Merge pull request #1 from vllm-project/main

merge master

Dockerfile

@@ -122,7 +122,7 @@ RUN --mount=type=bind,from=flash-attn-builder,src=/usr/src/flash-attention-v2,ta
 FROM vllm-base AS vllm-openai
 # install additional dependencies for openai api server
 RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install accelerate hf_transfer
+    pip install accelerate hf_transfer modelscope
 
 COPY --from=build /workspace/vllm/*.so /workspace/vllm/
 COPY vllm vllm

docs/requirements-docs.txt

@@ -1,3 +1,10 @@
 sphinx == 6.2.1
 sphinx-book-theme == 1.0.1
 sphinx-copybutton == 0.5.2
+myst-parser == 2.0.0
+sphinx-argparse
+
+# packages to install to build the documentation
+pydantic
+-f https://download.pytorch.org/whl/cpu
+torch

docs/source/conf.py

@@ -22,7 +22,7 @@
 # -- Project information -----------------------------------------------------
 
 project = 'vLLM'
-copyright = '2023, vLLM Team'
+copyright = '2024, vLLM Team'
 author = 'the vLLM Team'
 
 # -- General configuration ---------------------------------------------------
@@ -37,6 +37,8 @@
     "sphinx_copybutton",
     "sphinx.ext.autodoc",
     "sphinx.ext.autosummary",
+    "myst_parser",
+    "sphinxarg.ext",
 ]
 
 # Add any paths that contain templates here, relative to this directory.

docs/source/dev/sampling_params.rst

@@ -0,0 +1,4 @@
+Sampling Params
+===============
+
+.. automodule:: vllm.sampling_params.SamplingParams

docs/source/getting_started/amd-installation.rst

@@ -100,7 +100,7 @@ You can build and install vLLM from source:
 
 Build a docker image from `Dockerfile.rocm`, and launch a docker container.
 
-The `Dokerfile.rocm` is designed to support both ROCm 5.7 and ROCm 6.0 and later versions. It provides flexibility to customize the build of docker image using the following arguments:
+The `Dockerfile.rocm` is designed to support both ROCm 5.7 and ROCm 6.0 and later versions. It provides flexibility to customize the build of docker image using the following arguments:
 
 * `BASE_IMAGE`: specifies the base image used when running ``docker build``, specifically the PyTorch on ROCm base image. We have tested ROCm 5.7 and ROCm 6.0. The default is `rocm/pytorch:rocm6.0_ubuntu20.04_py3.9_pytorch_2.1.1`
 * `FX_GFX_ARCHS`: specifies the GFX architecture that is used to build flash-attention, for example, `gfx90a;gfx942` for MI200 and MI300. The default is `gfx90a;gfx942`

docs/source/getting_started/neuron-installation.rst

@@ -128,6 +128,7 @@ Once neuronx-cc and transformers-neuronx packages are installed, we will be able
 
 .. code-block:: console
 
+    $ git clone https://github.com/vllm-project/vllm.git
     $ cd vllm
     $ pip install -U -r requirements-neuron.txt
     $ pip install .

docs/source/index.rst

@@ -69,14 +69,11 @@ Documentation
    :maxdepth: 1
    :caption: Serving
 
-   serving/distributed_serving
-   serving/run_on_sky
-   serving/deploying_with_kserve
-   serving/deploying_with_triton
-   serving/deploying_with_bentoml
+   serving/openai_compatible_server
    serving/deploying_with_docker
-   serving/serving_with_langchain
+   serving/distributed_serving
    serving/metrics
+   serving/integrations
 
 .. toctree::
    :maxdepth: 1
@@ -98,6 +95,7 @@ Documentation
    :maxdepth: 2
    :caption: Developer Documentation
 
+   dev/sampling_params
    dev/engine/engine_index
    dev/kernel/paged_attention
 

docs/source/models/lora.rst

@@ -90,7 +90,7 @@ Requests can specify the LoRA adapter as if it were any other model via the ``mo
 processed according to the server-wide LoRA configuration (i.e. in parallel with base model requests, and potentially other
 LoRA adapter requests if they were provided and ``max_loras`` is set high enough).
 
-The following is an example request 
+The following is an example request
 
 .. code-block:: bash
 

docs/source/serving/integrations.rst

@@ -0,0 +1,11 @@
+Integrations
+------------
+
+.. toctree::
+   :maxdepth: 1
+
+   run_on_sky
+   deploying_with_kserve
+   deploying_with_triton
+   deploying_with_bentoml
+   serving_with_langchain

