Sean Silva
Can you list which ops are missing so we can add them to the e2e todo list? CC @ramiro050 who is tracking that TODO list.
Looks like the ops were added.
We don't want to randomly run the shape refinment pipeline too many times. Ideally we will have it in one good place in the pass pipeline. One idea is to...
We added canonicalization for aten.add.Tensor and aten.int.Tensor so this shouldn't be an issue anymore.
> I've tried to reproduce it as a test case but I wasn't able to. What is the original issue? Is this memref.copy issue related?
Can you reduce the test case to a single pass and a small (
The input to the function is a list of tensors, rather than a single tensor -- we don't currently support lists of tensors as inputs. Generally speaking, this kind of...
I'm going to close this, since I think Torch-MLIR is working as intended -- we don't support lists of tensors as input.
This is due to having two tensors in the program that share storage, but aren't the exact same tensor (e.g. they are two different views of the same underlying tensor)...
So is the blocker the "Unhandled tensor that shares storage with another tensor" error or aten.index.Tensor?