You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You can try with distributed data parallel: https://pytorch.org/tutorials/intermediate/ddp_tutorial.html. However, since different nav rollouts have different length and PyTorch DDP uses synchronized update, the speed of each step would be equal the the slowest one.
How to use multiple GPUs for training? Do I need to write the function of setting multiple GPUs into the model?
The text was updated successfully, but these errors were encountered: