From d573aeadcc891976f09d6d50f1a4f98c8ff809aa Mon Sep 17 00:00:00 2001 From: Michael Goin Date: Fri, 20 Dec 2024 14:03:50 -0500 Subject: [PATCH 1/3] [Bugfix] Don't log OpenAI field aliases as ignored (#11378) Signed-off-by: mgoin --- vllm/entrypoints/openai/protocol.py | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/vllm/entrypoints/openai/protocol.py b/vllm/entrypoints/openai/protocol.py index 1314de714215e..1d8b0d19f9516 100644 --- a/vllm/entrypoints/openai/protocol.py +++ b/vllm/entrypoints/openai/protocol.py @@ -46,7 +46,15 @@ class OpenAIBaseModel(BaseModel): @classmethod def __log_extra_fields__(cls, data): if isinstance(data, dict): - extra_fields = data.keys() - cls.model_fields.keys() + # Get all class field names and their potential aliases + field_names = set() + for field_name, field in cls.model_fields.items(): + field_names.add(field_name) + if hasattr(field, 'alias') and field.alias: + field_names.add(field.alias) + + # Compare against both field names and aliases + extra_fields = data.keys() - field_names if extra_fields: logger.warning( "The following fields were present in the request " From 5d2248d81ab1f83a2874bfa726f0a1933ef2d048 Mon Sep 17 00:00:00 2001 From: youkaichao Date: Fri, 20 Dec 2024 13:00:56 -0800 Subject: [PATCH 2/3] [doc] explain nccl requirements for rlhf (#11381) Signed-off-by: youkaichao --- docs/source/getting_started/debugging.rst | 1 + 1 file changed, 1 insertion(+) diff --git a/docs/source/getting_started/debugging.rst b/docs/source/getting_started/debugging.rst index 7f36d65a227f0..b123960533816 100644 --- a/docs/source/getting_started/debugging.rst +++ b/docs/source/getting_started/debugging.rst @@ -200,3 +200,4 @@ try this instead: Known Issues ---------------------------------------- - In ``v0.5.2``, ``v0.5.3``, and ``v0.5.3.post1``, there is a bug caused by `zmq `_ , which can occasionally cause vLLM to hang depending on the machine configuration. The solution is to upgrade to the latest version of ``vllm`` to include the `fix `_. +- To circumvent a NCCL `bug `__ , all vLLM processes will set an environment variable ``NCCL_CUMEM_ENABLE=0`` to disable NCCL's ``cuMem`` allocator. It does not affect performance but only gives memory benefits. When external processes want to set up a NCCL connection with vLLM's processes, they should also set this environment variable, otherwise, inconsistent environment setup will cause NCCL to hang or crash, as observed in `the RLHF integration `__ and the `discussion `__ . From 47a0b615b45efd0a9ed57049d8ca6eff1c249844 Mon Sep 17 00:00:00 2001 From: Jiaxin Shan Date: Fri, 20 Dec 2024 13:54:55 -0800 Subject: [PATCH 3/3] Add ray[default] to wget to run distributed inference out of box (#11265) Signed-off-by: Jiaxin Shan --- Dockerfile | 2 +- requirements-cuda.txt | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/Dockerfile b/Dockerfile index 84350cde59bfb..6226569e9d3b4 100644 --- a/Dockerfile +++ b/Dockerfile @@ -163,7 +163,7 @@ RUN PYTHON_VERSION_STR=$(echo ${PYTHON_VERSION} | sed 's/\.//g') && \ RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \ && echo 'tzdata tzdata/Zones/America select Los_Angeles' | debconf-set-selections \ && apt-get update -y \ - && apt-get install -y ccache software-properties-common git curl sudo vim python3-pip \ + && apt-get install -y ccache software-properties-common git curl wget sudo vim python3-pip \ && apt-get install -y ffmpeg libsm6 libxext6 libgl1 \ && add-apt-repository ppa:deadsnakes/ppa \ && apt-get update -y \ diff --git a/requirements-cuda.txt b/requirements-cuda.txt index 058ab7c1ee9df..8002fbd8ee5b9 100644 --- a/requirements-cuda.txt +++ b/requirements-cuda.txt @@ -2,7 +2,7 @@ -r requirements-common.txt # Dependencies for NVIDIA GPUs -ray >= 2.9 +ray[default] >= 2.9 nvidia-ml-py >= 12.560.30 # for pynvml package torch == 2.5.1 # These must be updated alongside torch