Inception_v3 for multi-GPU

1

I try to train inception_v3 in a multi-gpu environment, but it failed like:

Traceback (most recent call last):
  File "/home/ljy/anaconda3/lib/python3.6/site-packages/IPython/core/interactiveshell.py", line 2885, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-22-139a04d1be98>", line 1, in <module>
    model = torch.nn.parallel.DistributedDataParallel(model)
  File "/home/ljy/anaconda3/lib/python3.6/site-packages/torch/nn/parallel/distributed.py", line 103, in __init__
    dist.broadcast(p, 0)
  File "/home/ljy/anaconda3/lib/python3.6/site-packages/torch/distributed/__init__.py", line 197, in broadcast
    "collective only supported in process-group mode"
AssertionError: collective only supported in process-group mode

The code is as follows (pytorch 0.3.1):

import torchvision.models as models

model = models.inception_v3()
model.cuda()
model = torch.nn.parallel.DistributedDataParallel(model)
2

Sorry , Problem solved. But I don’t know how to close the issue.
For multi-GPU situation, I should use

model.cuda()
model = torch.nn.DataParallel(model)