Normalize to positive axis for comparison on GEMM#26152
Conversation
…ation.cc Co-authored-by: Edward Chen <18449977+edgchen1@users.noreply.github.com>
|
/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline |
|
Azure Pipelines successfully started running 4 pipeline(s). |
|
@edgchen1 could you help to check the CI runner status here? Thanks! |
|
@qti-yuduo Have you agreed to the CLA? It looks like the "license/cla" check is pending. |
I think so? Unless it's changed and I need to do it again. I've commited changes in the past: https://github.com/microsoft/onnxruntime/pulls?q=is%3Apr+author%3Aqti-yuduo+is%3Aclosed |
Ok, let me try to close and re-open the PR. |
|
/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline |
|
Azure Pipelines successfully started running 4 pipeline(s). |
GEMM has 2D input, when given `axis=-1` in DequantizeLinear node attributes. The check `axis == expected_axis` would fail. However, it sematically the same (-1 and 1 for 2D GEMM's input). Normalize the check so we can still perform the quantization for this case. --------- Co-authored-by: Edward Chen <18449977+edgchen1@users.noreply.github.com>
GEMM has 2D input, when given `axis=-1` in DequantizeLinear node attributes. The check `axis == expected_axis` would fail. However, it sematically the same (-1 and 1 for 2D GEMM's input). Normalize the check so we can still perform the quantization for this case. --------- Co-authored-by: Edward Chen <18449977+edgchen1@users.noreply.github.com>
2 participants
GEMM has 2D input, when given
axis=-1in DequantizeLinear node attributes. The checkaxis == expected_axiswould fail. However, it sematically the same (-1 and 1 for 2D GEMM's input). Normalize the check so we can still perform the quantization for this case.