👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

I suspect it has something to do with running the test case, but I can't reproduce the test case at my local env

all tests passed
and eagle failed but no OOM
FAILED tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py::test_eagle_correctness[eagle] - AssertionError: assert 6 > 6
FAILED tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py::test_eagle_correctness[eagle3] - AssertionError: assert 6 > 6

Let's see is it OK in this PR (only run all spec_decode test)

Failed as expected: https://github.com/vllm-project/vllm-ascend/actions/runs/18081163582/job/51444630922?pr=3250

Let's only run eagle then

pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py

only run eagle, PASSED: https://github.com/vllm-project/vllm-ascend/actions/runs/18081477286/job/51445359952?pr=3250

npu-smi info
pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_torchair_correctness.py
npu-smi info
pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py

Passed: https://github.com/vllm-project/vllm-ascend/actions/runs/18081893242/job/51446420430?pr=3250

This pull request has conflicts, please resolve those before we can evaluate the pull request.