Skip to content
test-full

[bugfix][torchair] fix wasted NPU memory buffer allocation for quantized deepseek with unquantized MTP layer #616

Sign in to view logs

[bugfix][torchair] fix wasted NPU memory buffer allocation for quantized deepseek with unquantized MTP layer

[bugfix][torchair] fix wasted NPU memory buffer allocation for quantized deepseek with unquantized MTP layer #616

Triggered via pull request September 20, 2025 14:38
@linfeng-yuanlinfeng-yuan
labeled #3068
linfeng-yuan:fix_torchair_redundant_process_group
Status Cancelled
Total duration 9s
Artifacts –

vllm_ascend_test_full.yaml

on: pull_request
changes
5s
changes
Matrix: multicard e2e test - full
Matrix: singlecard e2e test - full
Fit to window
Zoom out
Zoom in

Annotations

1 error
test-full
Canceling since a higher priority waiting request for test-full-refs/pull/3068/merge exists