Parallel tensors work with the graph fusion and kernel optimizations to help accelerate inference. PyTorch 2.1 is coming Ganti emphasized that IBM’s efforts to accelerate PyTorch for inferencing ...