Multi-CPU parallel training

[Taylor-Liu]

Hi, Wenlong. Thanks for sharing your code.

When I reproduced your code, I found that only one CPU was used for training. The training speed was a bit slow. Can your code train with multiple CPUs in parallel? I can't find the corresponding option in your code configuration.

[huangwl18]

Thanks for the interest in the paper! The code is not configured for multi-CPU training. Only vectorized environments (from OpenAI baselines) and torchfold (dynamic batching) are currently used for speedup in training.