Mengwei Liu comments

Results 60 comments of


                                            Mengwei Liu

NNPACK with Windows support

In case people are still looking for windows support, I just submit a PR #201

[MPS - DRAFT] Add support for slice_scatter; enable index_put

I see the metal kernel compilation path is not enabled. Any reason why indexing ops require metal kernels and any plan to enable metal kernel path? Asking because I'm thinking...

Refactor export_llama_lib.py

merged.

Duplicate registration of quantiation operators, e.g. quantized_decomposed::embedding_byte.out

Yeah I can take a look

Duplicate registration of quantiation operators, e.g. quantized_decomposed::embedding_byte.out

I'm iterating on this PR: https://github.com/pytorch/executorch/pull/3382. This should allow us to select the ops to be registered in `quantized_ops_aot_lib`

Duplicate registration of quantiation operators, e.g. quantized_decomposed::embedding_byte.out

> Thanks, commented on a small issue on the PR. > > I also find the flow with CMAKE is probably not tested even in ciflow/trunk at the moment. I've...

Duplicate registration of quantiation operators, e.g. quantized_decomposed::embedding_byte.out

Sent out https://github.com/pytorch/executorch/pull/3449 to fix it

[torchgen] Small fix for empty yaml file edge case

@pytorchbot merge

[torchgen] Small fix for empty yaml file edge case

@larryliu0820 has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D42729435).

[torchgen] Small fix for empty yaml file edge case

@larryliu0820 has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D42729435).