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, 01:00 UTC
[CI][HPU] update for v0 deprecate by switching to VLLM_TARGET_DEVICE=empty (#21006)
Signed-off-by: Chendi.Xue <chendi.xue@intel.com>
Chendi.Xue committed
Jul 15, 2025 at 22:07 UTC
e9534c7202e20b27c1c82f2736415a3a9ddc1be7
1 file changed
+3
-5
.buildkite/scripts/hardware_ci/run-hpu-test.sh
+3
-5
@@ -6,19 +6,17 @@ set -exuo pipefail
6
7
# Try building the docker image
8
cat <<EOF | docker build -t hpu-plugin-v1-test-env -f - .
9
-FROM 1.22-413-pt2.7.1:latest
9
+FROM gaudi-base-image:latest
10
11
COPY ./ /workspace/vllm
12
13
WORKDIR /workspace/vllm
14
15
-RUN pip install -v -r requirements/hpu.txt
16
-RUN pip install git+https://github.com/vllm-project/vllm-gaudi.git
17
-
15
ENV no_proxy=localhost,127.0.0.1
16
ENV PT_HPU_ENABLE_LAZY_COLLECTIVES=true
17
21
-RUN VLLM_TARGET_DEVICE=hpu python3 setup.py install
18
+RUN VLLM_TARGET_DEVICE=empty pip install .
19
+RUN pip install git+https://github.com/vllm-project/vllm-gaudi.git
20
21
# install development dependencies (for testing)
22
RUN python3 -m pip install -e tests/vllm_test_utils