docs/source/serving/openai_compatible_server.md

@@ -0,0 +1,114 @@
+# OpenAI Compatible Server
+
+vLLM provides an HTTP server that implements OpenAI's [Completions](https://platform.openai.com/docs/api-reference/completions) and [Chat](https://platform.openai.com/docs/api-reference/chat) API.
+
+You can start the server using Python, or using [Docker](deploying_with_docker.rst):
+```bash
+python -m vllm.entrypoints.openai.api_server --model meta-llama/Llama-2-7b-hf --dtype float32 --api-key token-abc123
+```
+
+To call the server, you can use the official OpenAI Python client library, or any other HTTP client.
+```python
+from openai import OpenAI
+client = OpenAI(
+    base_url="http://localhost:8000/v1",
+    api_key="token-abc123",
+)
+
+completion = client.chat.completions.create(
+  model="meta-llama/Llama-2-7b-hf",
+  messages=[
+    {"role": "system", "content": "You are a helpful assistant."},
+    {"role": "user", "content": "Hello!"}
+  ]
+)
+
+print(completion.choices[0].message)
+```
+
+## API Reference
+Please see the [OpenAI API Reference](https://platform.openai.com/docs/api-reference) for more information on the API. We support all parameters except:
+- Chat: `tools`, and `tool_choice`.
+- Completions: `suffix`.
+
+## Extra Parameters
+vLLM supports a set of parameters that are not part of the OpenAI API.
+In order to use them, you can pass them as extra parameters in the OpenAI client.
+Or directly merge them into the JSON payload if you are using HTTP call directly.
+
+```python
+completion = client.chat.completions.create(
+  model="meta-llama/Llama-2-7b-hf",
+  messages=[
+    {"role": "system", "content": "You are a helpful assistant."},
+    {"role": "user", "content": "Classify this sentiment: vLLM is wonderful!"}
+  ],
+  extra_body={
+    "guided_choice": ["positive", "negative"]
+  }
+)
+```
+
+### Extra Parameters for Chat API
+The following [sampling parameters (click through to see documentation)](../dev/sampling_params.rst) are supported.
+
+```{literalinclude} ../../../vllm/entrypoints/openai/protocol.py
+:language: python
+:start-after: begin-chat-completion-sampling-params
+:end-before: end-chat-completion-sampling-params
+```
+
+The following extra parameters are supported:
+
+```{literalinclude} ../../../vllm/entrypoints/openai/protocol.py
+:language: python
+:start-after: begin-chat-completion-extra-params
+:end-before: end-chat-completion-extra-params
+```
+
+### Extra Parameters for Completions API
+The following [sampling parameters (click through to see documentation)](../dev/sampling_params.rst) are supported.
+
+```{literalinclude} ../../../vllm/entrypoints/openai/protocol.py
+:language: python
+:start-after: begin-completion-sampling-params
+:end-before: end-completion-sampling-params
+```
+
+The following extra parameters are supported:
+
+```{literalinclude} ../../../vllm/entrypoints/openai/protocol.py
+:language: python
+:start-after: begin-completion-extra-params
+:end-before: end-completion-extra-params
+```
+
+## Chat Template
+
+In order for the language model to support chat protocol, vLLM requires the model to include
+a chat template in its tokenizer configuration. The chat template is a Jinja2 template that
+specifies how are roles, messages, and other chat-specific tokens are encoded in the input.
+
+An example chat template for `meta-llama/Llama-2-7b-chat-hf` can be found [here](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf/blob/09bd0f49e16738cdfaa6e615203e126038736eb0/tokenizer_config.json#L12)
+
+Some models do not provide a chat template even though they are instruction/chat fine-tuned. For those model,
+you can manually specify their chat template in the `--chat-template` parameter with the file path to the chat
+template, or the template in string form. Without a chat template, the server will not be able to process chat
+and all chat requests will error.
+
+```bash
+python -m vllm.entrypoints.openai.api_server \
+  --model ... \
+  --chat-template ./path-to-chat-template.jinja
+```
+
+vLLM community provides a set of chat templates for popular models. You can find them in the examples
+directory [here](https://github.com/vllm-project/vllm/tree/main/examples/)
+
+## Command line arguments for the server
+
+```{argparse}
+:module: vllm.entrypoints.openai.cli_args
+:func: make_arg_parser
+:prog: vllm-openai-server
+```

vllm/config.py

@@ -577,12 +577,12 @@ class DeviceConfig:
     def __init__(self, device: str = "auto") -> None:
         if device == "auto":
             # Automated device type detection
-            if torch.cuda.is_available():
-                self.device_type = "cuda"
-            elif is_neuron():
+            if is_neuron():
                 self.device_type = "neuron"
             else:
-                raise RuntimeError("No supported device detected.")
+                # We don't call torch.cuda.is_available() here to
+                # avoid initializing CUDA before workers are forked
+                self.device_type = "cuda"
         else:
             # Device type is assigned explicitly
             self.device_type = device

vllm/engine/ray_utils.py

@@ -33,8 +33,17 @@ def __getattr__(self, name):
             return getattr(self.worker, name)
 
