Stop Using torch.cuda! Unified Accelerator API in PyTorch!
#ai #pytorch #deeplearning
In this video, we discuss the significance of "accelerators" for deep learning. We also discuss why using torch.cuda is not optimal as we have more accelerators. Then, we discuss a new PyTorch API: torch.accelerator and show how you can write device-agnostic code in one line!
More information: https://docs.pytorch.org/docs/stable/...