hh-gpt-j
Created on March 17|Last edited on March 17
Comment
Reproduction of Max's (@reciprocate) HH example but saved with value heads.
https://github.com/CarperAI/trlx/blob/2f90ba0ecd640ae18cd62adb5e934a4b779f534b/examples/hh/ppo_hh.py
Results
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Run: ppo_hh/gpt-j-6B/7gpus:ppo-hh
1
Add a comment