template and isinstance(self.custom_chat_template, str):
+                            self.tokenizer.chat_template = self.custom_chat_template
+                    
+                    def apply_chat_template(self, input):
+                        if isinstance(input, list):
+                            if not self.has_chat_template:
+                                raise ValueError(
+                                    "Chat template does not exist for this model, you must provide a single string input instead of a list of messages"
+                                )
+                        elif isinstance(input, str):
+                            input = [{"role": "user", "content": input}]
+                        else:
+                            raise ValueError("Input must be a string or a list of messages")
+                        
+                        return self.tokenizer.apply_chat_template(
+                            input, tokenize=False, add_generation_prompt=True
+                        )
+                
+                return MinimalTokenizerWrapper(tokenizer)
+            except Exception as e:
+                logging.error(f"Failed to create fallback tokenizer: {e}")
+                raise e
+
    def dynamic_batch_size(self, current_batch_size, batch_size_growth_factor):
        return min(current_batch_size*batch_size_growth_factor, self.default_batch_size)
                           
@@ -55,7 +108,8 @@ class vLLMEngine:

    async def _generate_vllm(self, llm_input, validated_sampling_params, batch_size, stream, apply_chat_template, request_id, batch_size_growth_factor, min_batch_size: str) -> AsyncGenerator[dict, None]:
        if apply_chat_template or isinstance(llm_input, list):
-            llm_input = self.tokenizer.apply_chat_template(llm_input)
+            tokenizer_wrapper = self._get_tokenizer_for_chat_template()
+            llm_input = tokenizer_wrapper.apply_chat_template(llm_input)
        results_generator = self.llm.generate(llm_input, validated_sampling_params, request_id)
        n_responses, n_input_tokens, is_first_output = validated_sampling_params.n, 0, True
        last_output_texts, token_counters = ["" for _ in range(n_responses)], {"batch": 0, "total": 0}
@@ -122,41 +176,55 @@ class OpenAIvLLMEngine(vLLMEngine):
        super().__init__(vllm_engine)
        self.served_model_name = os.getenv("OPENAI_SERVED_MODEL_NAME_OVERRIDE") or self.engine_args.model
        self.response_role = os.getenv("OPENAI_RESPONSE_ROLE") or "assistant"
+        self.lora_adapters = self._load_lora_adapters()
        asyncio.run(self._initialize_engines())
        self.raw_openai_output = bool(int(os.getenv("RAW_OPENAI_OUTPUT", 1)))
-        
+
+    def _load_lora_adapters(self):
+        adapters = []
+        try:
+            adapters = json.loads(os.getenv("LORA_MODULES", '[]'))
+        except Exception as e:
+            logging.info(f"---Initialized adapter json load error: {e}")
+
+        for i, adapter in enumerate(adapters):
+            try:
+                adapters[i] = LoRAModulePath(**adapter)
+                logging.info(f"---Initialized adapter: {adapter}")
+            except Exception as e:
+                logging.info(f"---Initialized adapter not worked: {e}")
+                continue
+        return adapters
+
    async def _initialize_engines(self):
        self.model_config = await self.llm.get_model_config()
        self.base_model_paths = [
            BaseModelPath(name=self.engine_args.model, model_path=self.engine_args.model)
        ]

