Upload model files
Browse files- checkpoint-100/trainer_state.json +2134 -0
checkpoint-100/trainer_state.json
ADDED
|
@@ -0,0 +1,2134 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"episode": 2400,
|
| 5 |
+
"epoch": 0.34285714285714286,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 100,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"episode": 24,
|
| 14 |
+
"epoch": 0.0034285714285714284,
|
| 15 |
+
"eps": 3,
|
| 16 |
+
"loss/policy_avg": -0.00039759615901857615,
|
| 17 |
+
"loss/value_avg": 0.001173102529719472,
|
| 18 |
+
"lr": 0.0,
|
| 19 |
+
"objective/entropy": 9.603790283203125,
|
| 20 |
+
"objective/kl": 0.0522441528737545,
|
| 21 |
+
"objective/non_score_reward": -0.0026122077833861113,
|
| 22 |
+
"objective/rlhf_reward": -0.0026122077833861113,
|
| 23 |
+
"objective/scores": 0.0,
|
| 24 |
+
"policy/approxkl_avg": 0.00023563228023704141,
|
| 25 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 26 |
+
"policy/entropy_avg": 0.17652443051338196,
|
| 27 |
+
"step": 1,
|
| 28 |
+
"val/clipfrac_avg": 0.0,
|
| 29 |
+
"val/num_eos_tokens": 0,
|
| 30 |
+
"val/ratio": 0.9997509717941284,
|
| 31 |
+
"val/ratio_var": 1.331204543930653e-06
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"episode": 48,
|
| 35 |
+
"epoch": 0.006857142857142857,
|
| 36 |
+
"eps": 2,
|
| 37 |
+
"loss/policy_avg": 9.456602856516838e-05,
|
| 38 |
+
"loss/value_avg": 0.0012460823636502028,
|
| 39 |
+
"lr": 3.3333333333333334e-08,
|
| 40 |
+
"objective/entropy": 11.654745101928711,
|
| 41 |
+
"objective/kl": 0.04007173329591751,
|
| 42 |
+
"objective/non_score_reward": -0.002003586385399103,
|
| 43 |
+
"objective/rlhf_reward": -0.002003586385399103,
|
| 44 |
+
"objective/scores": 0.0,
|
| 45 |
+
"policy/approxkl_avg": 0.0005037242081016302,
|
| 46 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 47 |
+
"policy/entropy_avg": 0.22399930655956268,
|
| 48 |
+
"step": 2,
|
| 49 |
+
"val/clipfrac_avg": 0.0,
|
| 50 |
+
"val/num_eos_tokens": 0,
|
| 51 |
+
"val/ratio": 0.999730110168457,
|
| 52 |
+
"val/ratio_var": 1.496027721259452e-06
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"episode": 72,
|
| 56 |
+
"epoch": 0.010285714285714285,
|
| 57 |
+
"eps": 3,
|
| 58 |
+
"loss/policy_avg": 0.0005133129889145494,
|
| 59 |
+
"loss/value_avg": 0.0011833460303023458,
|
| 60 |
+
"lr": 6.666666666666667e-08,
|
| 61 |
+
"objective/entropy": 9.299583435058594,
|
| 62 |
+
"objective/kl": -0.02202693186700344,
|
| 63 |
+
"objective/non_score_reward": 0.0011013468028977513,
|
| 64 |
+
"objective/rlhf_reward": 0.0011013468028977513,
|
| 65 |
+
"objective/scores": 0.0,
|
| 66 |
+
"policy/approxkl_avg": 0.0002756627509370446,
|
| 67 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 68 |
+
"policy/entropy_avg": 0.16759081184864044,
|
| 69 |
+
"step": 3,
|
| 70 |
+
"val/clipfrac_avg": 0.0,
|
| 71 |
+
"val/num_eos_tokens": 0,
|
| 72 |
+
"val/ratio": 1.0005955696105957,
|
| 73 |
+
"val/ratio_var": 1.5924795206956333e-06
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"episode": 96,
|
| 77 |
+
"epoch": 0.013714285714285714,
|
| 78 |
+
"eps": 3,
|
| 79 |
+
"loss/policy_avg": -0.0007660436676815152,
|
| 80 |
+
"loss/value_avg": 0.0012440377613529563,
|
| 81 |
+
"lr": 1e-07,
|
| 82 |
+
"objective/entropy": 10.092122077941895,
|
| 83 |
+
"objective/kl": -0.001453061937354505,
|
| 84 |
+
"objective/non_score_reward": 7.26534053683281e-05,
|
| 85 |
+
"objective/rlhf_reward": 7.26534053683281e-05,
|
| 86 |
+
"objective/scores": 0.0,
|
| 87 |
+
"policy/approxkl_avg": 0.0004117934440728277,
|
| 88 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 89 |
+
"policy/entropy_avg": 0.18802356719970703,
|
| 90 |
+
"step": 4,
|
| 91 |
+
"val/clipfrac_avg": 0.0,
|
| 92 |
+
"val/num_eos_tokens": 0,
|
| 93 |
+
"val/ratio": 1.0004326105117798,
|
| 94 |
+
"val/ratio_var": 2.240923549834406e-06
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"episode": 120,
|
| 98 |
+
"epoch": 0.017142857142857144,
|
| 99 |
+
"eps": 3,
|
| 100 |
+
"loss/policy_avg": -0.0014500913675874472,
|
| 101 |
+
"loss/value_avg": 0.001122939051128924,
|
| 102 |
+
"lr": 1.3333333333333334e-07,
|
| 103 |
+
"objective/entropy": 10.164737701416016,
|
| 104 |
+
"objective/kl": 0.05467784032225609,
|
| 105 |
+
"objective/non_score_reward": -0.002733892295509577,
|
| 106 |
+
"objective/rlhf_reward": -0.002733892295509577,
|
| 107 |
+
"objective/scores": 0.0,
|
| 108 |
+
"policy/approxkl_avg": 0.0004754411638714373,
|
| 109 |
+
"policy/clipfrac_avg": 0.002358490601181984,
|
| 110 |
+
"policy/entropy_avg": 0.18593449890613556,
|
| 111 |
+
"step": 5,
|
| 112 |
+
"val/clipfrac_avg": 0.0,
|
| 113 |
+
"val/num_eos_tokens": 0,
|
| 114 |
+
"val/ratio": 1.0011851787567139,
|
| 115 |
+
"val/ratio_var": 2.860480663002818e-06
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"episode": 144,
|
| 119 |
+
"epoch": 0.02057142857142857,
|
| 120 |
+
"eps": 3,
|
| 121 |
+
"loss/policy_avg": -0.00038924324326217175,
|
| 122 |
+
"loss/value_avg": 0.0011501125991344452,
|
| 123 |
+
"lr": 1.6666666666666665e-07,
|
| 124 |
+
"objective/entropy": 10.74693489074707,
|
| 125 |
+
"objective/kl": -0.019970744848251343,
|
| 126 |
+
"objective/non_score_reward": 0.0009985374053940177,
|
| 127 |
+
"objective/rlhf_reward": 0.0009985374053940177,
|
| 128 |
+
"objective/scores": 0.0,
|
| 129 |
+
"policy/approxkl_avg": 0.0004121576203033328,
|
| 130 |
+
"policy/clipfrac_avg": 0.0013757861452177167,
|
| 131 |
+
"policy/entropy_avg": 0.20963406562805176,
|
| 132 |
+
"step": 6,
|
| 133 |
+
"val/clipfrac_avg": 0.0,
|
| 134 |
+
"val/num_eos_tokens": 0,
|
| 135 |
+
"val/ratio": 1.0004911422729492,
|
| 136 |
+
"val/ratio_var": 3.83469341613818e-06
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"episode": 168,
|
| 140 |
+
"epoch": 0.024,
|
| 141 |
+
"eps": 3,
|
| 142 |
+
"loss/policy_avg": -0.0006043941248208284,
|
| 143 |
+
"loss/value_avg": 0.0011014372576028109,
|
| 144 |
+
"lr": 2e-07,
|
| 145 |
+
"objective/entropy": 9.983142852783203,
|
| 146 |
+
"objective/kl": -0.023275122046470642,
|
| 147 |
+
"objective/non_score_reward": 0.0011637563584372401,
|
| 148 |
+
"objective/rlhf_reward": 0.0011637563584372401,
|
| 149 |
+
"objective/scores": 0.0,
|
| 150 |
+
"policy/approxkl_avg": 0.00038099882658571005,
|
| 151 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 152 |
+
"policy/entropy_avg": 0.18453757464885712,
|
| 153 |
+
"step": 7,
|
| 154 |
+
"val/clipfrac_avg": 0.0,
|
| 155 |
+
"val/num_eos_tokens": 0,
|
| 156 |
+
"val/ratio": 0.998975396156311,
|
| 157 |
+
"val/ratio_var": 2.2136255211080424e-06
|
| 158 |
+
},
|
| 159 |
+
{
|
| 160 |
+
"episode": 192,
|
| 161 |
+
"epoch": 0.027428571428571427,
|
| 162 |
+
"eps": 3,
|
| 163 |
+
"loss/policy_avg": -0.0012414506636559963,
|
| 164 |
+
"loss/value_avg": 0.0011209447402507067,
|
| 165 |
+
"lr": 2.3333333333333333e-07,
|
| 166 |
+
"objective/entropy": 10.798426628112793,
|
| 167 |
+
"objective/kl": 0.0037710717879235744,
|
| 168 |
+
"objective/non_score_reward": -0.00018855356029234827,
|
| 169 |
+
"objective/rlhf_reward": -0.00018855356029234827,
|
| 170 |
+
"objective/scores": 0.0,
|
| 171 |
+
"policy/approxkl_avg": 0.00043036130955442786,
|
| 172 |
+
"policy/clipfrac_avg": 0.0021619496401399374,
|
| 173 |
+
"policy/entropy_avg": 0.1988888382911682,
|
| 174 |
+
"step": 8,
|
| 175 |
+
"val/clipfrac_avg": 0.0,
|
| 176 |
+
"val/num_eos_tokens": 0,
|
| 177 |
+
"val/ratio": 0.999695360660553,
|
| 178 |
+
"val/ratio_var": 1.6811103478175937e-06
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"episode": 216,
|
| 182 |
+
"epoch": 0.030857142857142857,
|
| 183 |
+
"eps": 3,
|
| 184 |
+
"loss/policy_avg": -0.0001550057204440236,
|
| 185 |
+
"loss/value_avg": 0.0011778806801885366,
|
| 186 |
+
"lr": 2.6666666666666667e-07,
|
| 187 |
+
"objective/entropy": 11.547513961791992,
|
| 188 |
+
"objective/kl": -0.006700088735669851,
|
| 189 |
+
"objective/non_score_reward": 0.00033500464633107185,
|
| 190 |
+
"objective/rlhf_reward": 0.00033500464633107185,
|
| 191 |
+
"objective/scores": 0.0,
|
| 192 |
+
"policy/approxkl_avg": 0.0003637151967268437,
|
| 193 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 194 |
+
"policy/entropy_avg": 0.20382140576839447,
|
| 195 |
+
"step": 9,
|
| 196 |
+
"val/clipfrac_avg": 0.0,
|
| 197 |
+
"val/num_eos_tokens": 0,
|
| 198 |
+
"val/ratio": 1.000375509262085,
|
| 199 |
+
"val/ratio_var": 2.181258651035023e-06
|
| 200 |
+
},
|
| 201 |
+
{
|
| 202 |
+
"episode": 240,
|
| 203 |
+
"epoch": 0.03428571428571429,
|
| 204 |
+
"eps": 3,
|
| 205 |
+
"loss/policy_avg": 0.00045341846998780966,
|
| 206 |
+
"loss/value_avg": 0.0011125364108011127,
|
| 207 |
+
"lr": 3e-07,
|
| 208 |
+
"objective/entropy": 10.174510955810547,
|
| 209 |
+
"objective/kl": 0.07684259861707687,
|
| 210 |
+
"objective/non_score_reward": -0.0038421298377215862,
|
| 211 |
+
"objective/rlhf_reward": -0.0038421298377215862,
|
| 212 |
+
"objective/scores": 0.0,
|
| 213 |
+
"policy/approxkl_avg": 0.00037233877810649574,
|
| 214 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 215 |
+
"policy/entropy_avg": 0.20036275684833527,
|
| 216 |
+
"step": 10,
|
| 217 |
+
"val/clipfrac_avg": 0.0,
|
| 218 |
+
"val/num_eos_tokens": 0,
|
| 219 |
+
"val/ratio": 1.0004992485046387,
|
| 220 |
+
"val/ratio_var": 3.829253728326876e-06
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"episode": 264,
|
| 224 |
+
"epoch": 0.037714285714285714,
|
| 225 |
+
"eps": 4,
|
| 226 |
+
"loss/policy_avg": -0.0006368225440382957,
|
| 227 |
+
"loss/value_avg": 0.0012326778378337622,
|
| 228 |
+
"lr": 3.333333333333333e-07,
|
| 229 |
+
"objective/entropy": 9.915946006774902,
|
| 230 |
+
"objective/kl": 0.08295571058988571,
|
| 231 |
+
"objective/non_score_reward": -0.0041477857157588005,
|
| 232 |
+
"objective/rlhf_reward": -0.0041477857157588005,
|
| 233 |
+
"objective/scores": 0.0,
|
| 234 |
+
"policy/approxkl_avg": 0.0003957168955821544,
|
| 235 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 236 |
+
"policy/entropy_avg": 0.18829384446144104,
|
| 237 |
+
"step": 11,
|
| 238 |
+
"val/clipfrac_avg": 0.0,
|
| 239 |
+
"val/num_eos_tokens": 0,
|
| 240 |
+
"val/ratio": 1.0000991821289062,
|
| 241 |
+
"val/ratio_var": 2.684702167243813e-06
|
| 242 |
+
},
|
| 243 |
+
{
|
| 244 |
+
"episode": 288,
|
| 245 |
+
"epoch": 0.04114285714285714,
|
| 246 |
+
"eps": 4,
|
| 247 |
+
"loss/policy_avg": -0.00045989686623215675,
|
| 248 |
+
"loss/value_avg": 0.001033363863825798,
|
| 249 |
+
"lr": 3.666666666666666e-07,
|
| 250 |
+
"objective/entropy": 9.914142608642578,
|
| 251 |
+
"objective/kl": 0.001713767647743225,
|
| 252 |
+
"objective/non_score_reward": -8.56887418194674e-05,
|
| 253 |
+
"objective/rlhf_reward": -8.56887418194674e-05,
|
| 254 |
+
"objective/scores": 0.0,
|
| 255 |
+
"policy/approxkl_avg": 0.00041497976053506136,
|
| 256 |
+
"policy/clipfrac_avg": 0.0013757861452177167,
|
| 257 |
+
"policy/entropy_avg": 0.20258067548274994,
|
| 258 |
+
"step": 12,
|
| 259 |
+
"val/clipfrac_avg": 0.0,
|
| 260 |
+
"val/num_eos_tokens": 0,
|
| 261 |
+
"val/ratio": 1.000044345855713,
|
| 262 |
+
"val/ratio_var": 2.319321311006206e-06
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"episode": 312,
|
| 266 |
+
"epoch": 0.044571428571428574,
|
| 267 |
+
"eps": 4,
|
| 268 |
+
"loss/policy_avg": -4.981772508472204e-05,
|
| 269 |
+
"loss/value_avg": 0.0011141800787299871,
|
| 270 |
+
"lr": 4e-07,
|
| 271 |
+
"objective/entropy": 8.326139450073242,
|
| 272 |
+
"objective/kl": 0.03665222227573395,
|
| 273 |
+
"objective/non_score_reward": -0.001832611276768148,
|
| 274 |
+
"objective/rlhf_reward": -0.001832611276768148,
