disable empirical normalizer updates on resume training

leggedrobotics / rsl_rl

Fast and simple implementation of RL algorithms, designed to run fully on GPU.

Other

505 stars 156 forks source link

Open tasdep opened 2 months ago

tasdep commented 2 months ago

This completely disables the updates if "resume" is in the agent_cfg and True.

Should we check how many iterations/updates were done previously and only disable if over a threshold?

pascal-roth commented 1 month ago

@nikitardn should we merge this?