Quinn Dawkins

This was my concern as well, but some of the cases we are starting to see where we need support for aten.view were difficult with the previous approach so I...

I dumped the E2E cases into a file for IR unit tests, but without any filecheck stuff just so that we can see the IR. I can add the lowerings...

> Can you create small IR unit tests for all the cases that need to be handled so we can look for a solution together? Generally the `tensor.reshape` op is...

> I think the collapse to 1d + expand approach is a good one that is quite general. We should only use it when there is no other way to...

> tensor.reshape is at basically the same abstraction level as the op we are lowering here, so it is definitely easier to lower to. But then somebody else has to...

Ah I didn't realize that we didn't support dynamic size-1 broadcast. @vivekkhandelwal1 can you provide a compact test case that shows the case we need for v-diffusion? If we don't...

The current blocker for this model is adding support for multiple indexing tensors in aten.index.Tensor with dynamic dimensions. I opened an issue discussing adding support for this case at #1226.

Ah sorry I'm not up to date on the former error. @vivekkhandelwal1 can comment on whether that has been unblocked.

It looks like this was inadvertently fixed by #1082, however there is no test for this. Do we want to add one?

This upstream PyTorch PR includes the shape functions: https://github.com/pytorch/pytorch/pull/79194/files. The main problem function is this one for `aten.split.Tensor`: ```python def split_tensor(self: List[int], split_size: int, dim: int = 0) -> List[List[int]]:...