|
| 275 |
+
"objective/scores": 0.0,
|
| 276 |
+
"policy/approxkl_avg": 0.0002924801083281636,
|
| 277 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 278 |
+
"policy/entropy_avg": 0.15716251730918884,
|
| 279 |
+
"step": 13,
|
| 280 |
+
"val/clipfrac_avg": 0.0,
|
| 281 |
+
"val/num_eos_tokens": 0,
|
| 282 |
+
"val/ratio": 1.0000745058059692,
|
| 283 |
+
"val/ratio_var": 9.655751682657865e-07
|
| 284 |
+
},
|
| 285 |
+
{
|
| 286 |
+
"episode": 336,
|
| 287 |
+
"epoch": 0.048,
|
| 288 |
+
"eps": 4,
|
| 289 |
+
"loss/policy_avg": -0.0008058485109359026,
|
| 290 |
+
"loss/value_avg": 0.0011609160574153066,
|
| 291 |
+
"lr": 4.3333333333333335e-07,
|
| 292 |
+
"objective/entropy": 10.01640510559082,
|
| 293 |
+
"objective/kl": 0.02480611763894558,
|
| 294 |
+
"objective/non_score_reward": -0.001240305951796472,
|
| 295 |
+
"objective/rlhf_reward": -0.001240305951796472,
|
| 296 |
+
"objective/scores": 0.0,
|
| 297 |
+
"policy/approxkl_avg": 0.00044258785783313215,
|
| 298 |
+
"policy/clipfrac_avg": 0.001965408679097891,
|
| 299 |
+
"policy/entropy_avg": 0.19537438452243805,
|
| 300 |
+
"step": 14,
|
| 301 |
+
"val/clipfrac_avg": 0.0,
|
| 302 |
+
"val/num_eos_tokens": 0,
|
| 303 |
+
"val/ratio": 0.9992372989654541,
|
| 304 |
+
"val/ratio_var": 2.064736008833279e-06
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"episode": 360,
|
| 308 |
+
"epoch": 0.05142857142857143,
|
| 309 |
+
"eps": 4,
|
| 310 |
+
"loss/policy_avg": -0.0012431369395926595,
|
| 311 |
+
"loss/value_avg": 0.0011450606398284435,
|
| 312 |
+
"lr": 4.6666666666666666e-07,
|
| 313 |
+
"objective/entropy": 8.250815391540527,
|
| 314 |
+
"objective/kl": 0.026728516444563866,
|
| 315 |
+
"objective/non_score_reward": -0.0013364258920773864,
|
| 316 |
+
"objective/rlhf_reward": -0.0013364258920773864,
|
| 317 |
+
"objective/scores": 0.0,
|
| 318 |
+
"policy/approxkl_avg": 0.00041782623156905174,
|
| 319 |
+
"policy/clipfrac_avg": 0.0027515722904354334,
|
| 320 |
+
"policy/entropy_avg": 0.15008953213691711,
|
| 321 |
+
"step": 15,
|
| 322 |
+
"val/clipfrac_avg": 0.0,
|
| 323 |
+
"val/num_eos_tokens": 0,
|
| 324 |
+
"val/ratio": 0.9994743466377258,
|
| 325 |
+
"val/ratio_var": 1.3626447525894037e-06
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"episode": 384,
|
| 329 |
+
"epoch": 0.054857142857142854,
|
| 330 |
+
"eps": 4,
|
| 331 |
+
"loss/policy_avg": -0.0016457759775221348,
|
| 332 |
+
"loss/value_avg": 0.0012071160599589348,
|
| 333 |
+
"lr": 5e-07,
|
| 334 |
+
"objective/entropy": 9.560142517089844,
|
| 335 |
+
"objective/kl": 0.04084904119372368,
|
| 336 |
+
"objective/non_score_reward": -0.0020424523390829563,
|
| 337 |
+
"objective/rlhf_reward": -0.0020424523390829563,
|
| 338 |
+
"objective/scores": 0.0,
|
| 339 |
+
"policy/approxkl_avg": 0.00030206821975298226,
|
| 340 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 341 |
+
"policy/entropy_avg": 0.19114547967910767,
|
| 342 |
+
"step": 16,
|
| 343 |
+
"val/clipfrac_avg": 0.0,
|
| 344 |
+
"val/num_eos_tokens": 0,
|
| 345 |
+
"val/ratio": 0.9999871253967285,
|
| 346 |
+
"val/ratio_var": 1.3688204489881173e-06
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"episode": 408,
|
| 350 |
+
"epoch": 0.05828571428571429,
|
| 351 |
+
"eps": 4,
|
| 352 |
+
"loss/policy_avg": 0.00015279045328497887,
|
| 353 |
+
"loss/value_avg": 0.0011765253730118275,
|
| 354 |
+
"lr": 5.333333333333333e-07,
|
| 355 |
+
"objective/entropy": 10.473526000976562,
|
| 356 |
+
"objective/kl": 0.012600630521774292,
|
| 357 |
+
"objective/non_score_reward": -0.0006300313398241997,
|
| 358 |
+
"objective/rlhf_reward": -0.0006300313398241997,
|
| 359 |
+
"objective/scores": 0.0,
|
| 360 |
+
"policy/approxkl_avg": 0.000512410537339747,
|
| 361 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 362 |
+
"policy/entropy_avg": 0.1867837905883789,
|
| 363 |
+
"step": 17,
|
| 364 |
+
"val/clipfrac_avg": 0.0,
|
| 365 |
+
"val/num_eos_tokens": 0,
|
| 366 |
+
"val/ratio": 1.0001708269119263,
|
| 367 |
+
"val/ratio_var": 3.6231228932592785e-06
|
| 368 |
+
},
|
| 369 |
+
{
|
| 370 |
+
"episode": 432,
|
| 371 |
+
"epoch": 0.061714285714285715,
|
| 372 |
+
"eps": 4,
|
| 373 |
+
"loss/policy_avg": -0.0015998110175132751,
|
| 374 |
+
"loss/value_avg": 0.0011272402480244637,
|
| 375 |
+
"lr": 5.666666666666666e-07,
|
| 376 |
+
"objective/entropy": 9.374881744384766,
|
| 377 |
+
"objective/kl": -0.022168656811118126,
|
| 378 |
+
"objective/non_score_reward": 0.001108432887122035,
|
| 379 |
+
"objective/rlhf_reward": 0.001108432887122035,
|
| 380 |
+
"objective/scores": 0.0,
|
| 381 |
+
"policy/approxkl_avg": 0.0003468349459581077,
|
| 382 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 383 |
+
"policy/entropy_avg": 0.17824786901474,
|
| 384 |
+
"step": 18,
|
| 385 |
+
"val/clipfrac_avg": 0.0,
|
| 386 |
+
"val/num_eos_tokens": 0,
|
| 387 |
+
"val/ratio": 1.0005429983139038,
|
| 388 |
+
"val/ratio_var": 1.015850784824579e-06
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"episode": 456,
|
| 392 |
+
"epoch": 0.06514285714285714,
|
| 393 |
+
"eps": 4,
|
| 394 |
+
"loss/policy_avg": 0.0005391768645495176,
|
| 395 |
+
"loss/value_avg": 0.0010727113112807274,
|
| 396 |
+
"lr": 6e-07,
|
| 397 |
+
"objective/entropy": 9.847137451171875,
|
| 398 |
+
"objective/kl": 0.10544447600841522,
|
| 399 |
+
"objective/non_score_reward": -0.0052722240798175335,
|
| 400 |
+
"objective/rlhf_reward": -0.0052722240798175335,
|
| 401 |
+
"objective/scores": 0.0,
|
| 402 |
+
"policy/approxkl_avg": 0.0003828281769528985,
|
| 403 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 404 |
+
"policy/entropy_avg": 0.191776305437088,
|
| 405 |
+
"step": 19,
|
| 406 |
+
"val/clipfrac_avg": 0.0,
|
| 407 |
+
"val/num_eos_tokens": 0,
|
| 408 |
+
"val/ratio": 0.9989888072013855,
|
| 409 |
+
"val/ratio_var": 1.1592291002671118e-06
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"episode": 480,
|
| 413 |
+
"epoch": 0.06857142857142857,
|
| 414 |
+
"eps": 4,
|
| 415 |
+
"loss/policy_avg": 0.0004999454831704497,
|
| 416 |
+
"loss/value_avg": 0.0011963938595727086,
|
| 417 |
+
"lr": 6.333333333333332e-07,
|
| 418 |
+
"objective/entropy": 9.980831146240234,
|
| 419 |
+
"objective/kl": 0.009486228227615356,
|
| 420 |
+
"objective/non_score_reward": -0.0004743114986922592,
|
| 421 |
+
"objective/rlhf_reward": -0.0004743114986922592,
|
| 422 |
+
"objective/scores": 0.0,
|
| 423 |
+
"policy/approxkl_avg": 0.0003429501666687429,
|
| 424 |
+
"policy/clipfrac_avg": 0.0,
|
| 425 |
+
"policy/entropy_avg": 0.18964967131614685,
|
| 426 |
+
"step": 20,
|
| 427 |
+
"val/clipfrac_avg": 0.0,
|
| 428 |
+
"val/num_eos_tokens": 0,
|
| 429 |
+
"val/ratio": 0.9999833703041077,
|
| 430 |
+
"val/ratio_var": 3.019130190295982e-06
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"episode": 504,
|
| 434 |
+
"epoch": 0.072,
|
| 435 |
+
"eps": 4,
|
| 436 |
+
"loss/policy_avg": -0.0006457410054281354,
|
| 437 |
+
"loss/value_avg": 0.0011750897392630577,
|
| 438 |
+
"lr": 6.666666666666666e-07,
|
| 439 |
+
"objective/entropy": 8.674556732177734,
|
| 440 |
+
"objective/kl": 0.16487348079681396,
|
| 441 |
+
"objective/non_score_reward": -0.008243675343692303,
|
| 442 |
+
"objective/rlhf_reward": -0.008243675343692303,
|
| 443 |
+
"objective/scores": 0.0,
|
| 444 |
+
"policy/approxkl_avg": 0.00034728588070720434,
|
| 445 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 446 |
+
"policy/entropy_avg": 0.17806154489517212,
|
| 447 |
+
"step": 21,
|
| 448 |
+
"val/clipfrac_avg": 0.0,
|
| 449 |
+
"val/num_eos_tokens": 0,
|
| 450 |
+
"val/ratio": 0.9990400671958923,
|
| 451 |
+
"val/ratio_var": 1.2348247082627495e-06
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"episode": 528,
|
| 455 |
+
"epoch": 0.07542857142857143,
|
| 456 |
+
"eps": 4,
|
| 457 |
+
"loss/policy_avg": -0.0015686429105699062,
|
| 458 |
+
"loss/value_avg": 0.001164754037745297,
|
| 459 |
+
"lr": 7e-07,
|
| 460 |
+
"objective/entropy": 10.79302978515625,
|
| 461 |
+
"objective/kl": -0.08593261241912842,
|
| 462 |
+
"objective/non_score_reward": 0.004296630620956421,
|
| 463 |
+
"objective/rlhf_reward": 0.004296630620956421,
|
| 464 |
+
"objective/scores": 0.0,
|
| 465 |
+
"policy/approxkl_avg": 0.00043445732444524765,
|
| 466 |
+
"policy/clipfrac_avg": 0.001768867950886488,
|
| 467 |
+
"policy/entropy_avg": 0.1922873705625534,
|
| 468 |
+
"step": 22,
|
| 469 |
+
"val/clipfrac_avg": 0.0,
|
| 470 |
+
"val/num_eos_tokens": 0,
|
| 471 |
+
"val/ratio": 1.0014443397521973,
|
| 472 |
+
"val/ratio_var": 3.1963945730240084e-06
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"episode": 552,
|
| 476 |
+
"epoch": 0.07885714285714286,
|
| 477 |
+
"eps": 4,
|
| 478 |
+
"loss/policy_avg": -0.0007012488786131144,
|
| 479 |
+
"loss/value_avg": 0.0011740096379071474,
|
| 480 |
+
"lr": 7.333333333333332e-07,
|
| 481 |
+
"objective/entropy": 9.420793533325195,
|
| 482 |
+
"objective/kl": 0.13005897402763367,
|
| 483 |
+
"objective/non_score_reward": -0.006502948235720396,
|
| 484 |
+
"objective/rlhf_reward": -0.006502948235720396,
|
| 485 |
+
"objective/scores": 0.0,
|
| 486 |
+
"policy/approxkl_avg": 0.000359610392479226,
|
| 487 |
+
"policy/clipfrac_avg": 0.0013757860288023949,
|
| 488 |
+
"policy/entropy_avg": 0.1827276051044464,
|
| 489 |
+
"step": 23,
|
| 490 |
+
"val/clipfrac_avg": 0.0,
|
| 491 |
+
"val/num_eos_tokens": 0,
|
| 492 |
+
"val/ratio": 0.9988819360733032,
|
| 493 |
+
"val/ratio_var": 1.4016917475601076e-06
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"episode": 576,
|
| 497 |
+
"epoch": 0.08228571428571428,
|
| 498 |
+
"eps": 4,
|
| 499 |
+
"loss/policy_avg": -0.0009957049041986465,
|
| 500 |
+
"loss/value_avg": 0.001171375741250813,
|
| 501 |
+
"lr": 7.666666666666667e-07,
|
| 502 |
+
"objective/entropy": 12.150233268737793,
|
| 503 |
+
"objective/kl": 0.11499027907848358,
|
| 504 |
+
"objective/non_score_reward": -0.005749514326453209,
|
| 505 |
+
"objective/rlhf_reward": -0.005749514326453209,
|
| 506 |
+
"objective/scores": 0.0,
|
| 507 |
+
"policy/approxkl_avg": 0.0012159445323050022,
|
| 508 |
+
"policy/clipfrac_avg": 0.004323899280279875,
|
| 509 |
+
"policy/entropy_avg": 0.22310155630111694,
|
| 510 |
+
"step": 24,
|
| 511 |
+
"val/clipfrac_avg": 0.0,
|
| 512 |
+
"val/num_eos_tokens": 0,
|
| 513 |
+
"val/ratio": 0.9989429712295532,
|
| 514 |
+
"val/ratio_var": 1.6364158454962308e-06
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"episode": 600,
|
| 518 |
+
"epoch": 0.08571428571428572,
|
| 519 |
+
"eps": 4,
|
| 520 |
+
"loss/policy_avg": 5.847762804478407e-05,
|
| 521 |
+
"loss/value_avg": 0.0012743598781526089,
|
| 522 |
+
"lr": 8e-07,
|
| 523 |
+
"objective/entropy": 10.221760749816895,
|
| 524 |
+
"objective/kl": -0.07403554022312164,
|
| 525 |
+
"objective/non_score_reward": 0.0037017769645899534,
|
| 526 |
+
"objective/rlhf_reward": 0.0037017769645899534,
|
| 527 |
+
"objective/scores": 0.0,
|
| 528 |
+
"policy/approxkl_avg": 0.00038175773806869984,
|
| 529 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 530 |
+
"policy/entropy_avg": 0.1923346221446991,
|
| 531 |
+
"step": 25,
|
| 532 |
+
"val/clipfrac_avg": 0.0,
|
| 533 |
+
"val/num_eos_tokens": 0,
|
| 534 |
+
"val/ratio": 1.0002760887145996,
|
| 535 |
+
"val/ratio_var": 2.4850394311215496e-06
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"episode": 624,
|
| 539 |
+
"epoch": 0.08914285714285715,
|
| 540 |
+
"eps": 4,
|
| 541 |
+
"loss/policy_avg": 0.0005327528342604637,
|
| 542 |
+
"loss/value_avg": 0.001223960192874074,
|
| 543 |
+
"lr": 8.333333333333333e-07,
|
| 544 |
+