-        lora_modules = os.getenv('LORA_MODULES', None)
-        if lora_modules is not None:
-            try:
-                lora_modules = json.loads(lora_modules)
-                lora_modules = [LoRAModulePath(**lora_modules)]
-            except:
-                lora_modules = None
-
        self.serving_models = OpenAIServingModels(
            engine_client=self.llm,
            model_config=self.model_config,
            base_model_paths=self.base_model_paths,
-            lora_modules=None,
-            prompt_adapters=None,
+            lora_modules=self.lora_adapters,
        )
-
+        await self.serving_models.init_static_loras()
+        
+        # Get chat template from vLLM tokenizer if available
+        chat_template = None
+        if self.tokenizer and hasattr(self.tokenizer, 'tokenizer'):
+            chat_template = self.tokenizer.tokenizer.chat_template
+        
        self.chat_engine = OpenAIServingChat(
            engine_client=self.llm, 
            model_config=self.model_config,
            models=self.serving_models,
            response_role=self.response_role,
            request_logger=None,
-            chat_template=self.tokenizer.tokenizer.chat_template,
+            chat_template=chat_template,
            chat_template_content_format="auto",
            # enable_reasoning=os.getenv('ENABLE_REASONING', 'false').lower() == 'true',
-            # reasoning_parser=None,
+            reasoning_parser= os.getenv('REASONING_PARSER', "") or None,
            # return_token_as_token_ids=False,
            enable_auto_tools=os.getenv('ENABLE_AUTO_TOOL_CHOICE', 'false').lower() == 'true',
            tool_parser=os.getenv('TOOL_CALL_PARSER', "") or None,
@@ -26,6 +26,7 @@ DEFAULT_ARGS = {
    "trust_remote_code": os.getenv('TRUST_REMOTE_CODE', 'False').lower() == 'true',
    "download_dir": os.getenv('DOWNLOAD_DIR', None),
    "load_format": os.getenv('LOAD_FORMAT', 'auto'),
+    "config_format": os.getenv('CONFIG_FORMAT', 'auto'),
    "dtype": os.getenv('DTYPE', 'auto'),
    "kv_cache_dtype": os.getenv('KV_CACHE_DTYPE', 'auto'),
    "quantization_param_path": os.getenv('QUANTIZATION_PARAM_PATH', None),
@@ -92,6 +93,9 @@ DEFAULT_ARGS = {
    "otlp_traces_endpoint": os.getenv('OTLP_TRACES_ENDPOINT', None),
    "use_v2_block_manager": os.getenv('USE_V2_BLOCK_MANAGER', 'true'),
 }
+limit_mm_env = os.getenv('LIMIT_MM_PER_PROMPT')
+if limit_mm_env is not None:
+    DEFAULT_ARGS["limit_mm_per_prompt"] = convert_limit_mm_per_prompt(limit_mm_env)

 def match_vllm_args(args):
    """Rename args to match vllm by:
@@ -122,7 +126,7 @@ def get_local_args():
        local_args = json.load(f)

    if local_args.get("MODEL_NAME") is None:
-        raise ValueError("Model name not found in /local_model_args.json. There was a problem when baking the model in.")
+        logging.warning("Model name not found in /local_model_args.json. There maybe was a problem when baking the model in.")

    logging.info(f"Using baked in model with args: {local_args}")
    os.environ["TRANSFORMERS_OFFLINE"] = "1"
@@ -147,6 +151,9 @@ def get_engine_args():
    
    # Rename and match to vllm args
    args = match_vllm_args(args)
+
+    if args.get("load_format") == "bitsandbytes":
+        args["quantization"] = args["load_format"]
    
    # Set tensor parallel size and max parallel loading workers if more than 1 GPU is available
    num_gpus = device_count()
@@ -15,9 +15,14 @@ except ImportError:

 logging.basicConfig(level=logging.INFO)

+# Updated to parse multiple comma-separated multimodal limits (e.g., 'image=1,video=0')
 def convert_limit_mm_per_prompt(input_string: str):
-    key, value = input_string.split('=')
-    return {key: int(value)}
+    result = {}
+    pairs = input_string.split(',')
+    for pair in pairs:
+        key, value = pair.split('=')
+        result[key] = int(value)
+    return result

