Upload trainer_state.json with huggingface_hub
Browse files- trainer_state.json +1633 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,1633 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.7142857142857143,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 100,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"completion_length": 200.0,
|
| 13 |
+
"epoch": 0.007142857142857143,
|
| 14 |
+
"grad_norm": 0.00043654805631376803,
|
| 15 |
+
"kl": 1.1190734767296817e-05,
|
| 16 |
+
"learning_rate": 3.5714285714285714e-06,
|
| 17 |
+
"loss": 0.0,
|
| 18 |
+
"reward": 0.125,
|
| 19 |
+
"reward_std": 0.0,
|
| 20 |
+
"rewards/correctness_reward_func": 0.0,
|
| 21 |
+
"rewards/length_reward_func": 0.0,
|
| 22 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 23 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 24 |
+
"rewards/xmlcount_reward_func": 0.125,
|
| 25 |
+
"step": 1
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"completion_length": 292.875,
|
| 29 |
+
"epoch": 0.014285714285714285,
|
| 30 |
+
"grad_norm": 0.1742347776889801,
|
| 31 |
+
"kl": 5.753119239670923e-06,
|
| 32 |
+
"learning_rate": 7.142857142857143e-06,
|
| 33 |
+
"loss": 0.0,
|
| 34 |
+
"reward": -0.3997499942779541,
|
| 35 |
+
"reward_std": 0.15906040370464325,
|
| 36 |
+
"rewards/correctness_reward_func": 0.0,
|
| 37 |
+
"rewards/length_reward_func": 0.0,
|
| 38 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 39 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 40 |
+
"rewards/xmlcount_reward_func": -0.3997499942779541,
|
| 41 |
+
"step": 2
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"completion_length": 445.5,
|
| 45 |
+
"epoch": 0.02142857142857143,
|
| 46 |
+
"grad_norm": 0.0913790687918663,
|
| 47 |
+
"kl": 5.179101663088659e-06,
|
| 48 |
+
"learning_rate": 1.0714285714285714e-05,
|
| 49 |
+
"loss": 0.0,
|
| 50 |
+
"reward": -0.7059999704360962,
|
| 51 |
+
"reward_std": 0.729467511177063,
|
| 52 |
+
"rewards/correctness_reward_func": 0.25,
|
| 53 |
+
"rewards/length_reward_func": 0.0,
|
| 54 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 55 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 56 |
+
"rewards/xmlcount_reward_func": -0.9559999704360962,
|
| 57 |
+
"step": 3
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"completion_length": 397.875,
|
| 61 |
+
"epoch": 0.02857142857142857,
|
| 62 |
+
"grad_norm": 0.18434225022792816,
|
| 63 |
+
"kl": 1.612883534107823e-05,
|
| 64 |
+
"learning_rate": 1.4285714285714285e-05,
|
| 65 |
+
"loss": 0.0,
|
| 66 |
+
"reward": -0.7477500438690186,
|
| 67 |
+
"reward_std": 0.5131211876869202,
|
| 68 |
+
"rewards/correctness_reward_func": 0.0,
|
| 69 |
+
"rewards/length_reward_func": 0.0,
|
| 70 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 71 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 72 |
+
"rewards/xmlcount_reward_func": -0.7477500438690186,
|
| 73 |
+
"step": 4
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"completion_length": 374.25,
|
| 77 |
+
"epoch": 0.03571428571428571,
|
| 78 |
+
"grad_norm": 0.15434053540229797,
|
| 79 |
+
"kl": 3.564520739018917e-05,
|
| 80 |
+
"learning_rate": 1.785714285714286e-05,
|
| 81 |
+
"loss": 0.0,
|
| 82 |
+
"reward": -0.4762499928474426,
|
| 83 |
+
"reward_std": 0.7022607922554016,
|
| 84 |
+
"rewards/correctness_reward_func": 0.25,
|
| 85 |
+
"rewards/length_reward_func": 0.0,
|
| 86 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 87 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 88 |
+
"rewards/xmlcount_reward_func": -0.7262499928474426,
|
| 89 |
+
"step": 5
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"completion_length": 264.25,
|
| 93 |
+
"epoch": 0.04285714285714286,
|
| 94 |
+
"grad_norm": 0.14609746634960175,
|
| 95 |
+
"kl": 0.0002129438507836312,
|
| 96 |