"objective/entropy": 9.961271286010742,
|
| 545 |
+
"objective/kl": -0.03506360575556755,
|
| 546 |
+
"objective/non_score_reward": 0.0017531800549477339,
|
| 547 |
+
"objective/rlhf_reward": 0.0017531800549477339,
|
| 548 |
+
"objective/scores": 0.0,
|
| 549 |
+
"policy/approxkl_avg": 0.0003690699813887477,
|
| 550 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 551 |
+
"policy/entropy_avg": 0.19056138396263123,
|
| 552 |
+
"step": 26,
|
| 553 |
+
"val/clipfrac_avg": 0.0,
|
| 554 |
+
"val/num_eos_tokens": 0,
|
| 555 |
+
"val/ratio": 0.9998593926429749,
|
| 556 |
+
"val/ratio_var": 2.2398712644644547e-06
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"episode": 648,
|
| 560 |
+
"epoch": 0.09257142857142857,
|
| 561 |
+
"eps": 4,
|
| 562 |
+
"loss/policy_avg": -0.0012908531352877617,
|
| 563 |
+
"loss/value_avg": 0.0011609859066084027,
|
| 564 |
+
"lr": 8.666666666666667e-07,
|
| 565 |
+
"objective/entropy": 10.788581848144531,
|
| 566 |
+
"objective/kl": -0.035950906574726105,
|
| 567 |
+
"objective/non_score_reward": 0.0017975454684346914,
|
| 568 |
+
"objective/rlhf_reward": 0.0017975454684346914,
|
| 569 |
+
"objective/scores": 0.0,
|
| 570 |
+
"policy/approxkl_avg": 0.00042478417162783444,
|
| 571 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 572 |
+
"policy/entropy_avg": 0.20873138308525085,
|
| 573 |
+
"step": 27,
|
| 574 |
+
"val/clipfrac_avg": 0.0,
|
| 575 |
+
"val/num_eos_tokens": 0,
|
| 576 |
+
"val/ratio": 0.9998935461044312,
|
| 577 |
+
"val/ratio_var": 2.4417051918135257e-06
|
| 578 |
+
},
|
| 579 |
+
{
|
| 580 |
+
"episode": 672,
|
| 581 |
+
"epoch": 0.096,
|
| 582 |
+
"eps": 4,
|
| 583 |
+
"loss/policy_avg": 0.000542065070476383,
|
| 584 |
+
"loss/value_avg": 0.0012408210895955563,
|
| 585 |
+
"lr": 9e-07,
|
| 586 |
+
"objective/entropy": 8.87173080444336,
|
| 587 |
+
"objective/kl": -0.033091746270656586,
|
| 588 |
+
"objective/non_score_reward": 0.0016545872204005718,
|
| 589 |
+
"objective/rlhf_reward": 0.0016545872204005718,
|
| 590 |
+
"objective/scores": 0.0,
|
| 591 |
+
"policy/approxkl_avg": 0.00035692937672138214,
|
| 592 |
+
"policy/clipfrac_avg": 0.0,
|
| 593 |
+
"policy/entropy_avg": 0.17584078013896942,
|
| 594 |
+
"step": 28,
|
| 595 |
+
"val/clipfrac_avg": 0.0,
|
| 596 |
+
"val/num_eos_tokens": 0,
|
| 597 |
+
"val/ratio": 0.9999797940254211,
|
| 598 |
+
"val/ratio_var": 2.1010207547078608e-06
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"episode": 696,
|
| 602 |
+
"epoch": 0.09942857142857142,
|
| 603 |
+
"eps": 3,
|
| 604 |
+
"loss/policy_avg": 0.00025542714865878224,
|
| 605 |
+
"loss/value_avg": 0.0009407824254594743,
|
| 606 |
+
"lr": 9.333333333333333e-07,
|
| 607 |
+
"objective/entropy": 10.091318130493164,
|
| 608 |
+
"objective/kl": -0.0042562782764434814,
|
| 609 |
+
"objective/non_score_reward": 0.00021281388762872666,
|
| 610 |
+
"objective/rlhf_reward": 0.00021281388762872666,
|
| 611 |
+
"objective/scores": 0.0,
|
| 612 |
+
"policy/approxkl_avg": 0.0004177533555775881,
|
| 613 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 614 |
+
"policy/entropy_avg": 0.20332473516464233,
|
| 615 |
+
"step": 29,
|
| 616 |
+
"val/clipfrac_avg": 0.0,
|
| 617 |
+
"val/num_eos_tokens": 0,
|
| 618 |
+
"val/ratio": 1.0008416175842285,
|
| 619 |
+
"val/ratio_var": 4.070296654390404e-06
|
| 620 |
+
},
|
| 621 |
+
{
|
| 622 |
+
"episode": 720,
|
| 623 |
+
"epoch": 0.10285714285714286,
|
| 624 |
+
"eps": 3,
|
| 625 |
+
"loss/policy_avg": -0.0018287734128534794,
|
| 626 |
+
"loss/value_avg": 0.0010721104918047786,
|
| 627 |
+
"lr": 9.666666666666666e-07,
|
| 628 |
+
"objective/entropy": 11.005254745483398,
|
| 629 |
+
"objective/kl": 0.05638129264116287,
|
| 630 |
+
"objective/non_score_reward": -0.002819064538925886,
|
| 631 |
+
"objective/rlhf_reward": -0.002819064538925886,
|
| 632 |
+
"objective/scores": 0.0,
|
| 633 |
+
"policy/approxkl_avg": 0.00047117556096054614,
|
| 634 |
+
"policy/clipfrac_avg": 0.0015723269898444414,
|
| 635 |
+
"policy/entropy_avg": 0.20323297381401062,
|
| 636 |
+
"step": 30,
|
| 637 |
+
"val/clipfrac_avg": 0.0,
|
| 638 |
+
"val/num_eos_tokens": 0,
|
| 639 |
+
"val/ratio": 0.9995233416557312,
|
| 640 |
+
"val/ratio_var": 2.1699356693716254e-06
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"episode": 744,
|
| 644 |
+
"epoch": 0.10628571428571429,
|
| 645 |
+
"eps": 3,
|
| 646 |
+
"loss/policy_avg": -0.0013574458425864577,
|
| 647 |
+
"loss/value_avg": 0.0011641676537692547,
|
| 648 |
+
"lr": 1e-06,
|
| 649 |
+
"objective/entropy": 11.450809478759766,
|
| 650 |
+
"objective/kl": -0.010124157182872295,
|
| 651 |
+
"objective/non_score_reward": 0.000506207812577486,
|
| 652 |
+
"objective/rlhf_reward": 0.000506207812577486,
|
| 653 |
+
"objective/scores": 0.0,
|
| 654 |
+
"policy/approxkl_avg": 0.0004096939228475094,
|
| 655 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 656 |
+
"policy/entropy_avg": 0.2194271683692932,
|
| 657 |
+
"step": 31,
|
| 658 |
+
"val/clipfrac_avg": 0.0,
|
| 659 |
+
"val/num_eos_tokens": 0,
|
| 660 |
+
"val/ratio": 1.000313401222229,
|
| 661 |
+
"val/ratio_var": 2.678104010556126e-06
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"episode": 768,
|
| 665 |
+
"epoch": 0.10971428571428571,
|
| 666 |
+
"eps": 3,
|
| 667 |
+
"loss/policy_avg": -0.0002802757080644369,
|
| 668 |
+
"loss/value_avg": 0.001182440435513854,
|
| 669 |
+
"lr": 9.999676499856762e-07,
|
| 670 |
+
"objective/entropy": 9.76666259765625,
|
| 671 |
+
"objective/kl": 0.06829522550106049,
|
| 672 |
+
"objective/non_score_reward": -0.0034147612750530243,
|
| 673 |
+
"objective/rlhf_reward": -0.0034147612750530243,
|
| 674 |
+
"objective/scores": 0.0,
|
| 675 |
+
"policy/approxkl_avg": 0.00030742710805498064,
|
| 676 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 677 |
+
"policy/entropy_avg": 0.19262900948524475,
|
| 678 |
+
"step": 32,
|
| 679 |
+
"val/clipfrac_avg": 0.0,
|
| 680 |
+
"val/num_eos_tokens": 0,
|
| 681 |
+
"val/ratio": 1.0005744695663452,
|
| 682 |
+
"val/ratio_var": 2.010586285905447e-06
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"episode": 792,
|
| 686 |
+
"epoch": 0.11314285714285714,
|
| 687 |
+
"eps": 3,
|
| 688 |
+
"loss/policy_avg": -4.241405986249447e-05,
|
| 689 |
+
"loss/value_avg": 0.001140992040745914,
|
| 690 |
+
"lr": 9.998706045939205e-07,
|
| 691 |
+
"objective/entropy": 9.192793846130371,
|
| 692 |
+
"objective/kl": 0.011741682887077332,
|
| 693 |
+
"objective/non_score_reward": -0.0005870835739187896,
|
| 694 |
+
"objective/rlhf_reward": -0.0005870835739187896,
|
| 695 |
+
"objective/scores": 0.0,
|
| 696 |
+
"policy/approxkl_avg": 0.0003442139131948352,
|
| 697 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 698 |
+
"policy/entropy_avg": 0.18409989774227142,
|
| 699 |
+
"step": 33,
|
| 700 |
+
"val/clipfrac_avg": 0.0,
|
| 701 |
+
"val/num_eos_tokens": 0,
|
| 702 |
+
"val/ratio": 1.0005278587341309,
|
| 703 |
+
"val/ratio_var": 2.262753923787386e-06
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"episode": 816,
|
| 707 |
+
"epoch": 0.11657142857142858,
|
| 708 |
+
"eps": 3,
|
| 709 |
+
"loss/policy_avg": -0.0008663232438266277,
|
| 710 |
+
"loss/value_avg": 0.001219075871631503,
|
| 711 |
+
"lr": 9.997088777777095e-07,
|
| 712 |
+
"objective/entropy": 9.554176330566406,
|
| 713 |
+
"objective/kl": 0.09085823595523834,
|
| 714 |
+
"objective/non_score_reward": -0.004542911425232887,
|
| 715 |
+
"objective/rlhf_reward": -0.004542911425232887,
|
| 716 |
+
"objective/scores": 0.0,
|
| 717 |
+
"policy/approxkl_avg": 0.0003868684871122241,
|
| 718 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 719 |
+
"policy/entropy_avg": 0.2021941840648651,
|
| 720 |
+
"step": 34,
|
| 721 |
+
"val/clipfrac_avg": 0.0,
|
| 722 |
+
"val/num_eos_tokens": 0,
|
| 723 |
+
"val/ratio": 0.9998293519020081,
|
| 724 |
+
"val/ratio_var": 1.1972449556196807e-06
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"episode": 840,
|
| 728 |
+
"epoch": 0.12,
|
| 729 |
+
"eps": 3,
|
| 730 |
+
"loss/policy_avg": -1.8059727153740823e-05,
|
| 731 |
+
"loss/value_avg": 0.00110139069147408,
|
| 732 |
+
"lr": 9.994824927897762e-07,
|
| 733 |
+
"objective/entropy": 10.6250638961792,
|
| 734 |
+
"objective/kl": -0.057124651968479156,
|
| 735 |
+
"objective/non_score_reward": 0.002856232225894928,
|
| 736 |
+
"objective/rlhf_reward": 0.002856232225894928,
|
| 737 |
+
"objective/scores": 0.0,
|
| 738 |
+
"policy/approxkl_avg": 0.00036990485386922956,
|
| 739 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 740 |
+
"policy/entropy_avg": 0.19656652212142944,
|
| 741 |
+
"step": 35,
|
| 742 |
+
"val/clipfrac_avg": 0.0,
|
| 743 |
+
"val/num_eos_tokens": 0,
|
| 744 |
+
"val/ratio": 1.000251054763794,
|
| 745 |
+
"val/ratio_var": 3.1737645258544944e-06
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"episode": 864,
|
| 749 |
+
"epoch": 0.12342857142857143,
|
| 750 |
+
"eps": 3,
|
| 751 |
+
"loss/policy_avg": 0.0005211837124079466,
|
| 752 |
+
"loss/value_avg": 0.001098247361369431,
|
| 753 |
+
"lr": 9.99191482179265e-07,
|
| 754 |
+
"objective/entropy": 11.432059288024902,
|
| 755 |
+
"objective/kl": 0.03775228559970856,
|
| 756 |
+
"objective/non_score_reward": -0.001887614605948329,
|
| 757 |
+
"objective/rlhf_reward": -0.001887614605948329,
|
| 758 |
+
"objective/scores": 0.0,
|
| 759 |
+
"policy/approxkl_avg": 0.0015855329111218452,
|
| 760 |
+
"policy/clipfrac_avg": 0.001965408679097891,
|
| 761 |
+
"policy/entropy_avg": 0.2018502950668335,
|
| 762 |
+
"step": 36,
|
| 763 |
+
"val/clipfrac_avg": 0.0,
|
| 764 |
+
"val/num_eos_tokens": 0,
|
| 765 |
+
"val/ratio": 0.9998794794082642,
|
| 766 |
+
"val/ratio_var": 2.63856213678082e-06
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"episode": 888,
|
| 770 |
+
"epoch": 0.12685714285714286,
|
| 771 |
+
"eps": 3,
|
| 772 |
+
"loss/policy_avg": -0.0003723090048879385,
|
| 773 |
+
"loss/value_avg": 0.0011761891655623913,
|
| 774 |
+
"lr": 9.988358877870534e-07,
|
| 775 |
+
"objective/entropy": 13.564403533935547,
|
| 776 |
+
"objective/kl": 0.14848901331424713,
|
| 777 |
+
"objective/non_score_reward": -0.007424450945109129,
|
| 778 |
+
"objective/rlhf_reward": -0.007424450945109129,
|
| 779 |
+
"objective/scores": 0.0,
|
| 780 |
+
"policy/approxkl_avg": 0.0005069933249615133,
|
| 781 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 782 |
+
"policy/entropy_avg": 0.21795369684696198,
|
| 783 |
+
"step": 37,
|
| 784 |
+
"val/clipfrac_avg": 0.0,
|
| 785 |
+
"val/num_eos_tokens": 0,
|
| 786 |
+
"val/ratio": 0.9994122982025146,
|
| 787 |
+
"val/ratio_var": 3.516516017043614e-06
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"episode": 912,
|
| 791 |
+
"epoch": 0.13028571428571428,
|
| 792 |
+
"eps": 3,
|
| 793 |
+
"loss/policy_avg": -0.0005580224096775055,
|
| 794 |
+
"loss/value_avg": 0.0010930404532700777,
|
| 795 |
+
"lr": 9.984157607397357e-07,
|
| 796 |
+
"objective/entropy": 12.46867561340332,
|
| 797 |
+
"objective/kl": 0.04366648197174072,
|
| 798 |
+
"objective/non_score_reward": -0.00218332395888865,
|
| 799 |
+
"objective/rlhf_reward": -0.00218332395888865,
|
| 800 |
+
"objective/scores": 0.0,
|
| 801 |
+
"policy/approxkl_avg": 0.000503862276673317,
|
| 802 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 803 |
+
"policy/entropy_avg": 0.2015783190727234,
|
| 804 |
+
"step": 38,
|
| 805 |
+
"val/clipfrac_avg": 0.0,
|
| 806 |
+
"val/num_eos_tokens": 0,
|
| 807 |
+
"val/ratio": 1.0004000663757324,
|
| 808 |
+
"val/ratio_var": 6.4164978539338335e-06
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"episode": 936,
|
| 812 |
+
"epoch": 0.1337142857142857,
