This repository was archived by the owner on Apr 19, 2026. It is now read-only.
This repository was archived by the owner on Apr 19, 2026. It is now read-only.
sinusoid_table = torch.from_numpy(sinusoid_table).to("cuda:{}".format(self.args.local_rank)).float().unsqueeze(0) so slow #23
Why this step is so slow?