 def count_physical_cores():
    with open('/proc/cpuinfo') as f:
@@ -1,5 +1,432 @@
 {
  "versions": {
+    "0.10.0": {
+      "imageName": "runpod/worker-v1-vllm:v2.8.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
+    "0.9.0": {
+      "imageName": "runpod/worker-v1-vllm:v2.6.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
+    "0.9.1": {
+      "imageName": "runpod/worker-v1-vllm:v2.7.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
+    "0.9.0": {
+      "imageName": "runpod/worker-v1-vllm:v2.6.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
+    "0.8.5": {
+      "imageName": "runpod/worker-v1-vllm:v2.5.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
+    "0.8.4": {
+      "imageName": "runpod/worker-v1-vllm:v2.4.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
+    "0.8.3": {
+      "imageName": "runpod/worker-v1-vllm:v2.3.0stable-cuda12.1.0",
+      "minimumCudaVersion": "12.1",
+      "categories": [
+        {
+          "title": "LLM Settings",
+          "settings": [
+            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+            "DISABLE_SLIDING_WINDOW", "NUM_LOOKAHEAD_SLOTS",
+            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST",
+            "ENABLE_AUTO_TOOL_CHOICE", "TOOL_CALL_PARSER"
+          ]
+        },
+        {
+          "title": "Tokenizer Settings",
+          "settings": [
+            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+          ]
+        },
+        {
+          "title": "System Settings",
+          "settings": [
+            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+          ]
+        },
+        {
+          "title": "Streaming Settings",
+          "settings": [
+            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+          ]
+        },
+        {
+          "title": "OpenAI Settings",
+          "settings": [
+            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+          ]
+        },
+        {
+          "title": "Serverless Settings",
+          "settings": [
+            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+          ]
+        }
+      ]
+    },
    "0.8.2": {
      "imageName": "runpod/worker-v1-vllm:v2.2.0stable-cuda12.1.0",
      "minimumCudaVersion": "12.1",
@@ -365,126 +792,6 @@
          ]
        }
      ]
-    },
-    "0.6.2": {
-      "imageName": "runpod/worker-v1-vllm:v1.5.0stable-cuda12.1.0",
-      "minimumCudaVersion": "12.1",
-      "categories": [
-        {
-          "title": "LLM Settings",
-          "settings": [
-            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
-            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
-            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
-            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
-            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
-            "DISABLE_SLIDING_WINDOW", "USE_V2_BLOCK_MANAGER", "NUM_LOOKAHEAD_SLOTS",
-            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
-            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
-            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
-            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
-            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
-            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
-            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
-            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
-            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
-            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
-            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
-            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST"
-          ]
-        },
-        {
-          "title": "Tokenizer Settings",
-          "settings": [
-            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
-          ]
-        },
-        {
-          "title": "System Settings",
-          "settings": [
-            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
-            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
-          ]
-        },
-        {
-          "title": "Streaming Settings",
-          "settings": [