         def execute_method(self, method, *args, **kwargs):
-            executor = getattr(self, method)
-            return executor(*args, **kwargs)
+            try:
+                executor = getattr(self, method)
+                return executor(*args, **kwargs)
+            except Exception as e:
+                # exceptions in ray worker may cause deadlock
+                # see https://github.com/vllm-project/vllm/issues/3455
+                # print the error and inform the user to solve the error
+                msg = (f"Error executing method {method}. "
+                       "This might cause deadlock in distributed execution.")
+                logger.exception(msg)
+                raise e
 
         def get_node_ip(self) -> str:
             return get_ip()

vllm/entrypoints/openai/api_server.py

@@ -1,11 +1,8 @@
-import argparse
 import asyncio
-import json
 from contextlib import asynccontextmanager
 import os
 import importlib
 import inspect
-import ssl
 
 from prometheus_client import make_asgi_app
 import fastapi
@@ -23,9 +20,9 @@
                                               ChatCompletionRequest,
                                               ErrorResponse)
 from vllm.logger import init_logger
+from vllm.entrypoints.openai.cli_args import make_arg_parser
 from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
 from vllm.entrypoints.openai.serving_completion import OpenAIServingCompletion
-from vllm.entrypoints.openai.serving_engine import LoRA
 
 TIMEOUT_KEEP_ALIVE = 5  # seconds
 
@@ -51,109 +48,8 @@ async def _force_log():
 app = fastapi.FastAPI(lifespan=lifespan)
 
 
-class LoRAParserAction(argparse.Action):
-
-    def __call__(self, parser, namespace, values, option_string=None):
-        lora_list = []
-        for item in values:
-            name, path = item.split('=')
-            lora_list.append(LoRA(name, path))
-        setattr(namespace, self.dest, lora_list)
-
-
 def parse_args():
-    parser = argparse.ArgumentParser(
-        description="vLLM OpenAI-Compatible RESTful API server.")
-    parser.add_argument("--host", type=str, default=None, help="host name")
-    parser.add_argument("--port", type=int, default=8000, help="port number")
-    parser.add_argument(
-        "--uvicorn-log-level",
-        type=str,
-        default="info",
-        choices=['debug', 'info', 'warning', 'error', 'critical', 'trace'],
-        help="log level for uvicorn")
-    parser.add_argument("--allow-credentials",
-                        action="store_true",
-                        help="allow credentials")
-    parser.add_argument("--allowed-origins",
-                        type=json.loads,
-                        default=["*"],
-                        help="allowed origins")
-    parser.add_argument("--allowed-methods",
-                        type=json.loads,
-                        default=["*"],
-                        help="allowed methods")
-    parser.add_argument("--allowed-headers",
-                        type=json.loads,
-                        default=["*"],
-                        help="allowed headers")
-    parser.add_argument("--api-key",
-                        type=str,
-                        default=None,
-                        help="If provided, the server will require this key "
-                        "to be presented in the header.")
-    parser.add_argument("--served-model-name",
-                        type=str,
-                        default=None,
-                        help="The model name used in the API. If not "
-                        "specified, the model name will be the same as "
-                        "the huggingface name.")
-    parser.add_argument(
-        "--lora-modules",
-        type=str,
-        default=None,
-        nargs='+',
-        action=LoRAParserAction,
-        help="LoRA module configurations in the format name=path. "
-        "Multiple modules can be specified.")
-    parser.add_argument("--chat-template",
-                        type=str,
-                        default=None,
-                        help="The file path to the chat template, "
-                        "or the template in single-line form "
-                        "for the specified model")
-    parser.add_argument("--response-role",
-                        type=str,
-                        default="assistant",
-                        help="The role name to return if "
-                        "`request.add_generation_prompt=true`.")
-    parser.add_argument("--ssl-keyfile",
-                        type=str,
-                        default=None,
-                        help="The file path to the SSL key file")
-    parser.add_argument("--ssl-certfile",
-                        type=str,
-                        default=None,
-                        help="The file path to the SSL cert file")
-    parser.add_argument("--ssl-ca-certs",
-                        type=str,
-                        default=None,
-                        help="The CA certificates file")
-    parser.add_argument(
-        "--ssl-cert-reqs",
-        type=int,
-        default=int(ssl.CERT_NONE),
-        help="Whether client certificate is required (see stdlib ssl module's)"
-    )
-    parser.add_argument(
-        "--root-path",
-        type=str,
-        default=None,
-        help="FastAPI root_path when app is behind a path based routing proxy")
-    parser.add_argument(
-        "--middleware",
-        type=str,
-        action="append",
-        default=[],
-        help="Additional ASGI middleware to apply to the app. "
-        "We accept multiple --middleware arguments. "
-        "The value should be an import path. "
-        "If a function is provided, vLLM will add it to the server "
-        "using @app.middleware('http'). "
-        "If a class is provided, vLLM will add it to the server "
-        "using app.add_middleware(). ")
-
-    parser = AsyncEngineArgs.add_cli_args(parser)
+    parser = make_arg_parser()
     return parser.parse_args()