Justin Ngo

Hi @vivekkhandelwal1, @sjarus and I are also tracking fx_importer_tosa regressions from this update. Here are some additional tests that failed on fx_importer_tosa that haven't been included in your list. I...

I'm not sure if this has happened before the LLVM integration, but as of now, that is the problem.

If I recall correctly, tosa and onnx_tosa targets are not included in the CI; only the fx_importer_tosa target is included. Because those 2 targets are not checked by CI, I...

@vivekkhandelwal1 Can you take a look at this please? Is this a known issue, and are there any fixes? cc: @sjarus

I believe this [LLVM commit](https://github.com/llvm/llvm-project/pull/150561) has fixed this issue. @vivekkhandelwal1 can you help me bump LLVM in Torch-MLIR to get this change in place please? Thank you in advance!

> > I believe this [LLVM commit](https://github.com/llvm/llvm-project/pull/150561) has fixed this issue. [@vivekkhandelwal1](https://github.com/vivekkhandelwal1) can you help me bump LLVM in Torch-MLIR to get this change in place please? Thank you in...

Hi, the element types of your input don't match each other, which is [not allow as when lowering `torch.aten.mm` to TOSA](https://github.com/llvm/torch-mlir/blob/main/lib/Conversion/TorchToTosa/TorchToTosa.cpp#L1478-L1480).

Hi @HemKava, `scaled_dot_product_attention` doesn't have a legalization from Torch to TOSA yet, but you are welcomed to contribute one and add me or @sjarus as a reviewer.

I'm not entirely sure, but I believe that there is none. Others might have better suggestions though.

Hi @HemKava, I was looking at some older PR, and I found this one that was trying to add a decomposition for `ScaledDotProductAttention` a while back: https://github.com/llvm/torch-mlir/pull/3461. It was not...