[rank2]: AttributeError: module 'torch.distributed' has no attribute 'nn'
zimenglan-sysu-512 opened this issue · comments
when run the train.py, encounters the problem: [rank2]: AttributeError: module 'torch.distributed' has no attribute 'nn'
.
it can be solved by adding this line: import torch.distributed.nn as nn_dist
in model/model_longclip.py
file.
by the way, @beichenzbc can u share the training log file, i want to see whether can reproduce the result.
thanks