|
| 813 |
+
"eps": 3,
|
| 814 |
+
"loss/policy_avg": -0.0002338151098228991,
|
| 815 |
+
"loss/value_avg": 0.0011000724043697119,
|
| 816 |
+
"lr": 9.979311614422718e-07,
|
| 817 |
+
"objective/entropy": 11.266685485839844,
|
| 818 |
+
"objective/kl": 0.10951797664165497,
|
| 819 |
+
"objective/non_score_reward": -0.005475898738950491,
|
| 820 |
+
"objective/rlhf_reward": -0.005475898738950491,
|
| 821 |
+
"objective/scores": 0.0,
|
| 822 |
+
"policy/approxkl_avg": 0.0004268661141395569,
|
| 823 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 824 |
+
"policy/entropy_avg": 0.21669596433639526,
|
| 825 |
+
"step": 39,
|
| 826 |
+
"val/clipfrac_avg": 0.0,
|
| 827 |
+
"val/num_eos_tokens": 0,
|
| 828 |
+
"val/ratio": 0.9992050528526306,
|
| 829 |
+
"val/ratio_var": 1.6559835103180376e-06
|
| 830 |
+
},
|
| 831 |
+
{
|
| 832 |
+
"episode": 960,
|
| 833 |
+
"epoch": 0.13714285714285715,
|
| 834 |
+
"eps": 3,
|
| 835 |
+
"loss/policy_avg": -0.0007765925256535411,
|
| 836 |
+
"loss/value_avg": 0.001088449265807867,
|
| 837 |
+
"lr": 9.973821595693026e-07,
|
| 838 |
+
"objective/entropy": 12.338579177856445,
|
| 839 |
+
"objective/kl": 0.06209728866815567,
|
| 840 |
+
"objective/non_score_reward": -0.0031048643868416548,
|
| 841 |
+
"objective/rlhf_reward": -0.0031048643868416548,
|
| 842 |
+
"objective/scores": 0.0,
|
| 843 |
+
"policy/approxkl_avg": 0.0005026735016144812,
|
| 844 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 845 |
+
"policy/entropy_avg": 0.2244863510131836,
|
| 846 |
+
"step": 40,
|
| 847 |
+
"val/clipfrac_avg": 0.0,
|
| 848 |
+
"val/num_eos_tokens": 0,
|
| 849 |
+
"val/ratio": 0.9999690651893616,
|
| 850 |
+
"val/ratio_var": 5.0828230087063275e-06
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"episode": 984,
|
| 854 |
+
"epoch": 0.14057142857142857,
|
| 855 |
+
"eps": 3,
|
| 856 |
+
"loss/policy_avg": 0.00013598555233329535,
|
| 857 |
+
"loss/value_avg": 0.001187118818052113,
|
| 858 |
+
"lr": 9.967688340551327e-07,
|
| 859 |
+
"objective/entropy": 9.387279510498047,
|
| 860 |
+
"objective/kl": -0.033916082233190536,
|
| 861 |
+
"objective/non_score_reward": 0.0016958042979240417,
|
| 862 |
+
"objective/rlhf_reward": 0.0016958042979240417,
|
| 863 |
+
"objective/scores": 0.0,
|
| 864 |
+
"policy/approxkl_avg": 0.0004025290545541793,
|
| 865 |
+
"policy/clipfrac_avg": 0.0013757861452177167,
|
| 866 |
+
"policy/entropy_avg": 0.19263054430484772,
|
| 867 |
+
"step": 41,
|
| 868 |
+
"val/clipfrac_avg": 0.0,
|
| 869 |
+
"val/num_eos_tokens": 0,
|
| 870 |
+
"val/ratio": 1.0011463165283203,
|
| 871 |
+
"val/ratio_var": 3.6780629670829512e-06
|
| 872 |
+
},
|
| 873 |
+
{
|
| 874 |
+
"episode": 1008,
|
| 875 |
+
"epoch": 0.144,
|
| 876 |
+
"eps": 3,
|
| 877 |
+
"loss/policy_avg": -0.001398796564899385,
|
| 878 |
+
"loss/value_avg": 0.0012791166082024574,
|
| 879 |
+
"lr": 9.960912730823802e-07,
|
| 880 |
+
"objective/entropy": 9.443194389343262,
|
| 881 |
+
"objective/kl": 0.040594689548015594,
|
| 882 |
+
"objective/non_score_reward": -0.0020297346636652946,
|
| 883 |
+
"objective/rlhf_reward": -0.0020297346636652946,
|
| 884 |
+
"objective/scores": 0.0,
|
| 885 |
+
"policy/approxkl_avg": 0.00038655230309814215,
|
| 886 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 887 |
+
"policy/entropy_avg": 0.17716416716575623,
|
| 888 |
+
"step": 42,
|
| 889 |
+
"val/clipfrac_avg": 0.0,
|
| 890 |
+
"val/num_eos_tokens": 0,
|
| 891 |
+
"val/ratio": 0.9991838335990906,
|
| 892 |
+
"val/ratio_var": 1.4638164884672733e-06
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"episode": 1032,
|
| 896 |
+
"epoch": 0.14742857142857144,
|
| 897 |
+
"eps": 3,
|
| 898 |
+
"loss/policy_avg": -0.0008904861751943827,
|
| 899 |
+
"loss/value_avg": 0.0012651337310671806,
|
| 900 |
+
"lr": 9.953495740692994e-07,
|
| 901 |
+
"objective/entropy": 8.784332275390625,
|
| 902 |
+
"objective/kl": 0.048220910131931305,
|
| 903 |
+
"objective/non_score_reward": -0.0024110455997288227,
|
| 904 |
+
"objective/rlhf_reward": -0.0024110455997288227,
|
| 905 |
+
"objective/scores": 0.0,
|
| 906 |
+
"policy/approxkl_avg": 0.00036320951767265797,
|
| 907 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 908 |
+
"policy/entropy_avg": 0.19501766562461853,
|
| 909 |
+
"step": 43,
|
| 910 |
+
"val/clipfrac_avg": 0.0,
|
| 911 |
+
"val/num_eos_tokens": 0,
|
| 912 |
+
"val/ratio": 0.9992011785507202,
|
| 913 |
+
"val/ratio_var": 7.953665885906958e-07
|
| 914 |
+
},
|
| 915 |
+
{
|
| 916 |
+
"episode": 1056,
|
| 917 |
+
"epoch": 0.15085714285714286,
|
| 918 |
+
"eps": 3,
|
| 919 |
+
"loss/policy_avg": -0.00040724524296820164,
|
| 920 |
+
"loss/value_avg": 0.0012381336418911815,
|
| 921 |
+
"lr": 9.945438436557734e-07,
|
| 922 |
+
"objective/entropy": 10.724536895751953,
|
| 923 |
+
"objective/kl": 0.09735743701457977,
|
| 924 |
+
"objective/non_score_reward": -0.004867871757596731,
|
| 925 |
+
"objective/rlhf_reward": -0.004867871757596731,
|
| 926 |
+
"objective/scores": 0.0,
|
| 927 |
+
"policy/approxkl_avg": 0.0004834880237467587,
|
| 928 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 929 |
+
"policy/entropy_avg": 0.19912828505039215,
|
| 930 |
+
"step": 44,
|
| 931 |
+
"val/clipfrac_avg": 0.0,
|
| 932 |
+
"val/num_eos_tokens": 0,
|
| 933 |
+
"val/ratio": 0.9993612766265869,
|
| 934 |
+
"val/ratio_var": 3.948975518142106e-06
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"episode": 1080,
|
| 938 |
+
"epoch": 0.15428571428571428,
|
| 939 |
+
"eps": 3,
|
| 940 |
+
"loss/policy_avg": -0.0007694043451920152,
|
| 941 |
+
"loss/value_avg": 0.0012855801032856107,
|
| 942 |
+
"lr": 9.93674197687982e-07,
|
| 943 |
+
"objective/entropy": 10.774508476257324,
|
| 944 |
+
"objective/kl": 0.036663156002759933,
|
| 945 |
+
"objective/non_score_reward": -0.0018331576138734818,
|
| 946 |
+
"objective/rlhf_reward": -0.0018331576138734818,
|
| 947 |
+
"objective/scores": 0.0,
|
| 948 |
+
"policy/approxkl_avg": 0.0003691558085847646,
|
| 949 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 950 |
+
"policy/entropy_avg": 0.19582445919513702,
|
| 951 |
+
"step": 45,
|
| 952 |
+
"val/clipfrac_avg": 0.0,
|
| 953 |
+
"val/num_eos_tokens": 0,
|
| 954 |
+
"val/ratio": 1.0001473426818848,
|
| 955 |
+
"val/ratio_var": 3.0011931357876165e-06
|
| 956 |
+
},
|
| 957 |
+
{
|
| 958 |
+
"episode": 1104,
|
| 959 |
+
"epoch": 0.15771428571428572,
|
| 960 |
+
"eps": 3,
|
| 961 |
+
"loss/policy_avg": -0.0010882224887609482,
|
| 962 |
+
"loss/value_avg": 0.0011191555531695485,
|
| 963 |
+
"lr": 9.927407612017446e-07,
|
| 964 |
+
"objective/entropy": 8.94318962097168,
|
| 965 |
+
"objective/kl": 0.036568351089954376,
|
| 966 |
+
"objective/non_score_reward": -0.0018284174147993326,
|
| 967 |
+
"objective/rlhf_reward": -0.0018284174147993326,
|
| 968 |
+
"objective/scores": 0.0,
|
| 969 |
+
"policy/approxkl_avg": 0.00030608323868364096,
|
| 970 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 971 |
+
"policy/entropy_avg": 0.18505753576755524,
|
| 972 |
+
"step": 46,
|
| 973 |
+
"val/clipfrac_avg": 0.0,
|
| 974 |
+
"val/num_eos_tokens": 0,
|
| 975 |
+
"val/ratio": 0.9989262819290161,
|
| 976 |
+
"val/ratio_var": 2.5265023850806756e-06
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"episode": 1128,
|
| 980 |
+
"epoch": 0.16114285714285714,
|
| 981 |
+
"eps": 3,
|
| 982 |
+
"loss/policy_avg": 6.927899084985256e-05,
|
| 983 |
+
"loss/value_avg": 0.0010387528454884887,
|
| 984 |
+
"lr": 9.91743668404545e-07,
|
| 985 |
+
"objective/entropy": 13.284429550170898,
|
| 986 |
+
"objective/kl": 0.0841490626335144,
|
| 987 |
+
"objective/non_score_reward": -0.00420745275914669,
|
| 988 |
+
"objective/rlhf_reward": -0.00420745275914669,
|
| 989 |
+
"objective/scores": 0.0,
|
| 990 |
+
"policy/approxkl_avg": 0.0005432004109025002,
|
| 991 |
+
"policy/clipfrac_avg": 0.0011792451841756701,
|
| 992 |
+
"policy/entropy_avg": 0.2116156369447708,
|
| 993 |
+
"step": 47,
|
| 994 |
+
"val/clipfrac_avg": 0.0,
|
| 995 |
+
"val/num_eos_tokens": 0,
|
| 996 |
+
"val/ratio": 0.9999104738235474,
|
| 997 |
+
"val/ratio_var": 2.7350765776645858e-06
|
| 998 |
+
},
|
| 999 |
+
{
|
| 1000 |
+
"episode": 1152,
|
| 1001 |
+
"epoch": 0.16457142857142856,
|
| 1002 |
+
"eps": 3,
|
| 1003 |
+
"loss/policy_avg": -0.00019468856044113636,
|
| 1004 |
+
"loss/value_avg": 0.0011783144436776638,
|
| 1005 |
+
"lr": 9.906830626562331e-07,
|
| 1006 |
+
"objective/entropy": 11.494680404663086,
|
| 1007 |
+
"objective/kl": 0.04910100996494293,
|
| 1008 |
+
"objective/non_score_reward": -0.0024550508242100477,
|
| 1009 |
+
"objective/rlhf_reward": -0.0024550508242100477,
|
| 1010 |
+
"objective/scores": 0.0,
|
| 1011 |
+
"policy/approxkl_avg": 0.00043969100806862116,
|
| 1012 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 1013 |
+
"policy/entropy_avg": 0.23010241985321045,
|
| 1014 |
+
"step": 48,
|
| 1015 |
+
"val/clipfrac_avg": 0.0,
|
| 1016 |
+
"val/num_eos_tokens": 0,
|
| 1017 |
+
"val/ratio": 1.0009230375289917,
|
| 1018 |
+
"val/ratio_var": 1.998189873120282e-06
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"episode": 1176,
|
| 1022 |
+
"epoch": 0.168,
|
| 1023 |
+
"eps": 3,
|
| 1024 |
+
"loss/policy_avg": -0.0006489114603027701,
|
| 1025 |
+
"loss/value_avg": 0.001087805489078164,
|
| 1026 |
+
"lr": 9.89559096448414e-07,
|
| 1027 |
+
"objective/entropy": 11.260568618774414,
|
| 1028 |
+
"objective/kl": 0.0814318135380745,
|
| 1029 |
+
"objective/non_score_reward": -0.004071590956300497,
|
| 1030 |
+
"objective/rlhf_reward": -0.004071590956300497,
|
| 1031 |
+
"objective/scores": 0.0,
|
| 1032 |
+
"policy/approxkl_avg": 0.00044973386684432626,
|
| 1033 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 1034 |
+
"policy/entropy_avg": 0.1947825849056244,
|
| 1035 |
+
"step": 49,
|
| 1036 |
+
"val/clipfrac_avg": 0.0,
|
| 1037 |
+
"val/num_eos_tokens": 0,
|
| 1038 |
+
"val/ratio": 1.0001856088638306,
|
| 1039 |
+
"val/ratio_var": 8.380131362173415e-07
|
| 1040 |
+
},
|
| 1041 |
+
{
|
| 1042 |
+
"episode": 1200,
|
| 1043 |
+
"epoch": 0.17142857142857143,
|
| 1044 |
+
"eps": 3,
|
| 1045 |
+
"loss/policy_avg": -0.0005568858468905091,
|
| 1046 |
+
"loss/value_avg": 0.0010799263836815953,
|
| 1047 |
+
"lr": 9.883719313825227e-07,
|
| 1048 |
+
"objective/entropy": 12.4381742477417,
|
| 1049 |
+
"objective/kl": 0.00899545382708311,
|
| 1050 |
+
"objective/non_score_reward": -0.00044977269135415554,
|
| 1051 |
+
"objective/rlhf_reward": -0.00044977269135415554,
|
| 1052 |
+
"objective/scores": 0.0,
|
| 1053 |
+
"policy/approxkl_avg": 0.0008998862467706203,
|
| 1054 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1055 |
+
"policy/entropy_avg": 0.2059018909931183,
|
| 1056 |
+
"step": 50,
|
| 1057 |
+
"val/clipfrac_avg": 0.0,
|
| 1058 |
+
"val/num_eos_tokens": 0,
|
| 1059 |
+
"val/ratio": 0.9994922280311584,
|
| 1060 |
+
"val/ratio_var": 4.479263679968426e-06
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"episode": 1224,
|
| 1064 |
+
"epoch": 0.17485714285714285,
|
| 1065 |
+
"eps": 3,
|
| 1066 |
+
"loss/policy_avg": 0.00012879818677902222,
|
| 1067 |
+
"loss/value_avg": 0.0011688255472108722,
|
| 1068 |
+
"lr": 9.871217381465902e-07,
|
| 1069 |
+
"objective/entropy": 9.98705768585205,
|
| 1070 |
+
"objective/kl": -0.022387906908988953,
|
| 1071 |
+
"objective/non_score_reward": 0.0011193952523171902,
|
| 1072 |
+
"objective/rlhf_reward": 0.0011193952523171902,
|
| 1073 |
+
"objective/scores": 0.0,
|
| 1074 |
+
"policy/approxkl_avg": 0.000400608463678509,
|
| 1075 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1076 |
+
"policy/entropy_avg": 0.1886395812034607,
