merge main

docs/source/index.rst

+Welcome to vLLM!
+================
+
+.. figure:: ./assets/logos/vllm-logo-text-light.png
+  :width: 60%
+  :align: center
+  :alt: vLLM
+  :class: no-scaled-link
+
+.. raw:: html
+
+   <p style="text-align:center">
+   <strong>Easy, fast, and cheap LLM serving for everyone
+   </strong>
+   </p>
+
+   <p style="text-align:center">
+   <script async defer src="https://buttons.github.io/buttons.js"></script>
+   <a class="github-button" href="https://github.com/vllm-project/vllm" data-show-count="true" data-size="large" aria-label="Star">Star</a>
+   <a class="github-button" href="https://github.com/vllm-project/vllm/subscription" data-icon="octicon-eye" data-size="large" aria-label="Watch">Watch</a>
+   <a class="github-button" href="https://github.com/vllm-project/vllm/fork" data-icon="octicon-repo-forked" data-size="large" aria-label="Fork">Fork</a>
+   </p>
+
+
+
+vLLM is a fast and easy-to-use library for LLM inference and serving.
+
+vLLM is fast with:
+
+* State-of-the-art serving throughput
+* Efficient management of attention key and value memory with **PagedAttention**
+* Continuous batching of incoming requests
+* Fast model execution with CUDA/HIP graph
+* Quantization: `GPTQ <https://arxiv.org/abs/2210.17323>`_, `AWQ <https://arxiv.org/abs/2306.00978>`_, `SqueezeLLM <https://arxiv.org/abs/2306.07629>`_
+* Optimized CUDA kernels
+
+vLLM is flexible and easy to use with:
+
+* Seamless integration with popular HuggingFace models
+* High-throughput serving with various decoding algorithms, including *parallel sampling*, *beam search*, and more
+* Tensor parallelism support for distributed inference
+* Streaming outputs
+* OpenAI-compatible API server
+* Support NVIDIA GPUs and AMD GPUs
+
+For more information, check out the following:
+
+* `vLLM announcing blog post <https://vllm.ai>`_ (intro to PagedAttention)
+* `vLLM paper <https://arxiv.org/abs/2309.06180>`_ (SOSP 2023)
+* `How continuous batching enables 23x throughput in LLM inference while reducing p50 latency <https://www.anyscale.com/blog/continuous-batching-llm-inference>`_ by Cade Daniel et al.
+
+
+
+Documentation
+-------------
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Getting Started
+
+   getting_started/installation
+   getting_started/amd-installation
+   getting_started/quickstart
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Serving
+
+   serving/distributed_serving
+   serving/run_on_sky
+   serving/deploying_with_triton
+   serving/deploying_with_docker
+   serving/serving_with_langchain
+   serving/metrics
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Models
+
+   models/supported_models
+   models/adding_model
+   models/engine_args
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Quantization
+
+   quantization/auto_awq
\ No newline at end of file

