[MLIR][TORCH] Add missing ops in RefineTypes.cpp

[vivekkhandelwal1]

Signed-Off By: Vivek Khandelwal[email protected]

[AmosLewis]

LGTM

[AmosLewis]

Related Issue: nod-ai/SHARK-Studio#494

[ramiro050]

This should not be needed. These ops were removed from here to make sure the custom op code works:

torch-mlir/python/torch_mlir/dialects/torch/importer/jit_ir/build_tools/abstract_interp_lib_gen.py

Line 65 in 3260a1e

def aten〇tanh〡dtype(self_rank: int, self_dtype: int) -> int:

From looking at the issue linked, it seems that the problem is that some places in the code base call getDtype without first checking hasDtype. The solution should be to add the right checks

[vivekkhandelwal1]

This should not be needed. These ops were removed from here to make sure the custom op code works:

torch-mlir/python/torch_mlir/dialects/torch/importer/jit_ir/build_tools/abstract_interp_lib_gen.py

Line 65 in 3260a1e

def aten〇tanh〡dtype(self_rank: int, self_dtype: int) -> int:

From looking at the issue linked, it seems that the problem is that some places in the code base call getDtype without first checking hasDtype. The solution should be to add the right checks

Thanks for letting me know.