currently, main_ddpg uses timesteps and main_cmpc uses episodes. This is problematic b/c it changes the meaning of --save_every and --render_every, flags shared between the two.
Currently, for main_ddpg, *_every means "every N samples (which is every timesteps_per_sample * N timesteps).
For main_cmpc *_every means every N episodes.
Both are not in a consistent, desirable state. It should be that:
main_cmpc uses timesteps as well
*_every refers to the number TIMESTEP between saves/renders/etc. Currently neither main_ddpg or main_cmpc do this.
Then the save_every default should be changed accordingly.
currently, main_ddpg uses timesteps and main_cmpc uses episodes. This is problematic b/c it changes the meaning of
--save_every
and--render_every
, flags shared between the two.Currently, for main_ddpg,
*_every
means "everyN
samples (which is everytimesteps_per_sample * N
timesteps).For main_cmpc
*_every
means every N episodes.Both are not in a consistent, desirable state. It should be that:
main_cmpc
uses timesteps as well*_every
refers to the number TIMESTEP between saves/renders/etc. Currently neither main_ddpg or main_cmpc do this.Then the save_every default should be changed accordingly.