Wrong ETA for resumed trainings

leggedrobotics / rsl_rl

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

Other

505 stars 156 forks source link

Wrong ETA for resumed trainings #10

Open E-Mans-Application opened 1 year ago

E-Mans-Application commented 1 year ago

Hello,

I noticed the ETA is wrong when the training is resumed from a previous run.

I guess you should replace locs['it'] (two occurrences) with locs['it'] - self.current_learning_iteration in the computation of the ETA (in runners/on_policy_runner.py, lines 209/210)