search

xbpeng / DeepMimic

Motion imitation with deep reinforcement learning.
https://xbpeng.github.io/projects/DeepMimic/index.html
MIT License
2.27k stars 485 forks source link

What are the time_lim arguments to mpi_run.py? #183

Closed ajlangley closed 2 years ago

ajlangley commented 2 years ago

What do the following arguments mean?

--time_lim_min --time_lim_max --time_lim_exp --time_end_lim_min --time_end_lim_max --time_end_lim_exp

Thanks for the help!

xbpeng commented 2 years ago

We sometimes train using a curriculum, where we start with short episodes and then progressively make the episodes longer as training processes. time_lim_min and time_lim_max define the starting range for the episode lengths. The episode length will be sampled between [min, max]. Similarly time_end_lim_min and time_end_lim_max is the episode lengths at the later stage of training. These values are annealed between the start and end values over a number of samples defined by (anneal_samples). You can ignore the lim_exp parameters.

ajlangley commented 2 years ago

Great, that clears it up. Thanks so much for the quick response!