-            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
-          ]
-        },
-        {
-          "title": "OpenAI Settings",
-          "settings": [
-            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
-          ]
-        },
-        {
-          "title": "Serverless Settings",
-          "settings": [
-            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
-          ]
-        }
-      ]
-    },
-    "0.6.1": {
-      "imageName": "runpod/worker-v1-vllm:v1.4.0stable-cuda12.1.0",
-      "minimumCudaVersion": "12.1",
-      "categories": [
-        {
-          "title": "LLM Settings",
-          "settings": [
-            "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
-            "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
-            "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
-            "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
-            "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
-            "DISABLE_SLIDING_WINDOW", "USE_V2_BLOCK_MANAGER", "NUM_LOOKAHEAD_SLOTS",
-            "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
-            "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
-            "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
-            "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
-            "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
-            "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
-            "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
-            "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
-            "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
-            "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
-            "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
-            "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST"
-          ]
-        },
-        {
-          "title": "Tokenizer Settings",
-          "settings": [
-            "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
-          ]
-        },
-        {
-          "title": "System Settings",
-          "settings": [
-            "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
-            "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
-          ]
-        },
-        {
-          "title": "Streaming Settings",
-          "settings": [
-            "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
-          ]
-        },
-        {
-          "title": "OpenAI Settings",
-          "settings": [
-            "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
-          ]
-        },
-        {
-          "title": "Serverless Settings",
-          "settings": [
-            "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
-          ]
-        }
-      ]
    }
  },
  "schema": {
@@ -741,14 +1048,15 @@
    "env_var_name": "QUANTIZATION",
    "value": "",
    "title": "Quantization",
-    "description": "Method used to quantize the weights.",
+    "description": "Method used to quantize the weights.\nif the `Load Format` is 'bitsandbytes' then `Quantization` will be forced to 'bitsandbytes'",
    "required": false,
    "type": "select",
    "options": [
      { "value": "None", "label": "None" },
      { "value": "awq", "label": "AWQ" },
      { "value": "squeezellm", "label": "SqueezeLLM" },
-      { "value": "gptq", "label": "GPTQ" }
+      { "value": "gptq", "label": "GPTQ" },
+      { "value": "bitsandbytes", "label": "bitsandbytes" }
    ]
  },
  "ROPE_SCALING": {
Author	SHA1	Message	Date
Marut PandyaandGitHub	7aa17463d3	Merge pull request #200 from runpod-workers/release/0.10.0 chore(release): 0.10.0	2025-08-11 15:33:51 -07:00
Marut PandyaandGitHub	72a643dd0c	Merge pull request #207 from JhennerTigreros/main Update requirements and engine creation to support new 0.10.0 vLLM version	2025-08-09 09:08:19 -07:00
Marut PandyaandGitHub	15f569f970	Merge pull request #208 from runpod-workers/revert-202-feat/proper-deployment [Revert]"feat: added dev & release workflows; added conventions to support AI IDE"	2025-08-09 09:06:53 -07:00
Marut PandyaandGitHub	2f2bd4c749	Revert "feat: added dev & release workflows; added conventions to support AI IDE"	2025-08-09 09:01:52 -07:00
