Ramiro Leal-Cavazos
Ramiro Leal-Cavazos
> meet another error about chunk, when I use https://github.com/llvm/torch-mlir/pull/1022 @ramiro050 The PR is not yet in a working state.
Can you rebase your branch to fix the conflicts so that the CI can run?
> I can't reproduce the CI failure locally with the following environments The CI is failing because one of the e2e tests is failing on an assertion. For some reason...
This is currently failing because there is no support for broadcasting the `bias` tensor: https://github.com/llvm/torch-mlir/blob/df0b1e77a475cffcd758c99efea7f9ba64ae060a/lib/Conversion/TorchToLinalg/Linear.cpp#L493-L497 @Shukla-Gaurav, do you know if after your patch https://github.com/llvm/torch-mlir/pull/862, this will work?
The error you're getting seems to be caused by the fact that the decomposition currently expects the input tensor to have rank >= 2: https://github.com/llvm/torch-mlir/blob/4ef6e69ed49f309aa70ea40f12bf7488cd8e3434/lib/Dialect/Torch/Transforms/DecomposeComplexOps.cpp#L2078 However, this does not seem...
Waiting for https://github.com/pytorch/pytorch/pull/83357 to land.
@vivekkhandelwal1 @Shukla-Gaurav, PTAL
Hi @albertdmath, Yes, adding this in `DecomposeComplexOps.cpp` seems like a good strategy to me. A good way to get familiar with how to do decompositions in torch-mlir is to look...
> Fascinating. I always though that getUsers was deduped. Yeah, I thought it was weird too. > Did you actually run into this in a real test? I would have...
Can you rebase your branch to resolve the conflicts so that the CI can run?