docs/source/models/adding_model.rst

+.. _adding_a_new_model:
+
+Adding a New Model
+==================
+
+This document provides a high-level guide on integrating a `HuggingFace Transformers <https://github.com/huggingface/transformers>`_ model into vLLM.
+
+.. note::
+    The complexity of adding a new model depends heavily on the model's architecture.
+    The process is considerably straightforward if the model shares a similar architecture with an existing model in vLLM.
+    However, for models that include new operators (e.g., a new attention mechanism), the process can be a bit more complex.
+
+.. tip::
+    If you are encountering issues while integrating your model into vLLM, feel free to open an issue on our `GitHub <https://github.com/vllm-project/vllm/issues>`_ repository.
+    We will be happy to help you out!
+
+
+0. Fork the vLLM repository
+--------------------------------
+
+Start by forking our `GitHub`_ repository and then :ref:`build it from source <build_from_source>`.
+This gives you the ability to modify the codebase and test your model.
+
+
+1. Bring your model code
+------------------------
+
+Clone the PyTorch model code from the HuggingFace Transformers repository and put it into the `vllm/model_executor/models <https://github.com/vllm-project/vllm/tree/main/vllm/model_executor/models>`_ directory.
+For instance, vLLM's `OPT model <https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/models/opt.py>`_ was adapted from the HuggingFace's `modeling_opt.py <https://github.com/huggingface/transformers/blob/main/src/transformers/models/opt/modeling_opt.py>`_ file.
+
+.. warning::
+    When copying the model code, make sure to review and adhere to the code's copyright and licensing terms.
+
+
+2. Rewrite the :code:`forward` methods
+--------------------------------------
+
+Next, you need to rewrite the :code:`forward` methods of your model by following these steps:
+
+1. Remove any unnecessary code, such as the code only used for training.
+2. Change the input parameters:
+
+.. code-block:: diff
+
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+    -    attention_mask: Optional[torch.Tensor] = None,
+    -    position_ids: Optional[torch.LongTensor] = None,
+    -    past_key_values: Optional[List[torch.FloatTensor]] = None,
+    -    inputs_embeds: Optional[torch.FloatTensor] = None,
+    -    labels: Optional[torch.LongTensor] = None,
+    -    use_cache: Optional[bool] = None,
+    -    output_attentions: Optional[bool] = None,
+    -    output_hidden_states: Optional[bool] = None,
+    -    return_dict: Optional[bool] = None,
+    -) -> Union[Tuple, CausalLMOutputWithPast]:
+    +    positions: torch.Tensor,
+    +    kv_caches: List[KVCache],
+    +    input_metadata: InputMetadata,
+    +    cache_events: Optional[List[torch.cuda.Event]],
+    +) -> SamplerOutput:
+
+3. Update the code by considering that :code:`input_ids` and :code:`positions` are now flattened tensors.
+4. Replace the attention operation with either :code:`PagedAttention`, :code:`PagedAttentionWithRoPE`, or :code:`PagedAttentionWithALiBi` depending on the model's architecture.
+
+.. note::
+    Currently, vLLM supports the basic multi-head attention mechanism and its variant with rotary positional embeddings.
+    If your model employs a different attention mechanism, you will need to implement a new attention layer in vLLM.
+
+
+3. (Optional) Implement tensor parallelism and quantization support
+-------------------------------------------------------------------
+
+If your model is too large to fit into a single GPU, you can use tensor parallelism to manage it.
+To do this, substitute your model's linear and embedding layers with their tensor-parallel versions.
+For the embedding layer, you can simply replace :code:`nn.Embedding` with :code:`VocabParallelEmbedding`. For the output LM head, you can use :code:`ParallelLMHead`.
+When it comes to the linear layers, we provide the following options to parallelize them:
+
+* :code:`ReplicatedLinear`: Replicates the inputs and weights across multiple GPUs. No memory saving.
+* :code:`RowParallelLinear`: The input tensor is partitioned along the hidden dimension. The weight matrix is partitioned along the rows (input dimension). An *all-reduce* operation is performed after the matrix multiplication to reduce the results. Typically used for the second FFN layer and the output linear transformation of the attention layer.
+* :code:`ColumnParallelLinear`: The input tensor is replicated. The weight matrix is partitioned along the columns (output dimension). The result is partitioned along the column dimension. Typically used for the first FFN layer and the separated QKV transformation of the attention layer in the original Transformer.
+* :code:`MergedColumnParallelLinear`: Column-parallel linear that merges multiple `ColumnParallelLinear` operators. Typically used for the first FFN layer with weighted activation functions (e.g., SiLU). This class handles the sharded weight loading logic of multiple weight matrices.
+* :code:`QKVParallelLinear`: Parallel linear layer for the query, key, and value projections of the multi-head and grouped-query attention mechanisms. When number of key/value heads are less than the world size, this class replicates the key/value heads properly. This class handles the weight loading and replication of the weight matrices.
+
+Note that all the linear layers above take `linear_method` as an input. vLLM will set this parameter according to different quantization schemes to support weight quantization.
+
+4. Implement the weight loading logic
+-------------------------------------
+
+You now need to implement the :code:`load_weights` method in your :code:`*ForCausalLM` class.
+This method should load the weights from the HuggingFace's checkpoint file and assign them to the corresponding layers in your model. Specifically, for `MergedColumnParallelLinear` and `QKVParallelLinear` layers, if the original model has separated weight matrices, you need to load the different parts separately.
+
+5. Register your model
+----------------------
+
+Finally, include your :code:`*ForCausalLM` class in `vllm/model_executor/models/__init__.py <https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/models/__init__.py>`_ and register it to the :code:`_MODEL_REGISTRY` in `vllm/model_executor/model_loader.py <https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/model_loader.py>`_.