|
| 1077 |
+
"step": 51,
|
| 1078 |
+
"val/clipfrac_avg": 0.0,
|
| 1079 |
+
"val/num_eos_tokens": 0,
|
| 1080 |
+
"val/ratio": 0.9999980926513672,
|
| 1081 |
+
"val/ratio_var": 2.1912596821493935e-06
|
| 1082 |
+
},
|
| 1083 |
+
{
|
| 1084 |
+
"episode": 1248,
|
| 1085 |
+
"epoch": 0.1782857142857143,
|
| 1086 |
+
"eps": 3,
|
| 1087 |
+
"loss/policy_avg": -0.0006774354260414839,
|
| 1088 |
+
"loss/value_avg": 0.001140483422204852,
|
| 1089 |
+
"lr": 9.85808696490701e-07,
|
| 1090 |
+
"objective/entropy": 10.60033893585205,
|
| 1091 |
+
"objective/kl": -0.04120694845914841,
|
| 1092 |
+
"objective/non_score_reward": 0.002060347469523549,
|
| 1093 |
+
"objective/rlhf_reward": 0.002060347469523549,
|
| 1094 |
+
"objective/scores": 0.0,
|
| 1095 |
+
"policy/approxkl_avg": 0.0003651762963272631,
|
| 1096 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 1097 |
+
"policy/entropy_avg": 0.18694177269935608,
|
| 1098 |
+
"step": 52,
|
| 1099 |
+
"val/clipfrac_avg": 0.0,
|
| 1100 |
+
"val/num_eos_tokens": 0,
|
| 1101 |
+
"val/ratio": 1.0005837678909302,
|
| 1102 |
+
"val/ratio_var": 3.188570644852007e-06
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"episode": 1272,
|
| 1106 |
+
"epoch": 0.18171428571428572,
|
| 1107 |
+
"eps": 3,
|
| 1108 |
+
"loss/policy_avg": -0.0014181339647620916,
|
| 1109 |
+
"loss/value_avg": 0.0011729756370186806,
|
| 1110 |
+
"lr": 9.844329952011504e-07,
|
| 1111 |
+
"objective/entropy": 11.802154541015625,
|
| 1112 |
+
"objective/kl": 0.048689838498830795,
|
| 1113 |
+
"objective/non_score_reward": -0.0024344921112060547,
|
| 1114 |
+
"objective/rlhf_reward": -0.0024344921112060547,
|
| 1115 |
+
"objective/scores": 0.0,
|
| 1116 |
+
"policy/approxkl_avg": 0.00045496394159272313,
|
| 1117 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1118 |
+
"policy/entropy_avg": 0.21137630939483643,
|
| 1119 |
+
"step": 53,
|
| 1120 |
+
"val/clipfrac_avg": 0.0,
|
| 1121 |
+
"val/num_eos_tokens": 0,
|
| 1122 |
+
"val/ratio": 1.0002349615097046,
|
| 1123 |
+
"val/ratio_var": 1.731065481180849e-06
|
| 1124 |
+
},
|
| 1125 |
+
{
|
| 1126 |
+
"episode": 1296,
|
| 1127 |
+
"epoch": 0.18514285714285714,
|
| 1128 |
+
"eps": 3,
|
| 1129 |
+
"loss/policy_avg": -0.0007270214264281094,
|
| 1130 |
+
"loss/value_avg": 0.0010591228492558002,
|
| 1131 |
+
"lr": 9.829948320733e-07,
|
| 1132 |
+
"objective/entropy": 11.661518096923828,
|
| 1133 |
+
"objective/kl": 0.004553104750812054,
|
| 1134 |
+
"objective/non_score_reward": -0.00022765521134715527,
|
| 1135 |
+
"objective/rlhf_reward": -0.00022765521134715527,
|
| 1136 |
+
"objective/scores": 0.0,
|
| 1137 |
+
"policy/approxkl_avg": 0.000439059833297506,
|
| 1138 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 1139 |
+
"policy/entropy_avg": 0.19910752773284912,
|
| 1140 |
+
"step": 54,
|
| 1141 |
+
"val/clipfrac_avg": 0.0,
|
| 1142 |
+
"val/num_eos_tokens": 0,
|
| 1143 |
+
"val/ratio": 1.0010268688201904,
|
| 1144 |
+
"val/ratio_var": 2.5225742774637183e-06
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"episode": 1320,
|
| 1148 |
+
"epoch": 0.18857142857142858,
|
| 1149 |
+
"eps": 3,
|
| 1150 |
+
"loss/policy_avg": -0.0006840452551841736,
|
| 1151 |
+
"loss/value_avg": 0.0011034862836822867,
|
| 1152 |
+
"lr": 9.8149441388314e-07,
|
| 1153 |
+
"objective/entropy": 9.149593353271484,
|
| 1154 |
+
"objective/kl": 0.057699695229530334,
|
| 1155 |
+
"objective/non_score_reward": -0.002884984714910388,
|
| 1156 |
+
"objective/rlhf_reward": -0.002884984714910388,
|
| 1157 |
+
"objective/scores": 0.0,
|
| 1158 |
+
"policy/approxkl_avg": 0.00033010850893333554,
|
| 1159 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1160 |
+
"policy/entropy_avg": 0.1894957423210144,
|
| 1161 |
+
"step": 55,
|
| 1162 |
+
"val/clipfrac_avg": 0.0,
|
| 1163 |
+
"val/num_eos_tokens": 0,
|
| 1164 |
+
"val/ratio": 0.9992285966873169,
|
| 1165 |
+
"val/ratio_var": 8.357873753084277e-07
|
| 1166 |
+
},
|
| 1167 |
+
{
|
| 1168 |
+
"episode": 1344,
|
| 1169 |
+
"epoch": 0.192,
|
| 1170 |
+
"eps": 3,
|
| 1171 |
+
"loss/policy_avg": -0.0006162170320749283,
|
| 1172 |
+
"loss/value_avg": 0.0011356081813573837,
|
| 1173 |
+
"lr": 9.799319563575593e-07,
|
| 1174 |
+
"objective/entropy": 11.028936386108398,
|
| 1175 |
+
"objective/kl": -0.02875569462776184,
|
| 1176 |
+
"objective/non_score_reward": 0.0014377848710864782,
|
| 1177 |
+
"objective/rlhf_reward": 0.0014377848710864782,
|
| 1178 |
+
"objective/scores": 0.0,
|
| 1179 |
+
"policy/approxkl_avg": 0.00047001155326142907,
|
| 1180 |
+
"policy/clipfrac_avg": 0.001768867950886488,
|
| 1181 |
+
"policy/entropy_avg": 0.21663422882556915,
|
| 1182 |
+
"step": 56,
|
| 1183 |
+
"val/clipfrac_avg": 0.0,
|
| 1184 |
+
"val/num_eos_tokens": 0,
|
| 1185 |
+
"val/ratio": 1.000121831893921,
|
| 1186 |
+
"val/ratio_var": 3.489588834781898e-06
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"episode": 1368,
|
| 1190 |
+
"epoch": 0.19542857142857142,
|
| 1191 |
+
"eps": 3,
|
| 1192 |
+
"loss/policy_avg": 0.00045855995267629623,
|
| 1193 |
+
"loss/value_avg": 0.0011290146503597498,
|
| 1194 |
+
"lr": 9.783076841433279e-07,
|
| 1195 |
+
"objective/entropy": 10.055129051208496,
|
| 1196 |
+
"objective/kl": 0.09907305240631104,
|
| 1197 |
+
"objective/non_score_reward": -0.004953653551638126,
|
| 1198 |
+
"objective/rlhf_reward": -0.004953653551638126,
|
| 1199 |
+
"objective/scores": 0.0,
|
| 1200 |
+
"policy/approxkl_avg": 0.0004340106970630586,
|
| 1201 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 1202 |
+
"policy/entropy_avg": 0.180865079164505,
|
| 1203 |
+
"step": 57,
|
| 1204 |
+
"val/clipfrac_avg": 0.0,
|
| 1205 |
+
"val/num_eos_tokens": 0,
|
| 1206 |
+
"val/ratio": 0.9993646740913391,
|
| 1207 |
+
"val/ratio_var": 1.6994080169752124e-06
|
| 1208 |
+
},
|
| 1209 |
+
{
|
| 1210 |
+
"episode": 1392,
|
| 1211 |
+
"epoch": 0.19885714285714284,
|
| 1212 |
+
"eps": 3,
|
| 1213 |
+
"loss/policy_avg": -0.000757572939619422,
|
| 1214 |
+
"loss/value_avg": 0.001213478040881455,
|
| 1215 |
+
"lr": 9.76621830774799e-07,
|
| 1216 |
+
"objective/entropy": 10.346697807312012,
|
| 1217 |
+
"objective/kl": 0.06749838590621948,
|
| 1218 |
+
"objective/non_score_reward": -0.003374919295310974,
|
| 1219 |
+
"objective/rlhf_reward": -0.003374919295310974,
|
| 1220 |
+
"objective/scores": 0.0,
|
| 1221 |
+
"policy/approxkl_avg": 0.0004295073449611664,
|
| 1222 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 1223 |
+
"policy/entropy_avg": 0.18486541509628296,
|
| 1224 |
+
"step": 58,
|
| 1225 |
+
"val/clipfrac_avg": 0.0,
|
| 1226 |
+
"val/num_eos_tokens": 0,
|
| 1227 |
+
"val/ratio": 1.000084400177002,
|
| 1228 |
+
"val/ratio_var": 2.244758661618107e-06
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"episode": 1416,
|
| 1232 |
+
"epoch": 0.2022857142857143,
|
| 1233 |
+
"eps": 3,
|
| 1234 |
+
"loss/policy_avg": -0.00044112117029726505,
|
| 1235 |
+
"loss/value_avg": 0.001154930330812931,
|
| 1236 |
+
"lr": 9.748746386403305e-07,
|
| 1237 |
+
"objective/entropy": 11.446981430053711,
|
| 1238 |
+
"objective/kl": 0.04666760563850403,
|
| 1239 |
+
"objective/non_score_reward": -0.0023333802819252014,
|
| 1240 |
+
"objective/rlhf_reward": -0.0023333802819252014,
|
| 1241 |
+
"objective/scores": 0.0,
|
| 1242 |
+
"policy/approxkl_avg": 0.00045786722330376506,
|
| 1243 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 1244 |
+
"policy/entropy_avg": 0.19822391867637634,
|
| 1245 |
+
"step": 59,
|
| 1246 |
+
"val/clipfrac_avg": 0.0,
|
| 1247 |
+
"val/num_eos_tokens": 0,
|
| 1248 |
+
"val/ratio": 1.0004160404205322,
|
| 1249 |
+
"val/ratio_var": 1.9798999346676283e-06
|
| 1250 |
+
},
|
| 1251 |
+
{
|
| 1252 |
+
"episode": 1440,
|
| 1253 |
+
"epoch": 0.2057142857142857,
|
| 1254 |
+
"eps": 3,
|
| 1255 |
+
"loss/policy_avg": -0.0011781796347349882,
|
| 1256 |
+
"loss/value_avg": 0.0011596104595810175,
|
| 1257 |
+
"lr": 9.730663589474364e-07,
|
| 1258 |
+
"objective/entropy": 8.769947052001953,
|
| 1259 |
+
"objective/kl": 0.06850416958332062,
|
| 1260 |
+
"objective/non_score_reward": -0.0034252083860337734,
|
| 1261 |
+
"objective/rlhf_reward": -0.0034252083860337734,
|
| 1262 |
+
"objective/scores": 0.0,
|
| 1263 |
+
"policy/approxkl_avg": 0.00029326786170713603,
|
| 1264 |
+
"policy/clipfrac_avg": 0.0,
|
| 1265 |
+
"policy/entropy_avg": 0.18261712789535522,
|
| 1266 |
+
"step": 60,
|
| 1267 |
+
"val/clipfrac_avg": 0.0,
|
| 1268 |
+
"val/num_eos_tokens": 0,
|
| 1269 |
+
"val/ratio": 1.000569224357605,
|
| 1270 |
+
"val/ratio_var": 1.8464993445377331e-06
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"episode": 1464,
|
| 1274 |
+
"epoch": 0.20914285714285713,
|
| 1275 |
+
"eps": 3,
|
| 1276 |
+
"loss/policy_avg": -0.000520245754159987,
|
| 1277 |
+
"loss/value_avg": 0.001076899585314095,
|
| 1278 |
+
"lr": 9.711972516866678e-07,
|
| 1279 |
+
"objective/entropy": 10.513251304626465,
|
| 1280 |
+
"objective/kl": 0.008584271185100079,
|
| 1281 |
+
"objective/non_score_reward": -0.0004292135126888752,
|
| 1282 |
+
"objective/rlhf_reward": -0.0004292135126888752,
|
| 1283 |
+
"objective/scores": 0.0,
|
| 1284 |
+
"policy/approxkl_avg": 0.00044821258052252233,
|
| 1285 |
+
"policy/clipfrac_avg": 0.0015723269898444414,
|
| 1286 |
+
"policy/entropy_avg": 0.18403959274291992,
|
| 1287 |
+
"step": 61,
|
| 1288 |
+
"val/clipfrac_avg": 0.0,
|
| 1289 |
+
"val/num_eos_tokens": 0,
|
| 1290 |
+
"val/ratio": 1.0004669427871704,
|
| 1291 |
+
"val/ratio_var": 1.8683210782910464e-06
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"episode": 1488,
|
| 1295 |
+
"epoch": 0.21257142857142858,
|
| 1296 |
+
"eps": 3,
|
| 1297 |
+
"loss/policy_avg": -7.968657882884145e-05,
|
| 1298 |
+
"loss/value_avg": 0.0013330953661352396,
|
| 1299 |
+
"lr": 9.692675855942318e-07,
|
| 1300 |
+
"objective/entropy": 9.325455665588379,
|
| 1301 |
+
"objective/kl": -0.04262801632285118,
|
| 1302 |
+
"objective/non_score_reward": 0.002131400629878044,
|
| 1303 |
+
"objective/rlhf_reward": 0.002131400629878044,
|
| 1304 |
+
"objective/scores": 0.0,
|
| 1305 |
+
"policy/approxkl_avg": 0.0003892054664902389,
|
| 1306 |
+
"policy/clipfrac_avg": 0.0017688678344711661,
|
| 1307 |
+
"policy/entropy_avg": 0.1799222230911255,
|
| 1308 |
+
"step": 62,
|
| 1309 |
+
"val/clipfrac_avg": 0.0,
|
| 1310 |
+
"val/num_eos_tokens": 0,
|
| 1311 |
+
"val/ratio": 0.9998717308044434,
|
| 1312 |
+
"val/ratio_var": 2.2758265458833193e-06
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"episode": 1512,
|
| 1316 |
+
"epoch": 0.216,
|
| 1317 |
+
"eps": 3,
|
| 1318 |
+
"loss/policy_avg": -0.0009765516733750701,
|
| 1319 |
+
"loss/value_avg": 0.0009945188648998737,
|
| 1320 |
+
"lr": 9.67277638113354e-07,
|
| 1321 |
+
"objective/entropy": 13.20396614074707,
|
| 1322 |
+
"objective/kl": 0.04450948163866997,
|
| 1323 |
+
"objective/non_score_reward": -0.0022254744544625282,
|
| 1324 |
+
"objective/rlhf_reward": -0.0022254744544625282,
|
| 1325 |
+
"objective/scores": 0.0,
|
| 1326 |
+
"policy/approxkl_avg": 0.000418822281062603,
|
| 1327 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1328 |
+
"policy/entropy_avg": 0.24310021102428436,
|
| 1329 |
+
"step": 63,
|
| 1330 |
+
"val/clipfrac_avg": 0.0,
|
| 1331 |
+
"val/num_eos_tokens": 0,
|
| 1332 |
+
"val/ratio": 1.000614047050476,
|
| 1333 |
+
"val/ratio_var": 3.149192025375669e-06
|
| 1334 |
+
},
|
| 1335 |
+
{
|
| 1336 |
+
"episode": 1536,
|
| 1337 |
+
"epoch": 0.21942857142857142,
|
| 1338 |
+
"eps": 3,
|
| 1339 |
+
"loss/policy_avg": -0.00018674280727282166,
|
| 1340 |
+
"loss/value_avg": 0.0012268940918147564,
|
| 1341 |
+
