ttnn.split

ttnn.split = FastOperation(python_fully_qualified_name='ttnn.split', function=<ttnn._ttnn.operations.data_movement.split_t object>, preprocess_golden_function_inputs=None, golden_function=None, postprocess_golden_function_outputs=None, is_cpp_operation=True, is_experimental=False)

Returns a tensor that is in num_splits ways on dim.

Equivalent pytorch code:

output_tensor = torch.split(input_tensor, 2, 1)

:param * input_tensor: Input Tensor. :param * split_size: Single chunk size or list of chunk sizes. Output may be smaller if dim not evenly divisible. :type * split_size: Union[int, list[int]] :param * dim2: Dim to split. Defaults to 0.

:keyword * memory_config: Memory Config of the output tensor