docs/source/models/engine_args.rst

+.. _engine_args:
+
+Engine Arguments
+================
+
+Below, you can find an explanation of every engine argument for vLLM:
+
+.. option:: --model <model_name_or_path>
+
+    Name or path of the huggingface model to use.
+
+.. option:: --tokenizer <tokenizer_name_or_path>
+
+    Name or path of the huggingface tokenizer to use.
+
+.. option:: --revision <revision>
+
+    The specific model version to use. It can be a branch name, a tag name, or a commit id. If unspecified, will use the default version.
+
+.. option:: --tokenizer-revision <revision>
+
+    The specific tokenizer version to use. It can be a branch name, a tag name, or a commit id. If unspecified, will use the default version.
+
+.. option:: --tokenizer-mode {auto,slow}
+
+    The tokenizer mode.
+    
+    * "auto" will use the fast tokenizer if available.
+    * "slow" will always use the slow tokenizer.
+
+.. option:: --trust-remote-code
+
+    Trust remote code from huggingface.
+
+.. option:: --download-dir <directory>
+
+    Directory to download and load the weights, default to the default cache dir of huggingface.
+
+.. option:: --load-format {auto,pt,safetensors,npcache,dummy}
+
+    The format of the model weights to load.
+
+    * "auto" will try to load the weights in the safetensors format and fall back to the pytorch bin format if safetensors format is not available.
+    * "pt" will load the weights in the pytorch bin format.
+    * "safetensors" will load the weights in the safetensors format.
+    * "npcache" will load the weights in pytorch format and store a numpy cache to speed up the loading.
+    * "dummy" will initialize the weights with random values, mainly for profiling.
+
+.. option:: --dtype {auto,half,float16,bfloat16,float,float32}
+
+    Data type for model weights and activations.
+
+    * "auto" will use FP16 precision for FP32 and FP16 models, and BF16 precision for BF16 models.
+    * "half" for FP16. Recommended for AWQ quantization.
+    * "float16" is the same as "half".
+    * "bfloat16" for a balance between precision and range.
+    * "float" is shorthand for FP32 precision.
+    * "float32" for FP32 precision.
+
+.. option:: --max-model-len <length>
+
+    Model context length. If unspecified, will be automatically derived from the model config.
+
+.. option:: --worker-use-ray
+
+    Use Ray for distributed serving, will be automatically set when using more than 1 GPU.
+
+.. option:: --pipeline-parallel-size (-pp) <size>
+
+    Number of pipeline stages.
+
+.. option:: --tensor-parallel-size (-tp) <size>
+
+    Number of tensor parallel replicas.
+
+.. option:: --max-parallel-loading-workers <workers>
+
+    Load model sequentially in multiple batches, to avoid RAM OOM when using tensor parallel and large models.
+
+.. option:: --block-size {8,16,32}
+
+    Token block size for contiguous chunks of tokens.
+
+.. option:: --seed <seed>
+
+    Random seed for operations.
+
+.. option:: --swap-space <size>
+
+    CPU swap space size (GiB) per GPU.
+
+.. option:: --gpu-memory-utilization <fraction>
+
+    The fraction of GPU memory to be used for the model executor, which can range from 0 to 1. 
+    For example, a value of 0.5 would imply 50% GPU memory utilization.
+    If unspecified, will use the default value of 0.9.
+
+.. option:: --max-num-batched-tokens <tokens>
+
+    Maximum number of batched tokens per iteration.
+
+.. option:: --max-num-seqs <sequences>
+
+    Maximum number of sequences per iteration.
+
+.. option:: --max-paddings <paddings>
+
+    Maximum number of paddings in a batch.
+
+.. option:: --disable-log-stats
+
+    Disable logging statistics.
+
+.. option:: --quantization (-q) {awq,squeezellm,None}
+
+    Method used to quantize the weights.