"lr": 9.652276953543877e-07,
|
| 1342 |
+
"objective/entropy": 10.03724193572998,
|
| 1343 |
+
"objective/kl": 0.010379712097346783,
|
| 1344 |
+
"objective/non_score_reward": -0.0005189854418858886,
|
| 1345 |
+
"objective/rlhf_reward": -0.0005189854418858886,
|
| 1346 |
+
"objective/scores": 0.0,
|
| 1347 |
+
"policy/approxkl_avg": 0.000406760664191097,
|
| 1348 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1349 |
+
"policy/entropy_avg": 0.18875974416732788,
|
| 1350 |
+
"step": 64,
|
| 1351 |
+
"val/clipfrac_avg": 0.0,
|
| 1352 |
+
"val/num_eos_tokens": 0,
|
| 1353 |
+
"val/ratio": 0.9995108842849731,
|
| 1354 |
+
"val/ratio_var": 1.4013304507898283e-06
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"episode": 1560,
|
| 1358 |
+
"epoch": 0.22285714285714286,
|
| 1359 |
+
"eps": 3,
|
| 1360 |
+
"loss/policy_avg": 0.00012768094893544912,
|
| 1361 |
+
"loss/value_avg": 0.001114123035222292,
|
| 1362 |
+
"lr": 9.631180520536777e-07,
|
| 1363 |
+
"objective/entropy": 10.841609001159668,
|
| 1364 |
+
"objective/kl": 0.08112704008817673,
|
| 1365 |
+
"objective/non_score_reward": -0.004056352190673351,
|
| 1366 |
+
"objective/rlhf_reward": -0.004056352190673351,
|
| 1367 |
+
"objective/scores": 0.0,
|
| 1368 |
+
"policy/approxkl_avg": 0.00039341190131381154,
|
| 1369 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 1370 |
+
"policy/entropy_avg": 0.19926729798316956,
|
| 1371 |
+
"step": 65,
|
| 1372 |
+
"val/clipfrac_avg": 0.0,
|
| 1373 |
+
"val/num_eos_tokens": 0,
|
| 1374 |
+
"val/ratio": 0.999116063117981,
|
| 1375 |
+
"val/ratio_var": 2.176814632548485e-06
|
| 1376 |
+
},
|
| 1377 |
+
{
|
| 1378 |
+
"episode": 1584,
|
| 1379 |
+
"epoch": 0.22628571428571428,
|
| 1380 |
+
"eps": 3,
|
| 1381 |
+
"loss/policy_avg": 0.0002728139515966177,
|
| 1382 |
+
"loss/value_avg": 0.001176491379737854,
|
| 1383 |
+
"lr": 9.60949011531184e-07,
|
| 1384 |
+
"objective/entropy": 8.628952026367188,
|
| 1385 |
+
"objective/kl": 0.06317141652107239,
|
| 1386 |
+
"objective/non_score_reward": -0.0031585709657520056,
|
| 1387 |
+
"objective/rlhf_reward": -0.0031585709657520056,
|
| 1388 |
+
"objective/scores": 0.0,
|
| 1389 |
+
"policy/approxkl_avg": 0.00034158816561102867,
|
| 1390 |
+
"policy/clipfrac_avg": 0.0,
|
| 1391 |
+
"policy/entropy_avg": 0.17121586203575134,
|
| 1392 |
+
"step": 66,
|
| 1393 |
+
"val/clipfrac_avg": 0.0,
|
| 1394 |
+
"val/num_eos_tokens": 0,
|
| 1395 |
+
"val/ratio": 0.9997283220291138,
|
| 1396 |
+
"val/ratio_var": 1.2662626431847457e-06
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"episode": 1608,
|
| 1400 |
+
"epoch": 0.2297142857142857,
|
| 1401 |
+
"eps": 3,
|
| 1402 |
+
"loss/policy_avg": -0.0013626827858388424,
|
| 1403 |
+
"loss/value_avg": 0.001127548050135374,
|
| 1404 |
+
"lr": 9.587208856468713e-07,
|
| 1405 |
+
"objective/entropy": 10.164885520935059,
|
| 1406 |
+
"objective/kl": -0.05043753236532211,
|
| 1407 |
+
"objective/non_score_reward": 0.002521876711398363,
|
| 1408 |
+
"objective/rlhf_reward": 0.002521876711398363,
|
| 1409 |
+
"objective/scores": 0.0,
|
| 1410 |
+
"policy/approxkl_avg": 0.00039231160189956427,
|
| 1411 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 1412 |
+
"policy/entropy_avg": 0.18126381933689117,
|
| 1413 |
+
"step": 67,
|
| 1414 |
+
"val/clipfrac_avg": 0.0,
|
| 1415 |
+
"val/num_eos_tokens": 0,
|
| 1416 |
+
"val/ratio": 1.0005989074707031,
|
| 1417 |
+
"val/ratio_var": 3.3499827623018064e-06
|
| 1418 |
+
},
|
| 1419 |
+
{
|
| 1420 |
+
"episode": 1632,
|
| 1421 |
+
"epoch": 0.23314285714285715,
|
| 1422 |
+
"eps": 3,
|
| 1423 |
+
"loss/policy_avg": 0.00018104736227542162,
|
| 1424 |
+
"loss/value_avg": 0.001195235294289887,
|
| 1425 |
+
"lr": 9.564339947558697e-07,
|
| 1426 |
+
"objective/entropy": 8.992341995239258,
|
| 1427 |
+
"objective/kl": 0.0629781186580658,
|
| 1428 |
+
"objective/non_score_reward": -0.003148905700072646,
|
| 1429 |
+
"objective/rlhf_reward": -0.003148905700072646,
|
| 1430 |
+
"objective/scores": 0.0,
|
| 1431 |
+
"policy/approxkl_avg": 0.00031165953259915113,
|
| 1432 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 1433 |
+
"policy/entropy_avg": 0.1769479513168335,
|
| 1434 |
+
"step": 68,
|
| 1435 |
+
"val/clipfrac_avg": 0.0,
|
| 1436 |
+
"val/num_eos_tokens": 0,
|
| 1437 |
+
"val/ratio": 0.9993069171905518,
|
| 1438 |
+
"val/ratio_var": 2.6862960567086702e-06
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"episode": 1656,
|
| 1442 |
+
"epoch": 0.23657142857142857,
|
| 1443 |
+
"eps": 3,
|
| 1444 |
+
"loss/policy_avg": -0.0006377133540809155,
|
| 1445 |
+
"loss/value_avg": 0.001105936011299491,
|
| 1446 |
+
"lr": 9.540886676624145e-07,
|
| 1447 |
+
"objective/entropy": 9.024686813354492,
|
| 1448 |
+
"objective/kl": 0.07867498695850372,
|
| 1449 |
+
"objective/non_score_reward": -0.003933749161660671,
|
| 1450 |
+
"objective/rlhf_reward": -0.003933749161660671,
|
| 1451 |
+
"objective/scores": 0.0,
|
| 1452 |
+
"policy/approxkl_avg": 0.0004139389202464372,
|
| 1453 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 1454 |
+
"policy/entropy_avg": 0.18283474445343018,
|
| 1455 |
+
"step": 69,
|
| 1456 |
+
"val/clipfrac_avg": 0.0,
|
| 1457 |
+
"val/num_eos_tokens": 0,
|
| 1458 |
+
"val/ratio": 0.999620258808136,
|
| 1459 |
+
"val/ratio_var": 1.2224793408677215e-06
|
| 1460 |
+
},
|
| 1461 |
+
{
|
| 1462 |
+
"episode": 1680,
|
| 1463 |
+
"epoch": 0.24,
|
| 1464 |
+
"eps": 3,
|
| 1465 |
+
"loss/policy_avg": -0.00034369935747236013,
|
| 1466 |
+
"loss/value_avg": 0.0010239611146971583,
|
| 1467 |
+
"lr": 9.516852415725732e-07,
|
| 1468 |
+
"objective/entropy": 12.412653923034668,
|
| 1469 |
+
"objective/kl": 0.08225254714488983,
|
| 1470 |
+
"objective/non_score_reward": -0.004112627357244492,
|
| 1471 |
+
"objective/rlhf_reward": -0.004112627357244492,
|
| 1472 |
+
"objective/scores": 0.0,
|
| 1473 |
+
"policy/approxkl_avg": 0.0005226304056122899,
|
| 1474 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 1475 |
+
"policy/entropy_avg": 0.23866336047649384,
|
| 1476 |
+
"step": 70,
|
| 1477 |
+
"val/clipfrac_avg": 0.0,
|
| 1478 |
+
"val/num_eos_tokens": 0,
|
| 1479 |
+
"val/ratio": 1.0004711151123047,
|
| 1480 |
+
"val/ratio_var": 3.487248477540561e-06
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"episode": 1704,
|
| 1484 |
+
"epoch": 0.24342857142857144,
|
| 1485 |
+
"eps": 3,
|
| 1486 |
+
"loss/policy_avg": -0.0012513366527855396,
|
| 1487 |
+
"loss/value_avg": 0.0012419146951287985,
|
| 1488 |
+
"lr": 9.492240620457606e-07,
|
| 1489 |
+
"objective/entropy": 13.152254104614258,
|
| 1490 |
+
"objective/kl": 0.06805577129125595,
|
| 1491 |
+
"objective/non_score_reward": -0.0034027881920337677,
|
| 1492 |
+
"objective/rlhf_reward": -0.0034027881920337677,
|
| 1493 |
+
"objective/scores": 0.0,
|
| 1494 |
+
"policy/approxkl_avg": 0.0005608252249658108,
|
| 1495 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 1496 |
+
"policy/entropy_avg": 0.2300390750169754,
|
| 1497 |
+
"step": 71,
|
| 1498 |
+
"val/clipfrac_avg": 0.0,
|
| 1499 |
+
"val/num_eos_tokens": 0,
|
| 1500 |
+
"val/ratio": 1.0004956722259521,
|
| 1501 |
+
"val/ratio_var": 2.79126152236131e-06
|
| 1502 |
+
},
|
| 1503 |
+
{
|
| 1504 |
+
"episode": 1728,
|
| 1505 |
+
"epoch": 0.24685714285714286,
|
| 1506 |
+
"eps": 3,
|
| 1507 |
+
"loss/policy_avg": -0.0006021471926942468,
|
| 1508 |
+
"loss/value_avg": 0.0010983510874211788,
|
| 1509 |
+
"lr": 9.467054829450571e-07,
|
| 1510 |
+
"objective/entropy": 11.628573417663574,
|
| 1511 |
+
"objective/kl": 0.0897781252861023,
|
| 1512 |
+
"objective/non_score_reward": -0.004488906357437372,
|
| 1513 |
+
"objective/rlhf_reward": -0.004488906357437372,
|
| 1514 |
+
"objective/scores": 0.0,
|
| 1515 |
+
"policy/approxkl_avg": 0.0004015603626612574,
|
| 1516 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1517 |
+
"policy/entropy_avg": 0.20409011840820312,
|
| 1518 |
+
"step": 72,
|
| 1519 |
+
"val/clipfrac_avg": 0.0,
|
| 1520 |
+
"val/num_eos_tokens": 0,
|
| 1521 |
+
"val/ratio": 1.000084638595581,
|
| 1522 |
+
"val/ratio_var": 3.225579803256551e-06
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"episode": 1752,
|
| 1526 |
+
"epoch": 0.2502857142857143,
|
| 1527 |
+
"eps": 3,
|
| 1528 |
+
"loss/policy_avg": -0.00017274607671424747,
|
| 1529 |
+
"loss/value_avg": 0.0011638551950454712,
|
| 1530 |
+
"lr": 9.441298663863289e-07,
|
| 1531 |
+
"objective/entropy": 10.489394187927246,
|
| 1532 |
+
"objective/kl": 0.05405331775546074,
|
| 1533 |
+
"objective/non_score_reward": -0.0027026659809052944,
|
| 1534 |
+
"objective/rlhf_reward": -0.0027026659809052944,
|
| 1535 |
+
"objective/scores": 0.0,
|
| 1536 |
+
"policy/approxkl_avg": 0.0003594418230932206,
|
| 1537 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 1538 |
+
"policy/entropy_avg": 0.19105064868927002,
|
| 1539 |
+
"step": 73,
|
| 1540 |
+
"val/clipfrac_avg": 0.0,
|
| 1541 |
+
"val/num_eos_tokens": 0,
|
| 1542 |
+
"val/ratio": 1.0002226829528809,
|
| 1543 |
+
"val/ratio_var": 1.964122702702298e-06
|
| 1544 |
+
},
|
| 1545 |
+
{
|
| 1546 |
+
"episode": 1776,
|
| 1547 |
+
"epoch": 0.2537142857142857,
|
| 1548 |
+
"eps": 3,
|
| 1549 |
+
"loss/policy_avg": 4.641339182853699e-05,
|
| 1550 |
+
"loss/value_avg": 0.0012347290758043528,
|
| 1551 |
+
"lr": 9.414975826861651e-07,
|
| 1552 |
+
"objective/entropy": 8.907478332519531,
|
| 1553 |
+
"objective/kl": 0.04718625172972679,
|
| 1554 |
+
"objective/non_score_reward": -0.0023593127261847258,
|
| 1555 |
+
"objective/rlhf_reward": -0.0023593127261847258,
|
| 1556 |
+
"objective/scores": 0.0,
|
| 1557 |
+
"policy/approxkl_avg": 0.0003988428507000208,
|
| 1558 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1559 |
+
"policy/entropy_avg": 0.19814759492874146,
|
| 1560 |
+
"step": 74,
|
| 1561 |
+
"val/clipfrac_avg": 0.0,
|
| 1562 |
+
"val/num_eos_tokens": 0,
|
| 1563 |
+
"val/ratio": 1.000138521194458,
|
| 1564 |
+
"val/ratio_var": 1.9056941482631373e-06
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"episode": 1800,
|
| 1568 |
+
"epoch": 0.2571428571428571,
|
| 1569 |
+
"eps": 3,
|
| 1570 |
+
"loss/policy_avg": -0.0015050144866108894,
|
| 1571 |
+
"loss/value_avg": 0.0012774434871971607,
|
| 1572 |
+
"lr": 9.388090103086343e-07,
|
| 1573 |
+
"objective/entropy": 13.666069030761719,
|
| 1574 |
+
"objective/kl": 0.2161487638950348,
|
| 1575 |
+
"objective/non_score_reward": -0.01080743782222271,
|
| 1576 |
+
"objective/rlhf_reward": -0.01080743782222271,
|
| 1577 |
+
"objective/scores": 0.0,
|
| 1578 |
+
"policy/approxkl_avg": 0.0012116986326873302,
|
| 1579 |
+
"policy/clipfrac_avg": 0.003144653979688883,
|
| 1580 |
+
"policy/entropy_avg": 0.24723494052886963,
|
| 1581 |
+
"step": 75,
|
| 1582 |
+
"val/clipfrac_avg": 0.0,
|
| 1583 |
+
"val/num_eos_tokens": 0,
|
| 1584 |
+
"val/ratio": 0.9994445443153381,
|
| 1585 |
+
"val/ratio_var": 2.9358038773352746e-06
|
| 1586 |
+
},
|
| 1587 |
+
{
|
| 1588 |
+
"episode": 1824,
|
| 1589 |
+
"epoch": 0.26057142857142856,
|
| 1590 |
+
"eps": 3,
|
| 1591 |
+
"loss/policy_avg": -0.00010800664313137531,
|
| 1592 |
+
"loss/value_avg": 0.0011818609200417995,
|
| 1593 |
+
"lr": 9.360645358108695e-07,
|
| 1594 |
+
"objective/entropy": 8.11184310913086,
|
| 1595 |
+
"objective/kl": 0.04297180101275444,
|
| 1596 |
+
"objective/non_score_reward": -0.0021485905162990093,
|
| 1597 |
+
"objective/rlhf_reward": -0.0021485905162990093,
|
| 1598 |
+
"objective/scores": 0.0,
|
| 1599 |
+
"policy/approxkl_avg": 0.0003635706380009651,
|
| 1600 |
+
"policy/clipfrac_avg": 0.0017688678344711661,
|
| 1601 |
+
"policy/entropy_avg": 0.17554748058319092,
