mirror of https://github.com/llvm/torch-mlir
7f9f99c6f8
The existing TorchToTosa lowering logic for `torch.aten.avg_pool2d` doesn't handle some unsupported properties well, leading to a silent wrong answer(SWA) when we go through `torch-backend-to-tosa-backend-pipeline.` For instance, with the existing TOSA avgpool2d specification, we can not represent `count_include_pad` and `divisor_override,` so during TorchToTosa lowering, we silently ignore these properties which leads to SWA in some cases—the fix captured in this change errors for unsupported scenarios. For details on `count_include_pad` and `divisor_override,` please see the below link. https://pytorch.org/docs/stable/generated/torch.nn.AvgPool2d.html --------- Co-authored-by: Hanumanth Hanumantharayappa <hhanuman@ah-hhanuman-l.dhcp.mathworks.com> |
||
|---|---|---|
| .. | ||
| TorchConversionToMLProgram | ||
| TorchOnnxToTorch | ||
| TorchToArith | ||
| TorchToLinalg | ||
| TorchToSCF | ||
| TorchToStablehlo | ||
| TorchToTensor | ||
| TorchToTosa | ||