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, 03:30 UTC
[XPU][CI] add v1/core test in xpu hardware ci (#20537)
Signed-off-by: Ma, Liangliang <liangliang.ma@intel.com>
Liangliang Ma committed
Jul 7, 2025 at 16:16 UTC
2c5ebec064bf3684c8f02b70b5963615daa81b28
3 files changed
+6
-8
.buildkite/scripts/hardware_ci/run-xpu-test.sh
+4
-2
@@ -11,8 +11,8 @@ container_name="xpu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head
11
docker build -t ${image_name} -f docker/Dockerfile.xpu .
12
13
# Setup cleanup
14
-remove_docker_container() {
15
- docker rm -f "${container_name}" || true;
14
+remove_docker_container() {
15
+ docker rm -f "${container_name}" || true;
16
docker image rm -f "${image_name}" || true;
17
docker system prune -f || true;
18
}
@@ -27,4 +27,6 @@ docker run \
27
"${image_name}" \
28
sh -c '
29
VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
30
+ cd tests
31
+ pytest -v -s v1/core
32
'
docker/Dockerfile.xpu
+1
-1
@@ -47,7 +47,7 @@ FROM vllm-base AS vllm-openai
47
48
# install additional dependencies for openai api server
49
RUN --mount=type=cache,target=/root/.cache/pip \
50
- pip install accelerate hf_transfer 'modelscope!=1.15.0'
50
+ pip install accelerate hf_transfer pytest 'modelscope!=1.15.0'
51
52
ENV VLLM_USAGE_SOURCE production-docker-image \
53
TRITON_XPU_PROFILE 1
vllm/platforms/xpu.py
+1
-5
@@ -93,10 +93,6 @@ class XPUPlatform(Platform):
93
"mode.")
94
model_config.enforce_eager = True
95
96
- if vllm_config.speculative_config is not None:
97
- raise NotImplementedError(
98
- "XPU does not support speculative decoding")
99
-
96
if vllm_config.device_config is not None:
97
assert vllm_config.device_config.device_type == "xpu"
98
@@ -181,4 +177,4 @@ class XPUPlatform(Platform):
177
178
@classmethod
179
def device_count(cls) -> int:
184
- return torch.xpu.device_count()
\ No newline at end of file
180
+ return torch.xpu.device_count()