|
| 1602 |
+
"step": 76,
|
| 1603 |
+
"val/clipfrac_avg": 0.0,
|
| 1604 |
+
"val/num_eos_tokens": 0,
|
| 1605 |
+
"val/ratio": 1.0001165866851807,
|
| 1606 |
+
"val/ratio_var": 1.6492232361997594e-06
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"episode": 1848,
|
| 1610 |
+
"epoch": 0.264,
|
| 1611 |
+
"eps": 3,
|
| 1612 |
+
"loss/policy_avg": -0.000423701130785048,
|
| 1613 |
+
"loss/value_avg": 0.0011581408325582743,
|
| 1614 |
+
"lr": 9.332645537874899e-07,
|
| 1615 |
+
"objective/entropy": 9.99191951751709,
|
| 1616 |
+
"objective/kl": -0.004131610505282879,
|
| 1617 |
+
"objective/non_score_reward": 0.00020658024004660547,
|
| 1618 |
+
"objective/rlhf_reward": 0.00020658024004660547,
|
| 1619 |
+
"objective/scores": 0.0,
|
| 1620 |
+
"policy/approxkl_avg": 0.00039624073542654514,
|
| 1621 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 1622 |
+
"policy/entropy_avg": 0.18171575665473938,
|
| 1623 |
+
"step": 77,
|
| 1624 |
+
"val/clipfrac_avg": 0.0,
|
| 1625 |
+
"val/num_eos_tokens": 0,
|
| 1626 |
+
"val/ratio": 1.0006225109100342,
|
| 1627 |
+
"val/ratio_var": 2.7435085030447226e-06
|
| 1628 |
+
},
|
| 1629 |
+
{
|
| 1630 |
+
"episode": 1872,
|
| 1631 |
+
"epoch": 0.2674285714285714,
|
| 1632 |
+
"eps": 3,
|
| 1633 |
+
"loss/policy_avg": -0.0017575371311977506,
|
| 1634 |
+
"loss/value_avg": 0.0011438459623605013,
|
| 1635 |
+
"lr": 9.304094668138669e-07,
|
| 1636 |
+
"objective/entropy": 10.796030044555664,
|
| 1637 |
+
"objective/kl": 0.04405898600816727,
|
| 1638 |
+
"objective/non_score_reward": -0.002202949021011591,
|
| 1639 |
+
"objective/rlhf_reward": -0.002202949021011591,
|
| 1640 |
+
"objective/scores": 0.0,
|
| 1641 |
+
"policy/approxkl_avg": 0.0003987160453107208,
|
| 1642 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 1643 |
+
"policy/entropy_avg": 0.20184190571308136,
|
| 1644 |
+
"step": 78,
|
| 1645 |
+
"val/clipfrac_avg": 0.0,
|
| 1646 |
+
"val/num_eos_tokens": 0,
|
| 1647 |
+
"val/ratio": 1.0002810955047607,
|
| 1648 |
+
"val/ratio_var": 3.7199308735580416e-06
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"episode": 1896,
|
| 1652 |
+
"epoch": 0.27085714285714285,
|
| 1653 |
+
"eps": 3,
|
| 1654 |
+
"loss/policy_avg": -0.0006982747581787407,
|
| 1655 |
+
"loss/value_avg": 0.001091040438041091,
|
| 1656 |
+
"lr": 9.274996853882425e-07,
|
| 1657 |
+
"objective/entropy": 11.000476837158203,
|
| 1658 |
+
"objective/kl": 0.0654320940375328,
|
| 1659 |
+
"objective/non_score_reward": -0.0032716048881411552,
|
| 1660 |
+
"objective/rlhf_reward": -0.0032716048881411552,
|
| 1661 |
+
"objective/scores": 0.0,
|
| 1662 |
+
"policy/approxkl_avg": 0.000344442087225616,
|
| 1663 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 1664 |
+
"policy/entropy_avg": 0.21253754198551178,
|
| 1665 |
+
"step": 79,
|
| 1666 |
+
"val/clipfrac_avg": 0.0,
|
| 1667 |
+
"val/num_eos_tokens": 0,
|
| 1668 |
+
"val/ratio": 0.9996973872184753,
|
| 1669 |
+
"val/ratio_var": 1.350840307168255e-06
|
| 1670 |
+
},
|
| 1671 |
+
{
|
| 1672 |
+
"episode": 1920,
|
| 1673 |
+
"epoch": 0.2742857142857143,
|
| 1674 |
+
"eps": 3,
|
| 1675 |
+
"loss/policy_avg": -0.00022802106104791164,
|
| 1676 |
+
"loss/value_avg": 0.0011568802874535322,
|
| 1677 |
+
"lr": 9.245356278727093e-07,
|
| 1678 |
+
"objective/entropy": 12.402453422546387,
|
| 1679 |
+
"objective/kl": 0.017512556165456772,
|
| 1680 |
+
"objective/non_score_reward": -0.0008756277966313064,
|
| 1681 |
+
"objective/rlhf_reward": -0.0008756277966313064,
|
| 1682 |
+
"objective/scores": 0.0,
|
| 1683 |
+
"policy/approxkl_avg": 0.0004438001196831465,
|
| 1684 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 1685 |
+
"policy/entropy_avg": 0.21272364258766174,
|
| 1686 |
+
"step": 80,
|
| 1687 |
+
"val/clipfrac_avg": 0.0,
|
| 1688 |
+
"val/num_eos_tokens": 0,
|
| 1689 |
+
"val/ratio": 1.0009765625,
|
| 1690 |
+
"val/ratio_var": 1.2678815437539015e-06
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"episode": 1944,
|
| 1694 |
+
"epoch": 0.2777142857142857,
|
| 1695 |
+
"eps": 3,
|
| 1696 |
+
"loss/policy_avg": -0.0021045708563178778,
|
| 1697 |
+
"loss/value_avg": 0.0011669377563521266,
|
| 1698 |
+
"lr": 9.215177204330587e-07,
|
| 1699 |
+
"objective/entropy": 11.159307479858398,
|
| 1700 |
+
"objective/kl": -0.03767850622534752,
|
| 1701 |
+
"objective/non_score_reward": 0.0018839254043996334,
|
| 1702 |
+
"objective/rlhf_reward": 0.0018839254043996334,
|
| 1703 |
+
"objective/scores": 0.0,
|
| 1704 |
+
"policy/approxkl_avg": 0.00048406008863821626,
|
| 1705 |
+
"policy/clipfrac_avg": 0.002555031329393387,
|
| 1706 |
+
"policy/entropy_avg": 0.2017732560634613,
|
| 1707 |
+
"step": 81,
|
| 1708 |
+
"val/clipfrac_avg": 0.0,
|
| 1709 |
+
"val/num_eos_tokens": 0,
|
| 1710 |
+
"val/ratio": 1.001840353012085,
|
| 1711 |
+
"val/ratio_var": 3.842494152195286e-06
|
| 1712 |
+
},
|
| 1713 |
+
{
|
| 1714 |
+
"episode": 1968,
|
| 1715 |
+
"epoch": 0.28114285714285714,
|
| 1716 |
+
"eps": 3,
|
| 1717 |
+
"loss/policy_avg": 0.000858477724250406,
|
| 1718 |
+
"loss/value_avg": 0.0012590258847922087,
|
| 1719 |
+
"lr": 9.184463969775083e-07,
|
| 1720 |
+
"objective/entropy": 9.627300262451172,
|
| 1721 |
+
"objective/kl": -0.05987191200256348,
|
| 1722 |
+
"objective/non_score_reward": 0.002993595553562045,
|
| 1723 |
+
"objective/rlhf_reward": 0.002993595553562045,
|
| 1724 |
+
"objective/scores": 0.0,
|
| 1725 |
+
"policy/approxkl_avg": 0.00036076403921470046,
|
| 1726 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 1727 |
+
"policy/entropy_avg": 0.188791424036026,
|
| 1728 |
+
"step": 82,
|
| 1729 |
+
"val/clipfrac_avg": 0.0,
|
| 1730 |
+
"val/num_eos_tokens": 0,
|
| 1731 |
+
"val/ratio": 1.0008705854415894,
|
| 1732 |
+
"val/ratio_var": 2.1515820662898477e-06
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"episode": 1992,
|
| 1736 |
+
"epoch": 0.2845714285714286,
|
| 1737 |
+
"eps": 3,
|
| 1738 |
+
"loss/policy_avg": 0.00014637189451605082,
|
| 1739 |
+
"loss/value_avg": 0.0012444497551769018,
|
| 1740 |
+
"lr": 9.153220990943145e-07,
|
| 1741 |
+
"objective/entropy": 12.66540813446045,
|
| 1742 |
+
"objective/kl": -0.0041860491037368774,
|
| 1743 |
+
"objective/non_score_reward": 0.000209302845178172,
|
| 1744 |
+
"objective/rlhf_reward": 0.000209302845178172,
|
| 1745 |
+
"objective/scores": 0.0,
|
| 1746 |
+
"policy/approxkl_avg": 0.0004341425374150276,
|
| 1747 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1748 |
+
"policy/entropy_avg": 0.22873584926128387,
|
| 1749 |
+
"step": 83,
|
| 1750 |
+
"val/clipfrac_avg": 0.0,
|
| 1751 |
+
"val/num_eos_tokens": 0,
|
| 1752 |
+
"val/ratio": 0.9995162487030029,
|
| 1753 |
+
"val/ratio_var": 4.206229277770035e-06
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"episode": 2016,
|
| 1757 |
+
"epoch": 0.288,
|
| 1758 |
+
"eps": 3,
|
| 1759 |
+
"loss/policy_avg": -0.0008962788851931691,
|
| 1760 |
+
"loss/value_avg": 0.0012219806667417288,
|
| 1761 |
+
"lr": 9.121452759882831e-07,
|
| 1762 |
+
"objective/entropy": 9.737704277038574,
|
| 1763 |
+
"objective/kl": 0.05109195038676262,
|
| 1764 |
+
"objective/non_score_reward": -0.0025545977987349033,
|
| 1765 |
+
"objective/rlhf_reward": -0.0025545977987349033,
|
| 1766 |
+
"objective/scores": 0.0,
|
| 1767 |
+
"policy/approxkl_avg": 0.00039628270315006375,
|
| 1768 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1769 |
+
"policy/entropy_avg": 0.19647139310836792,
|
| 1770 |
+
"step": 84,
|
| 1771 |
+
"val/clipfrac_avg": 0.0,
|
| 1772 |
+
"val/num_eos_tokens": 0,
|
| 1773 |
+
"val/ratio": 0.9999145269393921,
|
| 1774 |
+
"val/ratio_var": 2.936012378995656e-06
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"episode": 2040,
|
| 1778 |
+
"epoch": 0.2914285714285714,
|
| 1779 |
+
"eps": 3,
|
| 1780 |
+
"loss/policy_avg": -0.001591079868376255,
|
| 1781 |
+
"loss/value_avg": 0.0010952729498967528,
|
| 1782 |
+
"lr": 9.08916384416183e-07,
|
| 1783 |
+
"objective/entropy": 11.758407592773438,
|
| 1784 |
+
"objective/kl": 0.14314039051532745,
|
| 1785 |
+
"objective/non_score_reward": -0.007157019339501858,
|
| 1786 |
+
"objective/rlhf_reward": -0.007157019339501858,
|
| 1787 |
+
"objective/scores": 0.0,
|
| 1788 |
+
"policy/approxkl_avg": 0.00046683146501891315,
|
| 1789 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 1790 |
+
"policy/entropy_avg": 0.2205151468515396,
|
| 1791 |
+
"step": 85,
|
| 1792 |
+
"val/clipfrac_avg": 0.0,
|
| 1793 |
+
"val/num_eos_tokens": 0,
|
| 1794 |
+
"val/ratio": 1.0003023147583008,
|
| 1795 |
+
"val/ratio_var": 2.956066737169749e-06
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"episode": 2064,
|
| 1799 |
+
"epoch": 0.2948571428571429,
|
| 1800 |
+
"eps": 3,
|
| 1801 |
+
"loss/policy_avg": -0.0015333988703787327,
|
| 1802 |
+
"loss/value_avg": 0.001151310047134757,
|
| 1803 |
+
"lr": 9.056358886210747e-07,
|
| 1804 |
+
"objective/entropy": 12.441439628601074,
|
| 1805 |
+
"objective/kl": 0.08128754049539566,
|
| 1806 |
+
"objective/non_score_reward": -0.0040643769316375256,
|
| 1807 |
+
"objective/rlhf_reward": -0.0040643769316375256,
|
| 1808 |
+
"objective/scores": 0.0,
|
| 1809 |
+
"policy/approxkl_avg": 0.0005020004464313388,
|
| 1810 |
+
"policy/clipfrac_avg": 0.001965408679097891,
|
| 1811 |
+
"policy/entropy_avg": 0.22599831223487854,
|
| 1812 |
+
"step": 86,
|
| 1813 |
+
"val/clipfrac_avg": 0.0,
|
| 1814 |
+
"val/num_eos_tokens": 0,
|
| 1815 |
+
"val/ratio": 0.9994167685508728,
|
| 1816 |
+
"val/ratio_var": 3.2662399007676868e-06
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"episode": 2088,
|
| 1820 |
+
"epoch": 0.29828571428571427,
|
| 1821 |
+
"eps": 3,
|
| 1822 |
+
"loss/policy_avg": 0.00033648847602307796,
|
| 1823 |
+
"loss/value_avg": 0.001143318135291338,
|
| 1824 |
+
"lr": 9.023042602655623e-07,
|
| 1825 |
+
"objective/entropy": 9.817390441894531,
|
| 1826 |
+
"objective/kl": -0.09843956679105759,
|
| 1827 |
+
"objective/non_score_reward": 0.004921978339552879,
|
| 1828 |
+
"objective/rlhf_reward": 0.004921978339552879,
|
| 1829 |
+
"objective/scores": 0.0,
|
| 1830 |
+
"policy/approxkl_avg": 0.00033533855457790196,
|
| 1831 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1832 |
+
"policy/entropy_avg": 0.19580014050006866,
|
| 1833 |
+
"step": 87,
|
| 1834 |
+
"val/clipfrac_avg": 0.0,
|
| 1835 |
+
"val/num_eos_tokens": 0,
|
| 1836 |
+
"val/ratio": 1.0008838176727295,
|
| 1837 |
+
"val/ratio_var": 1.9849708223773632e-06
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"episode": 2112,
|
| 1841 |
+
"epoch": 0.3017142857142857,
|
| 1842 |
+
"eps": 3,
|
| 1843 |
+
"loss/policy_avg": 0.00017705616482999176,
|
| 1844 |
+
"loss/value_avg": 0.0010654886718839407,
|
| 1845 |
+
"lr": 8.989219783639795e-07,
|
| 1846 |
+
"objective/entropy": 9.474151611328125,
|
| 1847 |
+
"objective/kl": 0.08406557887792587,
|
| 1848 |
+
"objective/non_score_reward": -0.004203279037028551,
|
| 1849 |
+
"objective/rlhf_reward": -0.004203279037028551,
|
| 1850 |
+
"objective/scores": 0.0,
|
| 1851 |
+
"policy/approxkl_avg": 0.0003710829478222877,
|
| 1852 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 1853 |
+
"policy/entropy_avg": 0.1872844099998474,
|
| 1854 |
+
"step": 88,
|
| 1855 |
+
"val/clipfrac_avg": 0.0,
|
| 1856 |
+
"val/num_eos_tokens": 0,
|
| 1857 |
+
"val/ratio": 1.0003993511199951,
|
| 1858 |
+
"val/ratio_var": 1.4447776948145474e-06
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"episode": 2136,
|
| 1862 |
+
"epoch": 0.30514285714285716,
|
| 1863 |
+
"eps": 3,
|
| 1864 |
+
"loss/policy_avg": -0.0004872347926720977,
|
| 1865 |
+
"loss/value_avg": 0.001096963882446289,
