Describe the bug
Even after fixing Reshape RM issue, it seems embedding is still not seeing the tensor as RM:
TT_FATAL @ ../ttnn/cpp/ttnn/operations/embedding/device/embedding_device_operation.cpp:25: a.get_layout() == Layout::ROW_MAJOR
Ex input:
{'weight_shape': [81, 768], 'indices_shape': [1, 24], 'padding_idx': 1}
To Reproduce
Can filter the cloud kibana by data_movement.embbedding.embedding_pytorch2 or run the gh workflow accordingly
Expected behavior
Should either convert to RM or support tilized
Describe the bug Even after fixing Reshape RM issue, it seems embedding is still not seeing the tensor as RM: TT_FATAL @ ../ttnn/cpp/ttnn/operations/embedding/device/embedding_device_operation.cpp:25: a.get_layout() == Layout::ROW_MAJOR
Ex input: {'weight_shape': [81, 768], 'indices_shape': [1, 24], 'padding_idx': 1}
To Reproduce Can filter the cloud kibana by data_movement.embbedding.embedding_pytorch2 or run the gh workflow accordingly
Expected behavior Should either convert to RM or support tilized