transformers

c300ce78 - feat(tensor_parallel): split up the tp slice(s) computation and introduce a method to navigate slices over the flattened tensor view