4 ms·
It is a huge improvement to PPO because you don’t need a separate critic model which cuts memory costs in half and stabilizes training.
by janalsncm 3mo ago
It is a huge improvement to PPO because you don’t need a separate critic model which cuts memory costs in half and stabilizes training.
- porridgeraisin 3mo agoYes, but monte carlo estimating the critic model is not new.