+
"learning_rate": 2.1428571428571428e-05,
|
| 97 |
+
"loss": 0.0,
|
| 98 |
+
"reward": -0.1522500216960907,
|
| 99 |
+
"reward_std": 0.6662349700927734,
|
| 100 |
+
"rewards/correctness_reward_func": 0.25,
|
| 101 |
+
"rewards/length_reward_func": 0.0,
|
| 102 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 103 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 104 |
+
"rewards/xmlcount_reward_func": -0.4022500216960907,
|
| 105 |
+
"step": 6
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"completion_length": 221.5,
|
| 109 |
+
"epoch": 0.05,
|
| 110 |
+
"grad_norm": 0.22101597487926483,
|
| 111 |
+
"kl": 0.0012037234846502542,
|
| 112 |
+
"learning_rate": 2.5e-05,
|
| 113 |
+
"loss": 0.0,
|
| 114 |
+
"reward": -0.11937500536441803,
|
| 115 |
+
"reward_std": 0.2714921534061432,
|
| 116 |
+
"rewards/correctness_reward_func": 0.0,
|
| 117 |
+
"rewards/length_reward_func": 0.0,
|
| 118 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 119 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 120 |
+
"rewards/xmlcount_reward_func": -0.11937499791383743,
|
| 121 |
+
"step": 7
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"completion_length": 247.875,
|
| 125 |
+
"epoch": 0.05714285714285714,
|
| 126 |
+
"grad_norm": 0.22249825298786163,
|
| 127 |
+
"kl": 0.0021207458339631557,
|
| 128 |
+
"learning_rate": 2.857142857142857e-05,
|
| 129 |
+
"loss": 0.0001,
|
| 130 |
+
"reward": -0.19712500274181366,
|
| 131 |
+
"reward_std": 0.36591196060180664,
|
| 132 |
+
"rewards/correctness_reward_func": 0.0,
|
| 133 |
+
"rewards/length_reward_func": 0.0,
|
| 134 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 135 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 136 |
+
"rewards/xmlcount_reward_func": -0.19712500274181366,
|
| 137 |
+
"step": 8
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"completion_length": 511.75,
|
| 141 |
+
"epoch": 0.06428571428571428,
|
| 142 |
+
"grad_norm": 0.11614850163459778,
|
| 143 |
+
"kl": 0.0037874511908739805,
|
| 144 |
+
"learning_rate": 3.2142857142857144e-05,
|
| 145 |
+
"loss": 0.0002,
|
| 146 |
+
"reward": -0.7689999938011169,
|
| 147 |
+
"reward_std": 1.3408536911010742,
|
| 148 |
+
"rewards/correctness_reward_func": 0.25,
|
| 149 |
+
"rewards/length_reward_func": 0.0,
|
| 150 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 151 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 152 |
+
"rewards/xmlcount_reward_func": -1.0189999341964722,
|
| 153 |
+
"step": 9
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"completion_length": 221.625,
|
| 157 |
+
"epoch": 0.07142857142857142,
|
| 158 |
+
"grad_norm": 0.21436981856822968,
|
| 159 |
+
"kl": 0.012746547348797321,
|
| 160 |
+
"learning_rate": 3.571428571428572e-05,
|
| 161 |
+
"loss": 0.0005,
|
| 162 |
+
"reward": -0.23862498998641968,
|
| 163 |
+
"reward_std": 0.11265870183706284,
|
| 164 |
+
"rewards/correctness_reward_func": 0.0,
|
| 165 |
+
"rewards/length_reward_func": 0.0,
|
| 166 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 167 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 168 |
+
"rewards/xmlcount_reward_func": -0.23862498998641968,
|
| 169 |
+
"step": 10
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"completion_length": 292.75,
|
| 173 |
+
"epoch": 0.07857142857142857,
|
| 174 |
+
"grad_norm": 0.19347621500492096,
|
| 175 |
+
"kl": 0.011033562943339348,
|
| 176 |
+
"learning_rate": 3.928571428571429e-05,
|
| 177 |
+
"loss": 0.0004,
|
| 178 |
+
"reward": -0.4468750059604645,
|
| 179 |
+
"reward_std": 0.16547630727291107,
|
| 180 |
+
"rewards/correctness_reward_func": 0.0,
|
| 181 |
+
"rewards/length_reward_func": 0.0,
|
| 182 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 183 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 184 |
+
