Skip to main content
costa-huang
Projects
cleanRL
Reports
ddpg jax debug
Log in
Sign up
Share
Comment
Star
Share
Comment
Star
ddpg jax debug
Costa
Created on June 23
|
Last edited on June 27
Comment
HalfCheetah-v2
HalfCheetah-v2
200k
400k
600k
800k
Step
0
2000
4000
6000
8000
10000
Episodic Return
HalfCheetah-v2
HalfCheetah-v2
20
40
60
80
Time (minutes)
0
2000
4000
6000
8000
10000
Episodic Return
GPU utilization
GPU utilization
20
40
60
80
Time (minutes)
0
20
40
60
80
SPS
SPS
200k
400k
600k
800k
Step
500
1000
1500
2000
2500
Critic Loss
Critic Loss
200k
400k
600k
800k
Step
50
100
150
200
250
300
Actor Loss
Actor Loss
200k
400k
600k
800k
Step
-1000
-800
-600
-400
-200
0
CleanRL's ddpg_continuous_action_jax.py
1
CleanRL's ddpg_continuous_action.py
9
Add a comment