Skip to main content
costa-huang
Projects
cleanRL
Reports
PPO: JAX vs Torch in Envpool
Log in
Sign up
Share
Comment
Star
Share
Comment
Star
PPO: JAX vs Torch in Envpool
Costa
Created on June 25
|
Last edited on July 12
Comment
Ant-v2
Ant-v2
2M
4M
6M
8M
Step
-1000
0
1000
2000
3000
4000
5000
Ant-v2
Ant-v2
100
200
300
400
500
Time (seconds)
0
1000
2000
3000
4000
5000
losses/policy_loss
losses/policy_loss
2M
4M
6M
8M
Step
-0.02
0
0.02
0.04
0.06
losses/approx_kl
losses/approx_kl
2M
4M
6M
8M
Step
0.05
0.1
0.15
losses/value_loss
losses/value_loss
2M
4M
6M
8M
Step
0.5
1
1.5
SPS
SPS
2M
4M
6M
8M
Step
5000
10000
15000
20000
25000
CleanRL's ppo_continuous_action_envpool_jax.py (JAX)
1
CleanRL's ppo_continuous_action_envpool_jax.py (torch)
1
CleanRL's ppo_continuous_action_envpool.py (torch)
1
rl_games’ PPO + envpool (64 envs)
4
Add a comment