### Isuru - Pytorch PRs

- [add decomposition for frexp](https://github.com/pytorch/pytorch/pull/119217) - ~~Memory error?~~ Waiting for CI
- [register torch.return_types in torch.fx._pytree](https://github.com/pytorch/pytorch/pull/120027) - Waiting for review from maintainers of pytree
- [119284](https://github.com/pytorch/pytorch/pull/119284) - Ready to merge

- [Use a dtype property in torch inductor nodes](https://github.com/pytorch/#114774pytorch/pull/119227) - Waiting for CI

- [Decomposition for upsample_linear{1d, 3d}](https://github.com/pytorch/pytorch/pull/114774) - Needs a rebase and merge

- [Improve perf of avg_pool2d by rewriting as a reduction](https://github.com/pytorch/pytorch/pull/116756) - Figure out slowdown in benchmark
- [Add lowering for avg_pool{1, 3}d](https://github.com/pytorch/pytorch/pull/116085) - Ready to merge

- [Add decompositions for copy variants of view ops](https://github.com/pytorch/pytorch/pull/119889) - test failures

- [Add decompositions for avg_poolnd](https://github.com/pytorch/pytorch/pull/116418) - Decomposition slower than lowering
- [Add aten._unsafe_masked_index](https://github.com/pytorch/pytorch/pull/116491) - Test failures
- [Increase unroll_reductions_threshold](https://github.com/pytorch/pytorch/pull/116960) - Run benchmark

- [Add out wrappers to some decompositions](https://github.com/pytorch/pytorch/pull/115437) - Test failures

- [Fix test_decomp test for ops with py_impl(CompositeImplicitAutograd)](https://github.com/pytorch/pytorch/pull/116832) - test failure with requires grad tensors

- [Add decomposition for channel_shuffle](https://github.com/pytorch/pytorch/pull/118775) - Needs work
- [Fix FakeTensorUpdater logic for updating fake tensors](https://github.com/pytorch/pytorch/pull/116168) - benchmark test `sam` fails