ttnn.eq incorrect output for uint16 tensors

[sraizada-tt]

ttnn.eq considers 1==0 True for uint16 tensors; unit test:

import ttnn 
from ttnn import ReplicateTensorToMesh, ConcatMeshToTensor
import torch 
def test_matmul(device_mesh):
    mask = ttnn.from_torch(
            torch.zeros(1, 1, 32, 32),
            dtype=ttnn.uint16,
            layout=ttnn.TILE_LAYOUT,
            device=device_mesh,
            mesh_mapper=ReplicateTensorToMesh(device_mesh),
        )
    output = ttnn.from_torch(
            torch.ones(1, 1, 32, 32),
            dtype=ttnn.uint16,
            layout=ttnn.TILE_LAYOUT,
            device=device_mesh,
            mesh_mapper=ReplicateTensorToMesh(device_mesh),
        )
    output_reduced = ttnn.eq(output, mask) 

    assert torch.all(ttnn.to_torch(output_reduced, mesh_composer=ConcatMeshToTensor(device_mesh, dim=0))==0 ).item()

[arakhmati]

Re-assigning this issue to @umadevimcw

[ruthreshx]

Hi @sraizada-tt , Could you please add the instructions in ticket to recreate the failure in VM ? Thanks in advance!

[sraizada-tt]

lowering priority because we have a workaround

[KalaivaniMCW]

will be revisited after TTNN migration