Python train.py

[xjuzyw0319]

Thank you very much for your work and for making your data public. I recently encountered a problem when reproducing your code. When I ran the code python train.py, the following data appeared:

Updates 180, num timesteps 86880, FPS 82
Last 100 training episodes: mean/median reward -6.2/-9.1, min/max reward -21.0/26.0.

Is visualization supported during training, or is it only possible when executing python test.py? Thank you again.

[Shuijing725]

The output looks normal. If you set the num_processes in arguments.py to 1 and set sim.render to True, you'll be able to visualize training. Visualization is not supported if num_processes > 1.

[xjuzyw0319]

Hello, I encountered two problems when reproducing your code: First, I found that the FPS was only 6 at the beginning, and as the training progressed, the PFS was only 2. Can I modify your default parameter --num-processes to increase the FPS? Second, how long does it take for you to finish the training? Thank you very much for your reply, and finally I wish you a smooth work.