def _test(tensor): return tensor.to(torch.int16).view(torch.bfloat16) test = torch.compile(_test) test(torch.zeros((8, 8), dtype=torch.int8, device="cuda")) It fails ...
In torchserve we can spawn multiple python processes for the same pytorch model to scale inference. We've recently added support for torch.compile() but one major problem it brings is that I need to ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results