"rewards/xmlcount_reward_func": -0.4468749761581421,
|
| 185 |
+
"step": 11
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"completion_length": 110.625,
|
| 189 |
+
"epoch": 0.08571428571428572,
|
| 190 |
+
"grad_norm": 0.44592323899269104,
|
| 191 |
+
"kl": 0.060783885419368744,
|
| 192 |
+
"learning_rate": 4.2857142857142856e-05,
|
| 193 |
+
"loss": 0.0024,
|
| 194 |
+
"reward": 1.424625039100647,
|
| 195 |
+
"reward_std": 1.0278345346450806,
|
| 196 |
+
"rewards/correctness_reward_func": 1.25,
|
| 197 |
+
"rewards/length_reward_func": 0.0,
|
| 198 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 199 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 200 |
+
"rewards/xmlcount_reward_func": 0.17462500929832458,
|
| 201 |
+
"step": 12
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"completion_length": 678.75,
|
| 205 |
+
"epoch": 0.09285714285714286,
|
| 206 |
+
"grad_norm": 0.1104876920580864,
|
| 207 |
+
"kl": 0.013322519138455391,
|
| 208 |
+
"learning_rate": 4.642857142857143e-05,
|
| 209 |
+
"loss": 0.0005,
|
| 210 |
+
"reward": 0.468500018119812,
|
| 211 |
+
"reward_std": 0.9677569270133972,
|
| 212 |
+
"rewards/correctness_reward_func": 1.0,
|
| 213 |
+
"rewards/length_reward_func": 0.125,
|
| 214 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 215 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 216 |
+
"rewards/xmlcount_reward_func": -0.656499981880188,
|
| 217 |
+
"step": 13
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"completion_length": 495.25,
|
| 221 |
+
"epoch": 0.1,
|
| 222 |
+
"grad_norm": 0.13062289357185364,
|
| 223 |
+
"kl": 0.021864598616957664,
|
| 224 |
+
"learning_rate": 5e-05,
|
| 225 |
+
"loss": 0.0009,
|
| 226 |
+
"reward": -0.5831249952316284,
|
| 227 |
+
"reward_std": 0.676399827003479,
|
| 228 |
+
"rewards/correctness_reward_func": 0.0,
|
| 229 |
+
"rewards/length_reward_func": 0.0,
|
| 230 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 231 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 232 |
+
"rewards/xmlcount_reward_func": -0.5831249952316284,
|
| 233 |
+
"step": 14
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"completion_length": 781.375,
|
| 237 |
+
"epoch": 0.10714285714285714,
|
| 238 |
+
"grad_norm": 0.09523279964923859,
|
| 239 |
+
"kl": 0.013723693788051605,
|
| 240 |
+
"learning_rate": 4.999222955002041e-05,
|
| 241 |
+
"loss": 0.0005,
|
| 242 |
+
"reward": -0.13512498140335083,
|
| 243 |
+
"reward_std": 0.9264968633651733,
|
| 244 |
+
"rewards/correctness_reward_func": 0.0,
|
| 245 |
+
"rewards/length_reward_func": 0.0,
|
| 246 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 247 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 248 |
+
"rewards/xmlcount_reward_func": -0.13512498140335083,
|
| 249 |
+
"step": 15
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"completion_length": 405.5,
|
| 253 |
+
"epoch": 0.11428571428571428,
|
| 254 |
+
"grad_norm": 0.15296146273612976,
|
| 255 |
+
"kl": 0.03523610904812813,
|
| 256 |
+
"learning_rate": 4.996892303047306e-05,
|
| 257 |
+
"loss": 0.0014,
|
| 258 |
+
"reward": 2.193000078201294,
|
| 259 |
+
"reward_std": 0.43767958879470825,
|
| 260 |
+
"rewards/correctness_reward_func": 2.0,
|
| 261 |
+
"rewards/length_reward_func": 0.0,
|
| 262 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 263 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 264 |
+
"rewards/xmlcount_reward_func": 0.19300000369548798,
|
| 265 |
+
"step": 16
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"completion_length": 214.75,
|
| 269 |
+
"epoch": 0.12142857142857143,
|
| 270 |
+
"grad_norm": 0.2985721230506897,
|
| 271 |
+
"kl": 0.0879557877779007,
|
| 272 |
+
"learning_rate": 4.9930094929529506e-05,
|
| 273 |
+
"loss": 0.0035,
|
| 274 |
+
"reward": -0.011874988675117493,
|
| 275 |
+
