Modified distributed adamw's local_rank and world_size arguments

DeepX-inc / machina

Control section: Deep Reinforcement Learning framework

MIT License

279 stars 43 forks source link

Modified distributed adamw's local_rank and world_size arguments #239

Closed iory closed 5 years ago

iory commented 5 years ago

This PR modified the arguments of distributed adamw. local_rank and world_size 's default value is None and automatically set dist.get_rank() and dist.get_world_size() values. I also add the test code of DistributedAdamW and DistributedSGD.

rarilurelo commented 5 years ago

Thanks!