torch.from_tensor() bug fix

train.py

@@ -225,7 +225,7 @@ def train(hyp, opt, device, tb_writer=None):
             if rank != -1:
                 indices = torch.zeros([dataset.n], dtype=torch.int)
                 if rank == 0:
-                    indices[:] = torch.from_tensor(dataset.indices, dtype=torch.int)
+                    indices[:] = torch.tensor(dataset.indices, dtype=torch.int)
                 dist.broadcast(indices, 0)
                 if rank != 0:
                     dataset.indices = indices.cpu().numpy()