@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / cfe712bf1a
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 16, 17:00 UTC

[CI/Build] Use python 3.12 in cuda image (#8133)

Signed-off-by: Joe Runde <Joseph.Runde@ibm.com>

Joe Runde committed Sep 7, 2024 at 14:03 UTC cfe712bf1aedbee4f26105737710ff80ae9d624e
3 files changed +10 -5
Dockerfile
+6 -2
@@ -10,7 +10,7 @@ ARG CUDA_VERSION=12.4.1
10 # prepare basic build environment
11 FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu20.04 AS base
12 ARG CUDA_VERSION=12.4.1
13 -ARG PYTHON_VERSION=3.10
13 +ARG PYTHON_VERSION=3.12
14 ENV DEBIAN_FRONTEND=noninteractive
15
16 # Install Python and other dependencies
@@ -133,7 +133,7 @@ RUN --mount=type=cache,target=/root/.cache/pip \
133 # image with vLLM installed
134 FROM nvidia/cuda:${CUDA_VERSION}-base-ubuntu20.04 AS vllm-base
135 ARG CUDA_VERSION=12.4.1
136 -ARG PYTHON_VERSION=3.10
136 +ARG PYTHON_VERSION=3.12
137 WORKDIR /vllm-workspace
138 ENV DEBIAN_FRONTEND=noninteractive
139
@@ -179,6 +179,10 @@ FROM vllm-base AS test
179 ADD . /vllm-workspace/
180
181 # install development dependencies (for testing)
182 +# A newer setuptools is required for installing some test dependencies from source that do not publish python 3.12 wheels
183 +# This installation must complete before the test dependencies are collected and installed.
184 +RUN --mount=type=cache,target=/root/.cache/pip \
185 + python3 -m pip install "setuptools>=74.1.1"
186 RUN --mount=type=cache,target=/root/.cache/pip \
187 python3 -m pip install -r requirements-dev.txt
188
requirements-common.txt
+1
@@ -27,3 +27,4 @@ gguf == 0.9.1
27 importlib_metadata
28 mistral_common >= 1.3.4
29 pyyaml
30 +six>=1.16.0; python_version > '3.11' # transitive dependency of pandas that needs to be the latest version for python 3.12
tests/test_logger.py
+3 -3
@@ -95,7 +95,7 @@ def test_logger_configuring_can_be_disabled():
95 config behavior, however mocks are used to ensure no changes in behavior or
96 configuration occur."""
97
98 - with patch("logging.config.dictConfig") as dict_config_mock:
98 + with patch("vllm.logger.dictConfig") as dict_config_mock:
99 _configure_vllm_root_logger()
100 dict_config_mock.assert_not_called()
101
@@ -175,9 +175,9 @@ def test_custom_logging_config_is_parsed_and_used_when_provided():
175 logging_config_file.flush()
176 with patch("vllm.logger.VLLM_LOGGING_CONFIG_PATH",
177 logging_config_file.name), patch(
178 - "logging.config.dictConfig") as dict_config_mock:
178 + "vllm.logger.dictConfig") as dict_config_mock:
179 _configure_vllm_root_logger()
180 - assert dict_config_mock.called_with(valid_logging_config)
180 + dict_config_mock.assert_called_with(valid_logging_config)
181
182
183 @patch("vllm.logger.VLLM_CONFIGURE_LOGGING", 0)