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 14, 23:45 UTC
[Bugfix] Fix environment variable setting in CPU Dockerfile (#21730)
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Li, Jiang committed
Jul 28, 2025 at 19:02 UTC
65e8466c37a3bea48dd68393d01d1e96cf2e2d30
2 files changed
+18
-19
.buildkite/scripts/hardware_ci/run-cpu-test.sh
+6
-6
@@ -78,6 +78,12 @@ function cpu_tests() {
78
# VLLM_USE_V1=0 pytest -s -v \
79
# tests/quantization/test_ipex_quant.py"
80
81
+ # Run multi-lora tests
82
+ docker exec cpu-test-"$NUMA_NODE" bash -c "
83
+ set -e
84
+ pytest -s -v \
85
+ tests/lora/test_qwen2vl.py"
86
+
87
# online serving
88
docker exec cpu-test-"$NUMA_NODE" bash -c '
89
set -e
@@ -89,12 +95,6 @@ function cpu_tests() {
95
--model meta-llama/Llama-3.2-3B-Instruct \
96
--num-prompts 20 \
97
--endpoint /v1/completions'
92
-
93
- # Run multi-lora tests
94
- docker exec cpu-test-"$NUMA_NODE" bash -c "
95
- set -e
96
- pytest -s -v \
97
- tests/lora/test_qwen2vl.py"
98
}
99
100
# All of CPU tests are expected to be finished less than 40 mins.
docker/Dockerfile.cpu
+12
-13
@@ -19,16 +19,14 @@
19
# VLLM_CPU_AVX512VNNI=false (default)|true
20
#
21
22
-######################### BASE IMAGE #########################
23
-FROM ubuntu:22.04 AS base
22
+######################### COMMON BASE IMAGE #########################
23
+FROM ubuntu:22.04 AS base-common
24
25
WORKDIR /workspace/
26
27
ARG PYTHON_VERSION=3.12
28
ARG PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu"
29
30
-ENV LD_PRELOAD=""
31
-
30
# Install minimal dependencies and uv
31
RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
32
--mount=type=cache,target=/var/lib/apt,sharing=locked \
@@ -63,17 +61,18 @@ RUN --mount=type=cache,target=/root/.cache/uv \
61
ARG TARGETARCH
62
ENV TARGETARCH=${TARGETARCH}
63
66
-RUN if [ "$TARGETARCH" = "arm64" ]; then \
67
- PRELOAD_PATH="/usr/lib/aarch64-linux-gnu/libtcmalloc_minimal.so.4"; \
68
- else \
69
- PRELOAD_PATH="/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4:/opt/venv/lib/libiomp5.so"; \
70
- fi && \
71
- echo "export LD_PRELOAD=$PRELOAD_PATH" >> ~/.bashrc
64
+######################### x86_64 BASE IMAGE #########################
65
+FROM base-common AS base-amd64
66
+
67
+ENV LD_PRELOAD="/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4:/opt/venv/lib/libiomp5.so"
68
73
-# Ensure that the LD_PRELOAD environment variable for export is in effect.
74
-SHELL ["/bin/bash", "-c"]
69
+######################### arm64 BASE IMAGE #########################
70
+FROM base-common AS base-arm64
71
76
-ENV LD_PRELOAD=${LD_PRELOAD}
72
+ENV LD_PRELOAD="/usr/lib/aarch64-linux-gnu/libtcmalloc_minimal.so.4"
73
+
74
+######################### BASE IMAGE #########################
75
+FROM base-${TARGETARCH} AS base
76
77
RUN echo 'ulimit -c 0' >> ~/.bashrc
78