Skip to main content
openrlbenchmark
Projects
sbx
Reports
SBX SAC - influence of learning rate
Log in
Sign up
Share
Comment
1 star
SBX SAC - influence of learning rate
lr=3e-4 vs lr=1e-3
Antonin RAFFIN
Created on January 11
|
Last edited on January 11
Comment
HalfCheetah-v4
time/fps
time/fps
200k
400k
600k
800k
1M
global_step
1000
2000
3000
algo: sac, saved_hyperparams.learning_rate: 0.001
TD3 RL Zoo 2.3.0
algo: sac
TD3 RL Zoo 2.2.1
eval/mean_reward
eval/mean_reward
200k
400k
600k
800k
1M
global_step
0
2000
4000
6000
8000
10000
12000
algo: sac, saved_hyperparams.learning_rate: 0.001
TD3 RL Zoo 2.3.0
algo: sac
TD3 RL Zoo 2.2.1
TD3 RL Zoo 2.3.0
3
TD3 RL Zoo 2.2.1
10
Ant-v4
TD3 RL Zoo 2.3.0
3
TD3 RL Zoo 2.2.1
10
Hopper-v4
TD3 RL Zoo 2.3.0
3
TD3 RL Zoo 2.2.1
10
Walker2d-v4
TD3 RL Zoo 2.3.0
3
TD3 RL Zoo 2.2.1
10
Swimmer-v4
TD3 RL Zoo 2.3.0
3
TD3 RL Zoo 2.2.1
10
Add a comment