"reward_std": 0.6198051571846008,
|
| 276 |
+
"rewards/correctness_reward_func": 0.0,
|
| 277 |
+
"rewards/length_reward_func": 0.0,
|
| 278 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 279 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 280 |
+
"rewards/xmlcount_reward_func": -0.011875003576278687,
|
| 281 |
+
"step": 17
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"completion_length": 121.25,
|
| 285 |
+
"epoch": 0.12857142857142856,
|
| 286 |
+
"grad_norm": 0.417566180229187,
|
| 287 |
+
"kl": 0.22521352767944336,
|
| 288 |
+
"learning_rate": 4.987576938413504e-05,
|
| 289 |
+
"loss": 0.009,
|
| 290 |
+
"reward": 2.4171249866485596,
|
| 291 |
+
"reward_std": 0.1779337227344513,
|
| 292 |
+
"rewards/correctness_reward_func": 2.0,
|
| 293 |
+
"rewards/length_reward_func": 0.0,
|
| 294 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 295 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 296 |
+
"rewards/xmlcount_reward_func": 0.41712498664855957,
|
| 297 |
+
"step": 18
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"completion_length": 179.0,
|
| 301 |
+
"epoch": 0.1357142857142857,
|
| 302 |
+
"grad_norm": 0.285080224275589,
|
| 303 |
+
"kl": 0.10962098836898804,
|
| 304 |
+
"learning_rate": 4.9805980165004304e-05,
|
| 305 |
+
"loss": 0.0044,
|
| 306 |
+
"reward": 1.75,
|
| 307 |
+
"reward_std": 1.0350983142852783,
|
| 308 |
+
"rewards/correctness_reward_func": 1.25,
|
| 309 |
+
"rewards/length_reward_func": 0.0,
|
| 310 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 311 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 312 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 313 |
+
"step": 19
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"completion_length": 282.875,
|
| 317 |
+
"epoch": 0.14285714285714285,
|
| 318 |
+
"grad_norm": 0.27518561482429504,
|
| 319 |
+
"kl": 0.06473983079195023,
|
| 320 |
+
"learning_rate": 4.972077065562821e-05,
|
| 321 |
+
"loss": 0.0026,
|
| 322 |
+
"reward": 1.2120000123977661,
|
| 323 |
+
"reward_std": 1.0715053081512451,
|
| 324 |
+
"rewards/correctness_reward_func": 0.75,
|
| 325 |
+
"rewards/length_reward_func": 0.0,
|
| 326 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 327 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 328 |
+
"rewards/xmlcount_reward_func": 0.4620000123977661,
|
| 329 |
+
"step": 20
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"completion_length": 146.0,
|
| 333 |
+
"epoch": 0.15,
|
| 334 |
+
"grad_norm": 0.01954607106745243,
|
| 335 |
+
"kl": 0.1001419946551323,
|
| 336 |
+
"learning_rate": 4.962019382530521e-05,
|
| 337 |
+
"loss": 0.004,
|
| 338 |
+
"reward": 2.5,
|
| 339 |
+
"reward_std": 0.0,
|
| 340 |
+
"rewards/correctness_reward_func": 2.0,
|
| 341 |
+
"rewards/length_reward_func": 0.0,
|
| 342 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 343 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 344 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 345 |
+
"step": 21
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"completion_length": 233.125,
|
| 349 |
+
"epoch": 0.15714285714285714,
|
| 350 |
+
"grad_norm": 0.019903944805264473,
|
| 351 |
+
"kl": 0.0685255229473114,
|
| 352 |
+
"learning_rate": 4.9504312196213596e-05,
|
| 353 |
+
"loss": 0.0027,
|
| 354 |
+
"reward": 2.5,
|
| 355 |
+
"reward_std": 0.0,
|
| 356 |
+
"rewards/correctness_reward_func": 2.0,
|
| 357 |
+
"rewards/length_reward_func": 0.0,
|
| 358 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 359 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 360 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 361 |
+
"step": 22
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"completion_length": 199.125,
|
| 365 |
+
"epoch": 0.16428571428571428,
|
| 366 |
+
"grad_norm": 0.022628186270594597,
