Skip to main content
sorry
Projects
trlx
Reports
Fix ppo calculation with unequal generation lenghts
Log in
Sign up
Share
Comment
Star
Fix ppo calculation with unequal generation lenghts
Sorry
Created on December 14
|
Last edited on December 14
Comment
Section 1
policy/clipfrac
policy/clipfrac
100
200
300
400
Step
0
0.02
0.04
0.06
0.08
ppo_sentiments/gpt2-imdb:fix!
ppo_sentiments/gpt2-imdb:fix?
ppo_sentiments/gpt2-imdb:main
policy/approx_kl
policy/approx_kl
100
200
300
400
Step
0
0.002
0.004
0.006
0.008
ppo_sentiments/gpt2-imdb:fix!
ppo_sentiments/gpt2-imdb:fix?
ppo_sentiments/gpt2-imdb:main
Run set
3
Run set
3
Run set
3
Run set
3
Run set
3
Run set
3
Add a comment