Stuck when using multiple GPU

[nlpmc]

Hi~ Thanks for your code.
When I try to use four GPUs to train the model as you mentioned in the Readme, the training process stuck. I find that the original training command is -visible_gpus 0,1,2,3 -gpu_ranks 0,1,2,4. And when I change it to -visible_gpus 0,1,2,3 -gpu_ranks 0,1,2,3, this problem get fixed.

[nlpyang]

Thanks, the readme has been fixed.