docs/source/models/supported_models.rst

+.. _supported_models:
+
+Supported Models
+================
+
+vLLM supports a variety of generative Transformer models in `HuggingFace Transformers <https://huggingface.co/models>`_.
+The following is the list of model architectures that are currently supported by vLLM.
+Alongside each architecture, we include some popular models that use it.
+
+.. list-table::
+  :widths: 25 25 50
+  :header-rows: 1
+
+  * - Architecture
+    - Models
+    - Example HuggingFace Models
+  * - :code:`AquilaForCausalLM`
+    - Aquila
+    - :code:`BAAI/Aquila-7B`, :code:`BAAI/AquilaChat-7B`, etc.
+  * - :code:`BaiChuanForCausalLM`
+    - Baichuan
+    - :code:`baichuan-inc/Baichuan2-13B-Chat`, :code:`baichuan-inc/Baichuan-7B`, etc.
+  * - :code:`ChatGLMModel`
+    - ChatGLM
+    - :code:`THUDM/chatglm2-6b`, :code:`THUDM/chatglm3-6b`, etc.
+  * - :code:`DeciLMForCausalLM`
+    - DeciLM
+    - :code:`Deci/DeciLM-7B`, :code:`Deci/DeciLM-7B-instruct`, etc.
+  * - :code:`BloomForCausalLM`
+    - BLOOM, BLOOMZ, BLOOMChat
+    - :code:`bigscience/bloom`, :code:`bigscience/bloomz`, etc.
+  * - :code:`FalconForCausalLM`
+    - Falcon
+    - :code:`tiiuae/falcon-7b`, :code:`tiiuae/falcon-40b`, :code:`tiiuae/falcon-rw-7b`, etc.
+  * - :code:`GPT2LMHeadModel`
+    - GPT-2
+    - :code:`gpt2`, :code:`gpt2-xl`, etc.
+  * - :code:`GPTBigCodeForCausalLM`
+    - StarCoder, SantaCoder, WizardCoder
+    - :code:`bigcode/starcoder`, :code:`bigcode/gpt_bigcode-santacoder`, :code:`WizardLM/WizardCoder-15B-V1.0`, etc.
+  * - :code:`GPTJForCausalLM`
+    - GPT-J
+    - :code:`EleutherAI/gpt-j-6b`, :code:`nomic-ai/gpt4all-j`, etc.
+  * - :code:`GPTNeoXForCausalLM`
+    - GPT-NeoX, Pythia, OpenAssistant, Dolly V2, StableLM
+    - :code:`EleutherAI/gpt-neox-20b`, :code:`EleutherAI/pythia-12b`, :code:`OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5`, :code:`databricks/dolly-v2-12b`, :code:`stabilityai/stablelm-tuned-alpha-7b`, etc.
+  * - :code:`InternLMForCausalLM`
+    - InternLM
+    - :code:`internlm/internlm-7b`, :code:`internlm/internlm-chat-7b`, etc.
+  * - :code:`LlamaForCausalLM`
+    - LLaMA, LLaMA-2, Vicuna, Alpaca, Koala, Guanaco
+    - :code:`meta-llama/Llama-2-13b-hf`, :code:`meta-llama/Llama-2-70b-hf`, :code:`openlm-research/open_llama_13b`, :code:`lmsys/vicuna-13b-v1.3`, :code:`young-geng/koala`, etc.
+  * - :code:`MistralForCausalLM`
+    - Mistral, Mistral-Instruct
+    - :code:`mistralai/Mistral-7B-v0.1`, :code:`mistralai/Mistral-7B-Instruct-v0.1`, etc.
+  * - :code:`MixtralForCausalLM`
+    - Mixtral-8x7B, Mixtral-8x7B-Instruct
+    - :code:`mistralai/Mixtral-8x7B-v0.1`, :code:`mistralai/Mixtral-8x7B-Instruct-v0.1`, etc.
+  * - :code:`MPTForCausalLM`
+    - MPT, MPT-Instruct, MPT-Chat, MPT-StoryWriter
+    - :code:`mosaicml/mpt-7b`, :code:`mosaicml/mpt-7b-storywriter`, :code:`mosaicml/mpt-30b`, etc.
+  * - :code:`OPTForCausalLM`
+    - OPT, OPT-IML
+    - :code:`facebook/opt-66b`, :code:`facebook/opt-iml-max-30b`, etc.
+  * - :code:`PhiForCausalLM`
+    - Phi
+    - :code:`microsoft/phi-1_5`, :code:`microsoft/phi-2`, etc.
+  * - :code:`QWenLMHeadModel`
+    - Qwen
+    - :code:`Qwen/Qwen-7B`, :code:`Qwen/Qwen-7B-Chat`, etc.
+  * - :code:`YiForCausalLM`
+    - Yi
+    - :code:`01-ai/Yi-6B`, :code:`01-ai/Yi-34B`, etc.
+
+If your model uses one of the above model architectures, you can seamlessly run your model with vLLM.
+Otherwise, please refer to :ref:`Adding a New Model <adding_a_new_model>` for instructions on how to implement support for your model.
+Alternatively, you can raise an issue on our `GitHub <https://github.com/vllm-project/vllm/issues>`_ project.
+
+.. note::
+    Currently, the ROCm version of vLLM supports Mistral and Mixtral only for context lengths up to 4096.
+
+.. tip::
+    The easiest way to check if your model is supported is to run the program below:
+
+    .. code-block:: python
+
+        from vllm import LLM
+
+        llm = LLM(model=...)  # Name or path of your model
+        output = llm.generate("Hello, my name is")
+        print(output)
+
+    If vLLM successfully generates text, it indicates that your model is supported.
+
+.. tip::
+    To use models from `ModelScope <https://www.modelscope.cn>`_ instead of HuggingFace Hub, set an environment variable:
+
+    .. code-block:: shell
+
+       $ export VLLM_USE_MODELSCOPE=True
+
+    And use with :code:`trust_remote_code=True`.
+
+    .. code-block:: python
+
+        from vllm import LLM
+
+        llm = LLM(model=..., revision=..., trust_remote_code=True)  # Name or path of your model
+        output = llm.generate("Hello, my name is")
+        print(output)

docs/source/quantization/auto_awq.rst

+.. _auto_awq:
+
+AutoAWQ
+==================
+
+.. warning::
+
+   Please note that AWQ support in vLLM is under-optimized at the moment. We would recommend using the unquantized version of the model for better
+   accuracy and higher throughput. Currently, you can use AWQ as a way to reduce memory footprint. As of now, it is more suitable for low latency
+   inference with small number of concurrent requests. vLLM's AWQ implementation have lower throughput than unquantized version.
+
+To create a new 4-bit quantized model, you can leverage `AutoAWQ <https://github.com/casper-hansen/AutoAWQ>`_. 
+Quantizing reduces the model's precision from FP16 to INT4 which effectively reduces the file size by ~70%.
+The main benefits are lower latency and memory usage.
+
+You can quantize your own models by installing AutoAWQ or picking one of the `400+ models on Huggingface <https://huggingface.co/models?sort=trending&search=awq>`_. 
+
+.. code-block:: console
+
+    $ pip install autoawq
+
+After installing AutoAWQ, you are ready to quantize a model. Here is an example of how to quantize Vicuna 7B v1.5:
+
+.. code-block:: python
+
+    from awq import AutoAWQForCausalLM
+    from transformers import AutoTokenizer
+
+    model_path = 'lmsys/vicuna-7b-v1.5'
+    quant_path = 'vicuna-7b-v1.5-awq'
+    quant_config = { "zero_point": True, "q_group_size": 128, "w_bit": 4, "version": "GEMM" }
+
+    # Load model
+    model = AutoAWQForCausalLM.from_pretrained(model_path, **{"low_cpu_mem_usage": True})
+    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+
+    # Quantize
+    model.quantize(tokenizer, quant_config=quant_config)
+
+    # Save quantized model
+    model.save_quantized(quant_path)
+    tokenizer.save_pretrained(quant_path)
+
+To run an AWQ model with vLLM, you can use `TheBloke/Llama-2-7b-Chat-AWQ <https://huggingface.co/TheBloke/Llama-2-7b-Chat-AWQ>`_ with the following command:
+
+.. code-block:: console
+
+    $ python examples/llm_engine_example.py --model TheBloke/Llama-2-7b-Chat-AWQ --quantization awq
+
+AWQ models are also supported directly through the LLM entrypoint:
+
+.. code-block:: python
+
+    from vllm import LLM, SamplingParams
+
+    # Sample prompts.
+    prompts = [
+        "Hello, my name is",
+        "The president of the United States is",
+        "The capital of France is",
+        "The future of AI is",
+    ]
+    # Create a sampling params object.
+    sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
+
+    # Create an LLM.
+    llm = LLM(model="TheBloke/Llama-2-7b-Chat-AWQ", quantization="AWQ")
+    # Generate texts from the prompts. The output is a list of RequestOutput objects
+    # that contain the prompt, generated text, and other information.
+    outputs = llm.generate(prompts, sampling_params)
+    # Print the outputs.
+    for output in outputs:
+        prompt = output.prompt
+        generated_text = output.outputs[0].text
+        print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")

