[2/N][refactor] torchair deepseek mla backend refactor · vllm-project/vllm-ascend@5c278a8

Triggered via pull request August 20, 2025 15:46

synchronize #2459

Status Failure

Total duration 5m 18s

Artifacts –

vllm_ascend_test.yaml

on: pull_request

Matrix: singlecard e2e test

Matrix: unit test

Matrix: multicard e2e test

5 errors

Process completed with exit code 1.

Ruff (F841): vllm_ascend/attention/mla_v1.py#L910

vllm_ascend/attention/mla_v1.py:910:9: F841 Local variable `MAX_O_PROJ_PREFETCH_SIZE` is assigned to but never used

Ruff (F841): vllm_ascend/attention/mla_v1.py#L858

vllm_ascend/attention/mla_v1.py:858:13: F841 Local variable `prefill_q_nope` is assigned to but never used

Ruff (F841): vllm_ascend/attention/mla_v1.py#L838

vllm_ascend/attention/mla_v1.py:838:9: F841 Local variable `prefill_hs` is assigned to but never used

Ruff (F841): vllm_ascend/attention/mla_v1.py#L227

vllm_ascend/attention/mla_v1.py:227:13: F841 Local variable `num_spec_tokens` is assigned to but never used