Convert aten.reshape to ttnn.reshape

[jdh8]

Ticket

• Subproblem of Failed to lower some tensor manipulation ops from aten to ttnn tt-metal#12853
• Resolves aten.view not converted? #173

Problem description

Successfully convert aten.reshape to ttnn.reshape

This PR paves the way for #148, too.

What's changed

• Convert aten.reshape to ttnn.reshape
• Convert aten.view to ttnn.reshape
• Test conversion from torch.reshape to ttnn.reshape

[ayerofieiev-tt]

Test started here
https://github.com/tenstorrent/pytorch2.0_ttnn/actions/runs/10836182163

[ayerofieiev-tt]

Fails during models test

path	passed	error	xfailed	subtotal
models/perceiver_io/test_perceiver_io.py	🟢		🔴	2
models/resnet/test_resnet.py	🟢	🔴		2
models/resnet50/test_resnet50.py	🟢		🔴	2
models/roberta/test_roberta.py	🟢		🔴	2
TOTAL	4	1	3	8

[jdh8]

I just ran pytest tests/models/resnet/test_resnet.py locally and got one error as well. However, the error message was rather wordy. Is there a way to extract failing input variations?

[ayerofieiev-tt]

@jdh8 can you check this?
Tensor<[1, 512, 1, 1]> self = ?, List[int]<> size = [1, 512]

[ayerofieiev-tt]

Keep in mind that right now compiler Tilizes tensors before most TT-NN operations.
After the op is tilized you can't reshape it to random shape - two last dims have to be divisible by tile size (32x32).
If it is divisible - you can proceed with it, but else you have to move to RowMajor and then reshape

[jdh8]

This case passes on its own. Maybe we need to check the next op or the synergy between reshape and the next op?