Results of AdaBelief and Adam on Reinforcement Learning with SAC (Soft Actor Critic)

Dependencies

• prfl
• gym
• torch

How to run

sh run_adabelief_walker2d.sh sh run_adam_walkerd.sh

You can change --env Walker2d-v2 to --env HalfCheetah-v2 for different tasks

Hyper-parameters

eps for AdaBelief is 1e-12, other parameters are default as in adabelief-pytorch==0.1.0

Optimizer	lr	beta1	beta2	epsilon	weight_decay	weight_decouple	rectify	fixed_decay	amsgrad
Adam	1e-3	0.9	0.999	1e-8	0.0	-	-	-	-
AdaBelief	1e-3	0.9	0.999	1e-12	0.0	True=False	True	False	False

Results