christianhidber / easyagents

Reinforcement Learning for Practitioners.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

add num_iterations_between_plots argument

christianhidber opened this issue · comments

Currently plots are typically redrawn after each training iteration. For algorithms like Dqn with few episodes steps but many iterations this yields too high an overhead with respect to the compute time used for training.

By adding a 'num_iterations_between_plots' we can control the plot redraw frequency in the same way we control the policy evaluation frequency using 'num_iterations_between_eval'. If set to None a (potentially good) default could be 1/3 of 'num_iterations_between_eval'.