Ramiro Leal-Cavazos comments

Results 77 comments of


                                            Ramiro Leal-Cavazos

(WIP) Refactor lowering of aten.view op to the tensor dialect

> I would like to factorize this PR into separate parts. Sounds good. I'm all for small PRs 🙂 > I'm relatively new to working on github -- I wonder,...

Fix return tuple with one element

See: https://github.com/llvm/torch-mlir/issues/1471#issuecomment-1273913856

Decompose aten.index_put_impl to aten.index_put.hacked_twin

Let me take a look at the xfails and see if I find anything

Decompose aten.index_put_impl to aten.index_put.hacked_twin

The issue is that the TorchToTMTensor pattern expects the op to only have 2 tensors in the indices list of tensors (for example: `(None, None, tensor1, tensor2)` would be valid)....

Add missing early return.

Just realized that this was addressed in a PR made this morning: https://github.com/llvm/torch-mlir/pull/2409

[TORCH-LINALG] Add recomposition for select+copy_

Hi @bilibiliGO283, thanks for the PR! I will take a look at it tomorrow.

Error while conversion of .mlir to .vmfb

This seems like an IREE issue. I think it would be better to raise an issue in the IREE repo, so that people with the right expertise can look at...

[TOSA] failure compiling model dividing tensor by constant (for the trace method)

@eric-k256, do you know what's going on here? The IR seems fine to me. Maybe `reciprocal` does not work for zero-rank tensors.

Error when trying to emit MLIR for Training MNIST kernel

Answered on discord: https://discord.com/channels/636084430946959380/742573221882364009/1139614233978474646

Failing to convert the model to MLIR using torch_mlir.compile

Both errors seem to be from PyTorch, not torch-mlir. When you call `torch_mlir.compile`, torch-mlir calls `torch.jit.script` on the module before importing it into torch-mlir. The first error is due to...