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 15, 18:30 UTC
[CI][Intel GPU] refine intel GPU ci docker build (#14860)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Kunshang Ji committed
Mar 15, 2025 at 04:58 UTC
f58aea002c585d1957d46f3b9ab23642cab88d82
1 file changed
+14
-3
.buildkite/run-xpu-test.sh
+14
-3
@@ -4,16 +4,27 @@
4
# It serves a sanity check for compilation and basic model usage.
5
set -ex
6
7
+image_name="xpu/vllm-ci:${BUILDKITE_COMMIT}"
8
+container_name="xpu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head -c 10; echo)"
9
+
10
# Try building the docker image
8
-docker build -t xpu-test -f Dockerfile.xpu .
11
+docker build -t ${image_name} -f Dockerfile.xpu .
12
13
# Setup cleanup
11
-remove_docker_container() { docker rm -f xpu-test || true; }
14
+remove_docker_container() {
15
+ docker rm -f "${container_name}" || docker image rm -f "${image_name}" || true;
16
+}
17
trap remove_docker_container EXIT
18
remove_docker_container
19
20
# Run the image and test offline inference/tensor parallel
16
-docker run --name xpu-test --device /dev/dri -v /dev/dri/by-path:/dev/dri/by-path --entrypoint="" xpu-test sh -c '
21
+docker run \
22
+ --device /dev/dri \
23
+ -v /dev/dri/by-path:/dev/dri/by-path \
24
+ --entrypoint="" \
25
+ --name "${container_name}" \
26
+ "${image_name}" \
27
+ sh -c '
28
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m
29
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m -tp 2
30
'