Support for creating a 0-dim tensor when a single scalar value or shape [ ] is given

[KalaivaniMCW]

Requirements from pytorch tracing sweeps https://github.com/tenstorrent/pytorch2.0_ttnn/blob/main/docs/operations/aten.add.Tensor.md

Case 1: when input parameters are such :

Tensor<[1, 1, 1024]> self = ?,
Tensor other = 1.0

The tensor other is expected to be created as a 0-dim tensor with value 1.0

Case 2: when input parameters are such :

Tensor<[]> self = ?,
Tensor other = 1

The tensor self is expected to be created as a 0-dim tensor The tensor other is expected to be created as a 0-dim tensor with value 1.0

Case:3 Empty Tensor or elements [0,1]

[KalaivaniMCW]

Adding some examples regarding [ ], below shapes are taken from ttnn.add sweep for pytorch2 When tested in torch , Torch: {'input_shape': {'self': [], 'other': [0, 1]},

Torch: {'input_shape': {'self': [], 'other': [1, 1, 768]}

[VirdhatchaniKN]

Adding some examples regarding [ ], below shapes are taken from ttnn.divide sweep for pytorch2 : https://github.com/tenstorrent/tt-metal/issues/14733 When tested in torch ,

• {"shape": [1, 12, 7, 7], "other": []}
• {"shape": [1, 16, 1, 6], "other": []}