[Perf] Add new npu_fused_infer_attention_score op to improve perfomance in splitfuse cases and resolve long-seq mask problems #12563
Triggered via pull request
September 16, 2025 12:31
Status
Failure
Total duration
5m 59s
Artifacts
–
vllm_ascend_test.yaml
on: pull_request
Matrix: singlecard e2e test - light
Matrix: unit test
Matrix: multicard e2e test - light
Annotations
10 errors