docs/source/serving/deploying_with_docker.rst

+.. _deploying_with_docker:
+
+Deploying with Docker
+============================
+
+vLLM offers official docker image for deployment.
+The image can be used to run OpenAI compatible server.
+The image is available on Docker Hub as `vllm/vllm-openai <https://hub.docker.com/r/vllm/vllm-openai/tags>`_.
+
+.. code-block:: console
+
+    $ docker run --runtime nvidia --gpus all \
+        -v ~/.cache/huggingface:/root/.cache/huggingface \
+        --env "HUGGING_FACE_HUB_TOKEN=<secret>" \
+        -p 8000:8000 \
+        --ipc=host \
+        vllm/vllm-openai:latest \
+        --model mistralai/Mistral-7B-v0.1
+
+
+.. note::
+
+        You can either use the ``ipc=host`` flag or ``--shm-size`` flag to allow the
+        container to access the host's shared memory. vLLM uses PyTorch, which uses shared
+        memory to share data between processes under the hood, particularly for tensor parallel inference.
+
+
+You can build and run vLLM from source via the provided dockerfile. To build vLLM:
+
+.. code-block:: console
+
+    $ DOCKER_BUILDKIT=1 docker build . --target vllm-openai --tag vllm/vllm-openai # optionally specifies: --build-arg max_jobs=8 --build-arg nvcc_threads=2
+
+
+.. note::
+
+        By default vLLM will build for all GPU types for widest distribution. If you are just building for the
+        current GPU type the machine is running on, you can add the argument ``--build-arg torch_cuda_arch_list=""``
+        for vLLM to find the current GPU type and build for that.
+
+
+To run vLLM:
+
+.. code-block:: console
+
+    $ docker run --runtime nvidia --gpus all \
+        -v ~/.cache/huggingface:/root/.cache/huggingface \
+        -p 8000:8000 \
+        --env "HUGGING_FACE_HUB_TOKEN=<secret>" \
+        vllm/vllm-openai <args...>
+

docs/source/serving/deploying_with_triton.rst

+.. _deploying_with_triton:
+
+Deploying with NVIDIA Triton
+============================
+
+The `Triton Inference Server <https://github.com/triton-inference-server>`_ hosts a tutorial demonstrating how to quickly deploy a simple `facebook/opt-125m <https://huggingface.co/facebook/opt-125m>`_ model using vLLM. Please see `Deploying a vLLM model in Triton <https://github.com/triton-inference-server/tutorials/blob/main/Quick_Deploy/vLLM/README.md#deploying-a-vllm-model-in-triton>`_ for more details.

docs/source/serving/distributed_serving.rst

+.. _distributed_serving:
+
+Distributed Inference and Serving
+=================================
+
+vLLM supports distributed tensor-parallel inference and serving. Currently, we support `Megatron-LM's tensor parallel algorithm <https://arxiv.org/pdf/1909.08053.pdf>`_. We manage the distributed runtime with `Ray <https://github.com/ray-project/ray>`_. To run distributed inference, install Ray with:
+
+.. code-block:: console
+
+    $ pip install ray
+
+To run multi-GPU inference with the :code:`LLM` class, set the :code:`tensor_parallel_size` argument to the number of GPUs you want to use. For example, to run inference on 4 GPUs:
+
+.. code-block:: python
+
+    from vllm import LLM
+    llm = LLM("facebook/opt-13b", tensor_parallel_size=4)
+    output = llm.generate("San Franciso is a")
+
+To run multi-GPU serving, pass in the :code:`--tensor-parallel-size` argument when starting the server. For example, to run API server on 4 GPUs:
+
+.. code-block:: console
+
+    $ python -m vllm.entrypoints.api_server \
+    $     --model facebook/opt-13b \
+    $     --tensor-parallel-size 4
+
+To scale vLLM beyond a single machine, start a `Ray runtime <https://docs.ray.io/en/latest/ray-core/starting-ray.html>`_ via CLI before running vLLM:
+
+.. code-block:: console
+
+    $ # On head node
+    $ ray start --head
+
+    $ # On worker nodes
+    $ ray start --address=<ray-head-address>
+
+After that, you can run inference and serving on multiple machines by launching the vLLM process on the head node by setting :code:`tensor_parallel_size` to the number of GPUs to be the total number of GPUs across all machines.
\ No newline at end of file