|
| 1866 |
+
"lr": 8.95489529213517e-07,
|
| 1867 |
+
"objective/entropy": 9.022795677185059,
|
| 1868 |
+
"objective/kl": 0.027196139097213745,
|
| 1869 |
+
"objective/non_score_reward": -0.0013598070945590734,
|
| 1870 |
+
"objective/rlhf_reward": -0.0013598070945590734,
|
| 1871 |
+
"objective/scores": 0.0,
|
| 1872 |
+
"policy/approxkl_avg": 0.0003935418208129704,
|
| 1873 |
+
"policy/clipfrac_avg": 0.001965408679097891,
|
| 1874 |
+
"policy/entropy_avg": 0.1806870847940445,
|
| 1875 |
+
"step": 89,
|
| 1876 |
+
"val/clipfrac_avg": 0.0,
|
| 1877 |
+
"val/num_eos_tokens": 0,
|
| 1878 |
+
"val/ratio": 0.999650239944458,
|
| 1879 |
+
"val/ratio_var": 1.598498329258291e-06
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"episode": 2160,
|
| 1883 |
+
"epoch": 0.30857142857142855,
|
| 1884 |
+
"eps": 3,
|
| 1885 |
+
"loss/policy_avg": 6.0095335356891155e-05,
|
| 1886 |
+
"loss/value_avg": 0.0011415532790124416,
|
| 1887 |
+
"lr": 8.920074063243045e-07,
|
| 1888 |
+
"objective/entropy": 10.911069869995117,
|
| 1889 |
+
"objective/kl": 0.06437133252620697,
|
| 1890 |
+
"objective/non_score_reward": -0.003218566533178091,
|
| 1891 |
+
"objective/rlhf_reward": -0.003218566533178091,
|
| 1892 |
+
"objective/scores": 0.0,
|
| 1893 |
+
"policy/approxkl_avg": 0.0003742141416296363,
|
| 1894 |
+
"policy/clipfrac_avg": 0.001179245300590992,
|
| 1895 |
+
"policy/entropy_avg": 0.18816140294075012,
|
| 1896 |
+
"step": 90,
|
| 1897 |
+
"val/clipfrac_avg": 0.0,
|
| 1898 |
+
"val/num_eos_tokens": 0,
|
| 1899 |
+
"val/ratio": 0.9992557764053345,
|
| 1900 |
+
"val/ratio_var": 9.944599241862306e-07
|
| 1901 |
+
},
|
| 1902 |
+
{
|
| 1903 |
+
"episode": 2184,
|
| 1904 |
+
"epoch": 0.312,
|
| 1905 |
+
"eps": 3,
|
| 1906 |
+
"loss/policy_avg": 2.5081797502934933e-05,
|
| 1907 |
+
"loss/value_avg": 0.0012206505052745342,
|
| 1908 |
+
"lr": 8.884761103484547e-07,
|
| 1909 |
+
"objective/entropy": 9.843175888061523,
|
| 1910 |
+
"objective/kl": 0.03545796871185303,
|
| 1911 |
+
"objective/non_score_reward": -0.0017728982493281364,
|
| 1912 |
+
"objective/rlhf_reward": -0.0017728982493281364,
|
| 1913 |
+
"objective/scores": 0.0,
|
| 1914 |
+
"policy/approxkl_avg": 0.0003428169875405729,
|
| 1915 |
+
"policy/clipfrac_avg": 0.0,
|
| 1916 |
+
"policy/entropy_avg": 0.1816243827342987,
|
| 1917 |
+
"step": 91,
|
| 1918 |
+
"val/clipfrac_avg": 0.0,
|
| 1919 |
+
"val/num_eos_tokens": 0,
|
| 1920 |
+
"val/ratio": 1.0000550746917725,
|
| 1921 |
+
"val/ratio_var": 7.711231546636554e-07
|
| 1922 |
+
},
|
| 1923 |
+
{
|
| 1924 |
+
"episode": 2208,
|
| 1925 |
+
"epoch": 0.31542857142857145,
|
| 1926 |
+
"eps": 3,
|
| 1927 |
+
"loss/policy_avg": -0.0016599048394709826,
|
| 1928 |
+
"loss/value_avg": 0.0011678379960358143,
|
| 1929 |
+
"lr": 8.848961490080805e-07,
|
| 1930 |
+
"objective/entropy": 10.757587432861328,
|
| 1931 |
+
"objective/kl": 0.02934662066400051,
|
| 1932 |
+
"objective/non_score_reward": -0.0014673307305201888,
|
| 1933 |
+
"objective/rlhf_reward": -0.0014673307305201888,
|
| 1934 |
+
"objective/scores": 0.0,
|
| 1935 |
+
"policy/approxkl_avg": 0.00047666137106716633,
|
| 1936 |
+
"policy/clipfrac_avg": 0.0027515720576047897,
|
| 1937 |
+
"policy/entropy_avg": 0.21150188148021698,
|
| 1938 |
+
"step": 92,
|
| 1939 |
+
"val/clipfrac_avg": 0.0,
|
| 1940 |
+
"val/num_eos_tokens": 0,
|
| 1941 |
+
"val/ratio": 1.0001424551010132,
|
| 1942 |
+
"val/ratio_var": 2.449437033646973e-06
|
| 1943 |
+
},
|
| 1944 |
+
{
|
| 1945 |
+
"episode": 2232,
|
| 1946 |
+
"epoch": 0.31885714285714284,
|
| 1947 |
+
"eps": 3,
|
| 1948 |
+
"loss/policy_avg": -0.00034765334567055106,
|
| 1949 |
+
"loss/value_avg": 0.0010505674872547388,
|
| 1950 |
+
"lr": 8.81268037022296e-07,
|
| 1951 |
+
"objective/entropy": 10.09364128112793,
|
| 1952 |
+
"objective/kl": 0.052005209028720856,
|
| 1953 |
+
"objective/non_score_reward": -0.002600260078907013,
|
| 1954 |
+
"objective/rlhf_reward": -0.002600260078907013,
|
| 1955 |
+
"objective/scores": 0.0,
|
| 1956 |
+
"policy/approxkl_avg": 0.00031719336402602494,
|
| 1957 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 1958 |
+
"policy/entropy_avg": 0.17978450655937195,
|
| 1959 |
+
"step": 93,
|
| 1960 |
+
"val/clipfrac_avg": 0.0,
|
| 1961 |
+
"val/num_eos_tokens": 0,
|
| 1962 |
+
"val/ratio": 0.9992406964302063,
|
| 1963 |
+
"val/ratio_var": 1.4672579027319443e-06
|
| 1964 |
+
},
|
| 1965 |
+
{
|
| 1966 |
+
"episode": 2256,
|
| 1967 |
+
"epoch": 0.3222857142857143,
|
| 1968 |
+
"eps": 3,
|
| 1969 |
+
"loss/policy_avg": -0.00022107455879449844,
|
| 1970 |
+
"loss/value_avg": 0.0010250682244077325,
|
| 1971 |
+
"lr": 8.775922960332108e-07,
|
| 1972 |
+
"objective/entropy": 10.98604965209961,
|
| 1973 |
+
"objective/kl": 0.036630094051361084,
|
| 1974 |
+
"objective/non_score_reward": -0.0018315049819648266,
|
| 1975 |
+
"objective/rlhf_reward": -0.0018315049819648266,
|
| 1976 |
+
"objective/scores": 0.0,
|
| 1977 |
+
"policy/approxkl_avg": 0.0004163292469456792,
|
| 1978 |
+
"policy/clipfrac_avg": 0.0007861634949222207,
|
| 1979 |
+
"policy/entropy_avg": 0.19221962988376617,
|
| 1980 |
+
"step": 94,
|
| 1981 |
+
"val/clipfrac_avg": 0.0,
|
| 1982 |
+
"val/num_eos_tokens": 0,
|
| 1983 |
+
"val/ratio": 1.0002532005310059,
|
| 1984 |
+
"val/ratio_var": 3.751965323317563e-06
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"episode": 2280,
|
| 1988 |
+
"epoch": 0.32571428571428573,
|
| 1989 |
+
"eps": 3,
|
| 1990 |
+
"loss/policy_avg": -0.001208254136145115,
|
| 1991 |
+
"loss/value_avg": 0.0010296719847247005,
|
| 1992 |
+
"lr": 8.738694545309298e-07,
|
| 1993 |
+
"objective/entropy": 9.24979019165039,
|
| 1994 |
+
"objective/kl": 0.005700349807739258,
|
| 1995 |
+
"objective/non_score_reward": -0.00028501744964160025,
|
| 1996 |
+
"objective/rlhf_reward": -0.00028501744964160025,
|
| 1997 |
+
"objective/scores": 0.0,
|
| 1998 |
+
"policy/approxkl_avg": 0.0003547874221112579,
|
| 1999 |
+
"policy/clipfrac_avg": 0.00039308174746111035,
|
| 2000 |
+
"policy/entropy_avg": 0.1905178278684616,
|
| 2001 |
+
"step": 95,
|
| 2002 |
+
"val/clipfrac_avg": 0.0,
|
| 2003 |
+
"val/num_eos_tokens": 0,
|
| 2004 |
+
"val/ratio": 0.9988124370574951,
|
| 2005 |
+
"val/ratio_var": 2.4154353468475165e-06
|
| 2006 |
+
},
|
| 2007 |
+
{
|
| 2008 |
+
"episode": 2304,
|
| 2009 |
+
"epoch": 0.3291428571428571,
|
| 2010 |
+
"eps": 3,
|
| 2011 |
+
"loss/policy_avg": -0.0005715511506423354,
|
| 2012 |
+
"loss/value_avg": 0.0010314981918781996,
|
| 2013 |
+
"lr": 8.701000477775687e-07,
|
| 2014 |
+
"objective/entropy": 10.696990966796875,
|
| 2015 |
+
"objective/kl": 0.015906011685729027,
|
| 2016 |
+
"objective/non_score_reward": -0.0007953002932481468,
|
| 2017 |
+
"objective/rlhf_reward": -0.0007953002932481468,
|
| 2018 |
+
"objective/scores": 0.0,
|
| 2019 |
+
"policy/approxkl_avg": 0.0003968015080317855,
|
| 2020 |
+
"policy/clipfrac_avg": 0.00019654087373055518,
|
| 2021 |
+
"policy/entropy_avg": 0.2042274922132492,
|
| 2022 |
+
"step": 96,
|
| 2023 |
+
"val/clipfrac_avg": 0.0,
|
| 2024 |
+
"val/num_eos_tokens": 0,
|
| 2025 |
+
"val/ratio": 0.9995560646057129,
|
| 2026 |
+
"val/ratio_var": 9.015092246045242e-07
|
| 2027 |
+
},
|
| 2028 |
+
{
|
| 2029 |
+
"episode": 2328,
|
| 2030 |
+
"epoch": 0.3325714285714286,
|
| 2031 |
+
"eps": 3,
|
| 2032 |
+
"loss/policy_avg": -0.00013049808330833912,
|
| 2033 |
+
"loss/value_avg": 0.0011974009685218334,
|
| 2034 |
+
"lr": 8.662846177302938e-07,
|
| 2035 |
+
"objective/entropy": 11.063315391540527,
|
| 2036 |
+
"objective/kl": 0.13328056037425995,
|
| 2037 |
+
"objective/non_score_reward": -0.006664028391242027,
|
| 2038 |
+
"objective/rlhf_reward": -0.006664028391242027,
|
| 2039 |
+
"objective/scores": 0.0,
|
| 2040 |
+
"policy/approxkl_avg": 0.0004643582506105304,
|
| 2041 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 2042 |
+
"policy/entropy_avg": 0.2180095911026001,
|
| 2043 |
+
"step": 97,
|
| 2044 |
+
"val/clipfrac_avg": 0.0,
|
| 2045 |
+
"val/num_eos_tokens": 0,
|
| 2046 |
+
"val/ratio": 0.9988436698913574,
|
| 2047 |
+
"val/ratio_var": 4.182348675385583e-06
|
| 2048 |
+
},
|
| 2049 |
+
{
|
| 2050 |
+
"episode": 2352,
|
| 2051 |
+
"epoch": 0.336,
|
| 2052 |
+
"eps": 3,
|
| 2053 |
+
"loss/policy_avg": -0.00021513155661523342,
|
| 2054 |
+
"loss/value_avg": 0.0011991425417363644,
|
| 2055 |
+
"lr": 8.624237129634014e-07,
|
| 2056 |
+
"objective/entropy": 8.599056243896484,
|
| 2057 |
+
"objective/kl": 0.03197745233774185,
|
| 2058 |
+
"objective/non_score_reward": -0.0015988722443580627,
|
| 2059 |
+
"objective/rlhf_reward": -0.0015988722443580627,
|
| 2060 |
+
"objective/scores": 0.0,
|
| 2061 |
+
"policy/approxkl_avg": 0.00039817538345232606,
|
| 2062 |
+
"policy/clipfrac_avg": 0.0009827043395489454,
|
| 2063 |
+
"policy/entropy_avg": 0.18570010364055634,
|
| 2064 |
+
"step": 98,
|
| 2065 |
+
"val/clipfrac_avg": 0.0,
|
| 2066 |
+
"val/num_eos_tokens": 0,
|
| 2067 |
+
"val/ratio": 0.9998412728309631,
|
| 2068 |
+
"val/ratio_var": 1.992225406866055e-06
|
| 2069 |
+
},
|
| 2070 |
+
{
|
| 2071 |
+
"episode": 2376,
|
| 2072 |
+
"epoch": 0.3394285714285714,
|
| 2073 |
+
"eps": 3,
|
| 2074 |
+
"loss/policy_avg": -0.0004443543148227036,
|
| 2075 |
+
"loss/value_avg": 0.0011885116109624505,
|
| 2076 |
+
"lr": 8.58517888589445e-07,
|
| 2077 |
+
"objective/entropy": 11.050878524780273,
|
| 2078 |
+
"objective/kl": 0.03590617701411247,
|
| 2079 |
+
"objective/non_score_reward": -0.001795308431610465,
|
| 2080 |
+
"objective/rlhf_reward": -0.001795308431610465,
|
| 2081 |
+
"objective/scores": 0.0,
|
| 2082 |
+
"policy/approxkl_avg": 0.0004836043808609247,
|
| 2083 |
+
"policy/clipfrac_avg": 0.0017688678344711661,
|
| 2084 |
+
"policy/entropy_avg": 0.21578478813171387,
|
| 2085 |
+
"step": 99,
|
| 2086 |
+
"val/clipfrac_avg": 0.0,
|
| 2087 |
+
"val/num_eos_tokens": 0,
|
| 2088 |
+
"val/ratio": 1.000609040260315,
|
| 2089 |
+
"val/ratio_var": 2.0210025013511768e-06
|
| 2090 |
+
},
|
| 2091 |
+
{
|
| 2092 |
+
"episode": 2400,
|
| 2093 |
+
"epoch": 0.34285714285714286,
|
| 2094 |
+
"eps": 3,
|
| 2095 |
+
"loss/policy_avg": -0.0003154139267280698,
|
| 2096 |
+
"loss/value_avg": 0.0011843700194731355,
|
| 2097 |
+
"lr": 8.54567706179422e-07,
|
| 2098 |
+
"objective/entropy": 10.662446975708008,
|
| 2099 |
+
"objective/kl": 0.029185041785240173,
|
| 2100 |
+
"objective/non_score_reward": -0.0014592523220926523,
|
| 2101 |
+
"objective/rlhf_reward": -0.0014592523220926523,
|
| 2102 |
+
"objective/scores": 0.0,
|
| 2103 |
+
"policy/approxkl_avg": 0.00036341603845357895,
|
| 2104 |
+
"policy/clipfrac_avg": 0.000589622650295496,
|
| 2105 |
+
"policy/entropy_avg": 0.18732158839702606,
|
| 2106 |
+
"step": 100,
|
| 2107 |
+
"val/clipfrac_avg": 0.0,
|
| 2108 |
+
"val/num_eos_tokens": 0,
|
| 2109 |
+
"val/ratio": 0.9997389316558838,
|
| 2110 |
+
"val/ratio_var": 2.346555902477121e-06
|
| 2111 |
+
}
|
| 2112 |
+
],
|
| 2113 |
+
"logging_steps": 1,
|
| 2114 |
+
"max_steps": 292,
|
| 2115 |
+
"num_input_tokens_seen": 0,
|
| 2116 |
+
"num_train_epochs": 1.0,
|
| 2117 |
+
"save_steps": 50,
|
| 2118 |
+
"stateful_callbacks": {
|
| 2119 |
+
"TrainerControl": {
|
| 2120 |
+
"args": {
|
| 2121 |
+
"should_epoch_stop": false,
|
| 2122 |
+
"should_evaluate": false,
|
| 2123 |
+
"should_log": true,
|
| 2124 |
+
"should_save": true,
|
| 2125 |
+
"should_training_stop": false
|
| 2126 |
+
},
|
| 2127 |
+
"attributes": {}
|
| 2128 |
+
}
|
| 2129 |
+
},
|
| 2130 |
+
"total_flos": 0,
|
| 2131 |
+
"train_batch_size": null,
|
| 2132 |
+
"trial_name": null,
|
| 2133 |
+
"trial_params": null
|
| 2134 |
+
}
|