Fix wheel glibc version incompatibility (#1582)
### What this PR does / why we need it?
- Fixes https://github.com/vllm-project/vllm-ascend/issues/1533
### How was this patch tested?
1. Run the image
```
docker run \
--name cann_container \
--device /dev/davinci6 \
--device /dev/davinci_manager \
--device /dev/devmm_svm \
--device /dev/hisi_hdc \
-v /usr/local/dcmi:/usr/local/dcmi \
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
-v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
-v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
-v /etc/ascend_install.info:/etc/ascend_install.info \
-it quay.io/ascend/cann:8.1.rc1-910b-openeuler22.03-py3.11 bash
```
2. Install package
torch=2.5.1
torch-npu=2.5.1.post1.dev20250619
vllm=0.9.1
vllm-ascend=vllm_ascend-0.1.dev1+g02ac443-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.whl
Artifact download URL:
https://github.com/vllm-project/vllm-ascend/actions/runs/16039661265/artifacts/3454481370
3. Test offline script
```
from vllm import LLM, SamplingParams
import os
os.environ["VLLM_USE_V1"] = "1"
prompts = [
"Hello, my name is",
]
llm = LLM(model="Qwen3/Qwen3-1.7B")
outputs = llm.generate(prompts)
for output in outputs:
prompt = output.prompt
generated_text = output.outputs[0].text
print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
```
4. Results

- vLLM version: v0.9.2
- vLLM main:
b942c094e3
---------
Signed-off-by: Icey <1790571317@qq.com>
This commit is contained in:
41
.github/workflows/release_whl.yml
vendored
41
.github/workflows/release_whl.yml
vendored
@@ -78,22 +78,51 @@ jobs:
|
||||
--build-arg PY_VERSION=${{ matrix.python-version }} \
|
||||
-t wheel:v1 .
|
||||
docker run --rm \
|
||||
-u $(id -u):$(id -g) \
|
||||
-v $(pwd):/outpwd \
|
||||
wheel:v1 \
|
||||
bash -c "cp -r /workspace/vllm-ascend/dist /outpwd"
|
||||
ls dist
|
||||
|
||||
- name: Archive wheel
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: vllm-ascend-${{ matrix.os }}-py${{ matrix.python-version }}-wheel
|
||||
path: dist/*
|
||||
|
||||
- name: Set up Python ${{ matrix.python-version }}
|
||||
if: startsWith(github.ref, 'refs/tags/')
|
||||
uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
|
||||
- name: Repair wheels with auditwheel
|
||||
run: |
|
||||
python3 -m pip install auditwheel
|
||||
python3 -m pip install patchelf
|
||||
mkdir -p dist/repaired
|
||||
for whl in dist/*.whl; do
|
||||
auditwheel repair "$whl" -w dist/repaired/ \
|
||||
--exclude libplatform.so \
|
||||
--exclude libregister.so \
|
||||
--exclude libge_common_base.so \
|
||||
--exclude libc10.so \
|
||||
--exclude libc_sec.so \
|
||||
--exclude "libascend*.so" \
|
||||
--exclude "libtorch*.so"
|
||||
done
|
||||
rm -f dist/*.whl
|
||||
mv dist/repaired/*.whl dist/
|
||||
rmdir dist/repaired
|
||||
ls dist
|
||||
|
||||
- name: Verify automatic platform tags
|
||||
run: |
|
||||
cd dist
|
||||
for wheel in *.whl; do
|
||||
echo "verification file: $wheel"
|
||||
auditwheel show "$wheel"
|
||||
done
|
||||
|
||||
- name: Archive wheel
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: vllm-ascend-${{ matrix.os }}-py${{ matrix.python-version }}-wheel
|
||||
path: dist/*
|
||||
|
||||
- name: Release
|
||||
if: startsWith(github.ref, 'refs/tags/')
|
||||
|
||||
Reference in New Issue
Block a user