docs/source/serving/metrics.rst

+Production Metrics
+==================
+
+vLLM exposes a number of metrics that can be used to monitor the health of the
+system. These metrics are exposed via the `/metrics` endpoint on the vLLM
+OpenAI compatible API server.
+
+The following metrics are exposed:
+
+.. literalinclude:: ../../../vllm/engine/metrics.py
+    :language: python
+    :start-after: begin-metrics-definitions
+    :end-before: end-metrics-definitions

docs/source/serving/run_on_sky.rst

+.. _on_cloud:
+
+Running on clouds with SkyPilot
+===============================
+
+.. raw:: html
+
+    <p align="center">
+        <img src="https://imgur.com/yxtzPEu.png" alt="vLLM"/>
+    </p>
+
+vLLM can be run on the cloud to scale to multiple GPUs with `SkyPilot <https://github.com/skypilot-org/skypilot>`__, an open-source framework for running LLMs on any cloud.
+
+To install SkyPilot and setup your cloud credentials, run:
+
+.. code-block:: console
+
+    $ pip install skypilot
+    $ sky check
+
+See the vLLM SkyPilot YAML for serving, `serving.yaml <https://github.com/skypilot-org/skypilot/blob/master/llm/vllm/serve.yaml>`__.
+
+.. code-block:: yaml
+
+    resources:
+        accelerators: A100
+
+    envs:
+        MODEL_NAME: decapoda-research/llama-13b-hf
+        TOKENIZER: hf-internal-testing/llama-tokenizer
+
+    setup: |
+        conda create -n vllm python=3.9 -y
+        conda activate vllm
+        git clone https://github.com/vllm-project/vllm.git
+        cd vllm
+        pip install .
+        pip install gradio
+
+    run: |
+        conda activate vllm
+        echo 'Starting vllm api server...'
+        python -u -m vllm.entrypoints.api_server \
+                        --model $MODEL_NAME \
+                        --tensor-parallel-size $SKYPILOT_NUM_GPUS_PER_NODE \
+                        --tokenizer $TOKENIZER 2>&1 | tee api_server.log &
+        echo 'Waiting for vllm api server to start...'
+        while ! `cat api_server.log | grep -q 'Uvicorn running on'`; do sleep 1; done
+        echo 'Starting gradio server...'
+        python vllm/examples/gradio_webserver.py
+
+Start the serving the LLaMA-13B model on an A100 GPU:
+
+.. code-block:: console
+
+    $ sky launch serving.yaml
+
+Check the output of the command. There will be a shareable gradio link (like the last line of the following). Open it in your browser to use the LLaMA model to do the text completion.
+
+.. code-block:: console
+
+    (task, pid=7431) Running on public URL: https://<gradio-hash>.gradio.live
+
+**Optional**: Serve the 65B model instead of the default 13B and use more GPU:
+
+.. code-block:: console
+
+    sky launch -c vllm-serve-new -s serve.yaml --gpus A100:8 --env MODEL_NAME=decapoda-research/llama-65b-hf
+

docs/source/serving/serving_with_langchain.rst

+.. _run_on_langchain:
+
+Serving with Langchain
+============================
+
+vLLM is also available via `Langchain <https://github.com/langchain-ai/langchain>`_ .
+
+To install langchain, run
+
+.. code-block:: console
+
+    $ pip install langchain -q
+
+To run inference on a single or multiple GPUs, use ``VLLM`` class from ``langchain``.
+
+.. code-block:: python
+
+    from langchain.llms import VLLM
+
+    llm = VLLM(model="mosaicml/mpt-7b",
+               trust_remote_code=True,  # mandatory for hf models
+               max_new_tokens=128,
+               top_k=10,
+               top_p=0.95,
+               temperature=0.8,
+               # tensor_parallel_size=... # for distributed inference
+    )
+
+    print(llm("What is the capital of France ?"))
+
+Please refer to this `Tutorial <https://github.com/langchain-ai/langchain/blob/master/docs/docs/integrations/llms/vllm.ipynb>`_ for more details.
\ No newline at end of file