Jhenner Tigreros	fb0c030797	fix initialization on openaiservingmodels	2025-08-07 16:40:06 -05:00
Jhenner Tigreros	8b02a703b4	fix issues	2025-08-07 15:59:43 -05:00
Jhenner Tigreros	d8863139d6	add model to test	2025-08-07 15:28:02 -05:00
Jhenner TigrerosandGitHub	f5a063956e	Fix requirements.txt to support gpt-oss models	2025-08-07 15:13:37 -05:00
Marut PandyaandGitHub	18748fd73e	Merge pull request #202 from runpod-workers/feat/proper-deployment feat: added dev & release workflows; added conventions to support AI IDE	2025-08-04 17:13:47 -07:00
Tim Pietrusky	0133c23be8	ci: added manual workflow trigger for releases	2025-08-04 09:58:23 +02:00
Tim Pietrusky	a129cff47d	docs: use "version" instead of actual version, so that people can check the releases	2025-07-31 12:03:17 +02:00
Tim Pietrusky	30f2c4630e	refactor: use correct version	2025-07-31 12:02:46 +02:00
Tim Pietrusky	b98636e432	feat: added "dev" and "release" workflows; removed "vllm-base-image" as it's not needed	2025-07-28 16:40:20 +02:00
pandyamarut	185205c750	chore(release): 0.10.0 Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-07-25 10:18:06 -07:00
pandyamarut	b948e530a1	chore(release): 0.10.0 Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-07-25 10:17:50 -07:00
Marut PandyaandGitHub	4f22d6f107	Merge pull request #193 from runpod-workers/release/0.9.1 chore(release): v0.9.1	2025-06-26 13:23:33 -07:00
pandyamarut	8839689132	chore(release): v0.9.1 Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-06-26 11:06:23 -07:00
Marut PandyaandGitHub	5ddc2326cd	Merge pull request #191 from runpod-workers/feat/0.9.1 feat: update to 0.9.1 & added CONFIG_FORMAT to run magistral	2025-06-26 10:17:54 -07:00
Tim Pietrusky	46a3300cbb	chore: reverted changeds to only focus on vllm update	2025-06-20 14:56:39 +02:00
Tim Pietrusky	1e9a731380	Fix Mistral tokenizer initialization: let vLLM handle tokenizer for mistral models	2025-06-14 15:31:56 +02:00
Tim Pietrusky	c47e649a24	Add CONFIG_FORMAT environment variable support	2025-06-14 15:12:46 +02:00
Tim Pietrusky	7192bcaeef	Trigger build automatically on feat/0.9.1 branch	2025-06-14 13:49:22 +02:00
Tim Pietrusky	8665ffb78d	Revert workflow back to original configuration	2025-06-14 13:47:16 +02:00
Tim Pietrusky	57431b30ad	Fix workflow: use standard GitHub runners and actions	2025-06-14 13:42:46 +02:00
Tim Pietrusky	437a84c77a	ci: added workflow to build the image	2025-06-14 13:31:36 +02:00
Tim Pietrusky	a4062fc488	feat: update to 0.9.1	2025-06-14 13:31:27 +02:00
Marut PandyaandGitHub	9631407c1d	Merge pull request #189 from runpod-workers/hf-mm add multi modal env var	2025-06-10 17:03:18 -07:00
pandyamarut	1a93932ab2	add multi modal env var Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-06-10 17:02:42 -07:00
Marut PandyaandGitHub	4c6c88c3b9	Merge pull request #188 from WorstDev01/add-mm-limit-parameter Add multimodal limit parameter support	2025-06-10 14:59:21 -07:00
WorstDev01	a58a76783d	Add multimodal limit parameter support - Updated convert_limit_mm_per_prompt to handle multiple types - Added limit_mm_per_prompt parameter for image and video limits Note: Consider adjusting default values - perhaps image limit > 1 or video=1	2025-06-10 23:36:47 +02:00
Marut PandyaandGitHub	26919c8849	Merge pull request #185 from runpod-workers/up-0.9.0 Version upgrade	2025-06-05 12:00:05 -07:00
pandyamarut	70cd1c8113	update vllm version 0.9.0 Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-06-05 11:59:15 -07:00
Marut PandyaandGitHub	deeff579f8	Merge pull request #183 from SorenDreano/fix/model_name_in_local_args remove requirements for MODEL_NAME in local_model_args.json	2025-06-04 12:45:00 -07:00
Soren Dreano	11f96a09d7	remove requirements for MODEL_NAME in local_model_args.json We want to use the same local_model_args.json for multiple models which have different names. It would be very convenient to only have a single local_args file and not have to create it every time A warning should be enough for users	2025-05-23 17:59:19 +02:00
Marut PandyaandGitHub	23e8ecf85b	Merge pull request #169 from RedHitMark/main fix lora and multi-lora	2025-05-14 16:25:26 -07:00
Marut PandyaandGitHub	6db2c44d3b	Update Dockerfile	2025-05-09 09:27:54 -07:00
Marut PandyaandGitHub	9b7ca4d0b0	Update tests.json	2025-05-08 17:12:59 -07:00
Marut PandyaandGitHub	2a4eaf0356	Merge pull request #182 from runpod-workers/up-0.8.5 update vllm	2025-05-08 11:40:23 -07:00
