[bugfix][torchair] fix wasted NPU memory buffer allocation for quantized deepseek with unquantized MTP layer #617
Triggered via pull request
September 20, 2025 14:38
Status
Success
Total duration
1h 28m 33s
Artifacts
–
vllm_ascend_test_full.yaml
on: pull_request
changes
7s
Matrix: multicard e2e test - full
Matrix: singlecard e2e test - full