examples/api_client.py

+"""Example Python client for vllm.entrypoints.api_server"""
+
+import argparse
+import json
+from typing import Iterable, List
+
+import requests
+
+
+def clear_line(n: int = 1) -> None:
+    LINE_UP = '\033[1A'
+    LINE_CLEAR = '\x1b[2K'
+    for _ in range(n):
+        print(LINE_UP, end=LINE_CLEAR, flush=True)
+
+
+def post_http_request(prompt: str,
+                      api_url: str,
+                      n: int = 1,
+                      stream: bool = False) -> requests.Response:
+    headers = {"User-Agent": "Test Client"}
+    pload = {
+        "prompt": prompt,
+        "n": n,
+        "use_beam_search": True,
+        "temperature": 0.0,
+        "max_tokens": 16,
+        "stream": stream,
+    }
+    response = requests.post(api_url, headers=headers, json=pload, stream=True)
+    return response
+
+
+def get_streaming_response(response: requests.Response) -> Iterable[List[str]]:
+    for chunk in response.iter_lines(chunk_size=8192,
+                                     decode_unicode=False,
+                                     delimiter=b"\0"):
+        if chunk:
+            data = json.loads(chunk.decode("utf-8"))
+            output = data["text"]
+            yield output
+
+
+def get_response(response: requests.Response) -> List[str]:
+    data = json.loads(response.content)
+    output = data["text"]
+    return output
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--host", type=str, default="localhost")
+    parser.add_argument("--port", type=int, default=8000)
+    parser.add_argument("--n", type=int, default=4)
+    parser.add_argument("--prompt", type=str, default="San Francisco is a")
+    parser.add_argument("--stream", action="store_true")
+    args = parser.parse_args()
+    prompt = args.prompt
+    api_url = f"http://{args.host}:{args.port}/generate"
+    n = args.n
+    stream = args.stream
+
+    print(f"Prompt: {prompt!r}\n", flush=True)
+    response = post_http_request(prompt, api_url, n, stream)
+
+    if stream:
+        num_printed_lines = 0
+        for h in get_streaming_response(response):
+            clear_line(num_printed_lines)
+            num_printed_lines = 0
+            for i, line in enumerate(h):
+                num_printed_lines += 1
+                print(f"Beam candidate {i}: {line!r}", flush=True)
+    else:
+        output = get_response(response)
+        for i, line in enumerate(output):
+            print(f"Beam candidate {i}: {line!r}", flush=True)

examples/gradio_webserver.py

+import argparse
+import json
+
+import gradio as gr
+import requests
+
+
+def http_bot(prompt):
+    headers = {"User-Agent": "vLLM Client"}
+    pload = {
+        "prompt": prompt,
+        "stream": True,
+        "max_tokens": 128,
+    }
+    response = requests.post(args.model_url,
+                             headers=headers,
+                             json=pload,
+                             stream=True)
+
+    for chunk in response.iter_lines(chunk_size=8192,
+                                     decode_unicode=False,
+                                     delimiter=b"\0"):
+        if chunk:
+            data = json.loads(chunk.decode("utf-8"))
+            output = data["text"][0]
+            yield output
+
+
+def build_demo():
+    with gr.Blocks() as demo:
+        gr.Markdown("# vLLM text completion demo\n")
+        inputbox = gr.Textbox(label="Input",
+                              placeholder="Enter text and press ENTER")
+        outputbox = gr.Textbox(label="Output",
+                               placeholder="Generated result from the model")
+        inputbox.submit(http_bot, [inputbox], [outputbox])
+    return demo
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--host", type=str, default=None)
+    parser.add_argument("--port", type=int, default=8001)
+    parser.add_argument("--model-url",
+                        type=str,
+                        default="http://localhost:8000/generate")
+    args = parser.parse_args()
+
+    demo = build_demo()
+    demo.queue(concurrency_count=100).launch(server_name=args.host,
+                                             server_port=args.port,
+                                             share=True)

examples/llm_engine_example.py

+import argparse
+from typing import List, Tuple
+
+from vllm import EngineArgs, LLMEngine, SamplingParams, RequestOutput
+
+
+def create_test_prompts() -> List[Tuple[str, SamplingParams]]:
+    """Create a list of test prompts with their sampling parameters."""
+    return [
+        ("A robot may not injure a human being",
+         SamplingParams(temperature=0.0, logprobs=1, prompt_logprobs=1)),
+        ("To be or not to be,",
+         SamplingParams(temperature=0.8, top_k=5, presence_penalty=0.2)),
+        ("What is the meaning of life?",
+         SamplingParams(n=2,
+                        best_of=5,
+                        temperature=0.8,
+                        top_p=0.95,
+                        frequency_penalty=0.1)),
+        ("It is only with the heart that one can see rightly",
+         SamplingParams(n=3, best_of=3, use_beam_search=True,
+                        temperature=0.0)),
+    ]
+
+
+def process_requests(engine: LLMEngine,
+                     test_prompts: List[Tuple[str, SamplingParams]]):
+    """Continuously process a list of prompts and handle the outputs."""
+    request_id = 0
+
+    while test_prompts or engine.has_unfinished_requests():
+        if test_prompts:
+            prompt, sampling_params = test_prompts.pop(0)
+            engine.add_request(str(request_id), prompt, sampling_params)
+            request_id += 1
+
+        request_outputs: List[RequestOutput] = engine.step()
+
+        for request_output in request_outputs:
+            if request_output.finished:
+                print(request_output)
+
+
+def initialize_engine(args: argparse.Namespace) -> LLMEngine:
+    """Initialize the LLMEngine from the command line arguments."""
+    engine_args = EngineArgs.from_cli_args(args)
+    return LLMEngine.from_engine_args(engine_args)
+
+
+def main(args: argparse.Namespace):
+    """Main function that sets up and runs the prompt processing."""
+    engine = initialize_engine(args)
+    test_prompts = create_test_prompts()
+    process_requests(engine, test_prompts)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        description='Demo on using the LLMEngine class directly')
+    parser = EngineArgs.add_cli_args(parser)
+    args = parser.parse_args()
+    main(args)

