多机多卡GPU分布式训练

Traceback (most recent call last):
  File "train_erfnet_cluster.py", line 714, in
    os.environ['MASTER_ADDR'] = os.environ['PAI_HOST_IP_worker_0']
  File "/opt/conda/lib/python3.7/os.py", line 681, in __getitem__
    raise KeyError(key) from None
KeyError: 'PAI_HOST_IP_worker_0'
 

你可能感兴趣的:(分布式,p2p,cnn)