Accelerating PyTorch Transformers with Intel Sapphire Rapids - part 1