examples/offline_inference.py

+from vllm import LLM, SamplingParams
+
+# Sample prompts.
+prompts = [
+    "Hello, my name is",
+    "The president of the United States is",
+    "The capital of France is",
+    "The future of AI is",
+]
+# Create a sampling params object.
+sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
+
+# Create an LLM.
+llm = LLM(model="facebook/opt-125m")
+# Generate texts from the prompts. The output is a list of RequestOutput objects
+# that contain the prompt, generated text, and other information.
+outputs = llm.generate(prompts, sampling_params)
+# Print the outputs.
+for output in outputs:
+    prompt = output.prompt
+    generated_text = output.outputs[0].text
+    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")

examples/openai_chatcompletion_client.py

+from openai import OpenAI
+
+# Modify OpenAI's API key and API base to use vLLM's API server.
+openai_api_key = "EMPTY"
+openai_api_base = "http://localhost:8000/v1"
+
+client = OpenAI(
+    # defaults to os.environ.get("OPENAI_API_KEY")
+    api_key=openai_api_key,
+    base_url=openai_api_base,
+)
+
+models = client.models.list()
+model = models.data[0].id
+
+chat_completion = client.chat.completions.create(
+    messages=[{
+        "role": "system",
+        "content": "You are a helpful assistant."
+    }, {
+        "role": "user",
+        "content": "Who won the world series in 2020?"
+    }, {
+        "role":
+        "assistant",
+        "content":
+        "The Los Angeles Dodgers won the World Series in 2020."
+    }, {
+        "role": "user",
+        "content": "Where was it played?"
+    }],
+    model=model,
+)
+
+
+print("Chat completion results:")
+print(chat_completion)

examples/openai_completion_client.py

+from openai import OpenAI
+
+# Modify OpenAI's API key and API base to use vLLM's API server.
+openai_api_key = "EMPTY"
+openai_api_base = "http://localhost:8000/v1"
+
+client = OpenAI(
+    # defaults to os.environ.get("OPENAI_API_KEY")
+    api_key=openai_api_key,
+    base_url=openai_api_base,
+)
+
+models = client.models.list()
+model = models.data[0].id
+
+# Completion API
+stream = False
+completion = client.completions.create(
+    model=model,
+    prompt="A robot may not injure a human being",
+    echo=False,
+    n=2,
+    stream=stream,
+    logprobs=3
+)
+
+print("Completion results:")
+if stream:
+    for c in completion:
+        print(c)
+else:
+    print(completion)

examples/template_alpaca.jinja

+{{ (messages|selectattr('role', 'equalto', 'system')|list|last).content|trim if (messages|selectattr('role', 'equalto', 'system')|list) else '' }}
+
+{% for message in messages %}
+{% if message['role'] == 'user' %}
+### Instruction:
+{{ message['content']|trim -}}
+{% if not loop.last %}
+
+
+{% endif %}
+{% elif message['role'] == 'assistant' %}
+### Response:
+{{ message['content']|trim -}}
+{% if not loop.last %}
+
+
+{% endif %}
+{% elif message['role'] == 'user_context' %}
+### Input:
+{{ message['content']|trim -}}
+{% if not loop.last %}
+
+
+{% endif %}
+{% endif %}
+{% endfor %}
+{% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}
+### Response:
+{% endif %}
\ No newline at end of file

examples/template_chatml.jinja

+{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content']}}{% if (loop.last and add_generation_prompt) or not loop.last %}{{ '<|im_end|>' + '\n'}}{% endif %}{% endfor %}
+{% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}{{ '<|im_start|>assistant\n' }}{% endif %}
\ No newline at end of file

examples/template_inkbot.jinja

+<#meta#>
+- Date: {{ (messages|selectattr('role', 'equalto', 'meta-current_date')|list|last).content|trim if (messages|selectattr('role', 'equalto', 'meta-current_date')|list) else '' }}
+- Task: {{ (messages|selectattr('role', 'equalto', 'meta-task_name')|list|last).content|trim if (messages|selectattr('role', 'equalto', 'meta-task_name')|list) else '' }}
+<#system#>
+{{ (messages|selectattr('role', 'equalto', 'system')|list|last).content|trim if (messages|selectattr('role', 'equalto', 'system')|list) else '' }}
+<#chat#>
+{% for message in messages %}
+{% if message['role'] == 'user' %}
+<#user#>
+{{ message['content']|trim -}}
+{% if not loop.last %}
+
+{% endif %}
+{% elif message['role'] == 'assistant' %}
+<#bot#>
+{{ message['content']|trim -}}
+{% if not loop.last %}
+
+{% endif %}
+{% elif message['role'] == 'user_context' %}
+<#user_context#>
+{{ message['content']|trim -}}
+{% if not loop.last %}
+
+{% endif %}
+{% endif %}
+{% endfor %}
+{% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}
+<#bot#>
+{% endif %}
\ No newline at end of file