@PyTorch
Interested in torch.compile? @ezyang’s new post walks through the ways to use torch.compile including improving training and inference efficiency as well as optimization steps for both memory and performance with torch.compile https://t.co/a8cOSKJM8I