pandyamarut	ba19cc97bf	update vllm Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-05-08 11:34:58 -07:00
Marut PandyaandGitHub	6075f2c590	Merge pull request #181 from runpod-workers/revert-177-main Revert "fix: added back limit_mm_per_prompt to engine args"	2025-05-07 12:31:15 -07:00
Marut PandyaandGitHub	a9786a2481	Revert "fix: added back limit_mm_per_prompt to engine args"	2025-05-07 12:29:33 -07:00
Marut PandyaandGitHub	0b6bc7a2be	Update tests.json	2025-05-07 11:14:50 -07:00
Marut PandyaandGitHub	d0ab58ee17	Merge pull request #180 from muhsinking/patch-1 Update README.md table to fix table of contents links	2025-05-03 21:02:37 -07:00
Marut PandyaandGitHub	4e474c41c8	Merge pull request #177 from aleksandar-babic/main fix: added back limit_mm_per_prompt to engine args	2025-05-03 21:01:58 -07:00
Mo KingandGitHub	e9d13c155e	Update README.md to fix table of contents links to environment variable sections Separates environment variables into multiple tables, so that the table of contents will correctly jump to the appropriate section when clicked.	2025-04-30 12:45:54 -04:00
Marut PandyaandGitHub	e807342e90	Merge pull request #178 from runpod-workers/up-0.8.4 update vllm	2025-04-21 13:15:03 -07:00
pandyamarut	f33e8d2bcd	update vllm Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-04-21 13:14:32 -07:00
Marut PandyaandGitHub	6b64bb93bd	Merge pull request #147 from mohamednaji7/BitsAndBytes completing the "bitsandbytes" option - based on https://docs.vllm.ai/en/stable/quantization/bnb.html	2025-04-21 11:39:53 -07:00
Aleksandar Babic	a53cf777ff	chore: added limit_mm_per_prompt to worker config	2025-04-21 10:14:05 -04:00
Aleksandar Babic	cfc258674b	chore: added trailing comma to the final arg	2025-04-21 10:06:08 -04:00
Aleksandar Babic	8beafed06b	fix: added back limit_mm_per_prompt to engine args	2025-04-21 10:02:46 -04:00
Marut PandyaandGitHub	d77c53c3b7	Merge pull request #175 from runpod-workers/up-0.8.3 update vllm	2025-04-07 11:58:03 -07:00
pandyamarut	3d067cd472	update vllm Signed-off-by: pandyamarut <pandyamarut@gmail.com>	2025-04-07 11:54:57 -07:00
Marut PandyaandGitHub	f1360ccae7	Merge pull request #173 from KAJdev/patch-1 Update hub.json	2025-04-03 14:52:06 -07:00
Marut PandyaandGitHub	96f1b86126	Merge pull request #174 from KAJdev/patch-2 Update tests.json	2025-04-03 10:14:49 -07:00
Ezekiel WotringandGitHub	3a095f3e10	Update tests.json	2025-04-03 09:10:01 -08:00
Ezekiel WotringandGitHub	7cf7f3e4e6	Update hub.json	2025-04-03 09:04:28 -08:00
Marut PandyaandGitHub	4395d0c67b	Update hub.json	2025-04-03 08:10:28 -07:00
Marut PandyaandGitHub	5864fa6843	Update hub.json	2025-03-31 15:55:35 -07:00
Marut PandyaandGitHub	067f0bc173	Update tests.json	2025-03-31 15:49:31 -07:00
Marut PandyaandGitHub	0869068e7c	Create tests.json	2025-03-31 15:32:06 -07:00
Marut PandyaandGitHub	19f25b17de	Update hub.json	2025-03-31 15:14:40 -07:00
Marut PandyaandGitHub	70aa748c30	Update hub.json	2025-03-31 15:14:24 -07:00
RedHitMark	084d000324	fix lora and multi-lora	2025-03-14 21:45:09 +01:00
Mohamed NagyandGitHub	04288240f6	updating the `.` in ['awq', 'squeezellm', 'gptq'. 'bitsandbytes'] for the QUNATIZATION row	2025-02-02 14:58:48 +02:00
Mohamed NagyandGitHub	9299f43b8b	solving "typing_extensions" compatibility with "bitsandbytes" ``` 2025-01-22 18:04:01 [INFO] > [stage-0 5/8] RUN --mount=type=cache,target=/root/.cache/pip python3 -m pip install --upgrade pip && python3 -m pip install --upgrade -r /requirements.txt: 2025-01-22 18:04:01 [INFO] #13 8.904 2025-01-22 18:04:01 [INFO] #13 8.904 The conflict is caused by: 2025-01-22 18:04:01 [INFO] #13 8.904 The user requested typing-extensions==4.7.1 2025-01-22 18:04:01 [INFO] #13 8.904 bitsandbytes 0.45.0 depends on typing_extensions>=4.8.0 ```	2025-01-22 18:06:46 +02:00
mohamednaji7	131c17569f	correct access to "args" dictionary	2025-01-21 22:33:28 +02:00
mohamednaji7	8882f6d50b	adding 'bitsandbytes' to `QUANTIZATION`	2025-01-21 14:46:40 +02:00
mohamednaji7	331bc30101	adding 'bitsandbytes' option	2025-01-21 14:08:37 +02:00
mohamednaji7	a27f72a33a	inforce args.quantization for bnb load_froamt	2025-01-21 14:01:32 +02:00
mohamednaji7	7167985f23	including bitsandbytes "src:https://docs.vllm.ai/en/stable/quantization/bnb.html "	2025-01-21 13:49:36 +02:00