|
| 367 |
+
"kl": 0.08269491791725159,
|
| 368 |
+
"learning_rate": 4.937319780454559e-05,
|
| 369 |
+
"loss": 0.0033,
|
| 370 |
+
"reward": 2.5,
|
| 371 |
+
"reward_std": 0.0,
|
| 372 |
+
"rewards/correctness_reward_func": 2.0,
|
| 373 |
+
"rewards/length_reward_func": 0.0,
|
| 374 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 375 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 376 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 377 |
+
"step": 23
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"completion_length": 388.25,
|
| 381 |
+
"epoch": 0.17142857142857143,
|
| 382 |
+
"grad_norm": 0.12237099558115005,
|
| 383 |
+
"kl": 0.031046129763126373,
|
| 384 |
+
"learning_rate": 4.922693215572695e-05,
|
| 385 |
+
"loss": 0.0012,
|
| 386 |
+
"reward": 1.5,
|
| 387 |
+
"reward_std": 1.0690449476242065,
|
| 388 |
+
"rewards/correctness_reward_func": 1.0,
|
| 389 |
+
"rewards/length_reward_func": 0.0,
|
| 390 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 391 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 392 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 393 |
+
"step": 24
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"completion_length": 262.875,
|
| 397 |
+
"epoch": 0.17857142857142858,
|
| 398 |
+
"grad_norm": 0.026575006544589996,
|
| 399 |
+
"kl": 0.08844916522502899,
|
| 400 |
+
"learning_rate": 4.90656061737503e-05,
|
| 401 |
+
"loss": 0.0035,
|
| 402 |
+
"reward": 2.5,
|
| 403 |
+
"reward_std": 0.0,
|
| 404 |
+
"rewards/correctness_reward_func": 2.0,
|
| 405 |
+
"rewards/length_reward_func": 0.0,
|
| 406 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 407 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 408 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 409 |
+
"step": 25
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"completion_length": 290.125,
|
| 413 |
+
"epoch": 0.18571428571428572,
|
| 414 |
+
"grad_norm": 0.5788246989250183,
|
| 415 |
+
"kl": 0.06681957095861435,
|
| 416 |
+
"learning_rate": 4.888932014465352e-05,
|
| 417 |
+
"loss": 0.0027,
|
| 418 |
+
"reward": 0.375,
|
| 419 |
+
"reward_std": 0.2314550280570984,
|
| 420 |
+
"rewards/correctness_reward_func": 0.0,
|
| 421 |
+
"rewards/length_reward_func": 0.0,
|
| 422 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 423 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 424 |
+
"rewards/xmlcount_reward_func": 0.375,
|
| 425 |
+
"step": 26
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"completion_length": 224.375,
|
| 429 |
+
"epoch": 0.19285714285714287,
|
| 430 |
+
"grad_norm": 0.4455007314682007,
|
| 431 |
+
"kl": 0.2504517138004303,
|
| 432 |
+
"learning_rate": 4.86981836541783e-05,
|
| 433 |
+
"loss": 0.01,
|
| 434 |
+
"reward": 1.625,
|
| 435 |
+
"reward_std": 1.2174328565597534,
|
| 436 |
+
"rewards/correctness_reward_func": 1.25,
|
| 437 |
+
"rewards/length_reward_func": 0.0,
|
| 438 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 439 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 440 |
+
"rewards/xmlcount_reward_func": 0.375,
|
| 441 |
+
"step": 27
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"completion_length": 232.125,
|
| 445 |
+
"epoch": 0.2,
|
| 446 |
+
"grad_norm": 0.23774927854537964,
|
| 447 |
+
"kl": 0.1069759875535965,
|
| 448 |
+
"learning_rate": 4.849231551964771e-05,
|
| 449 |
+
"loss": 0.0043,
|
| 450 |
+
"reward": 1.921875,
|
| 451 |
+
"reward_std": 0.8937718272209167,
|
| 452 |
+
"rewards/correctness_reward_func": 1.5,
|
| 453 |
+
"rewards/length_reward_func": 0.0,
|
| 454 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 455 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 456 |
+
"rewards/xmlcount_reward_func": 0.421875,
|
| 457 |
+
"step": 28
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"completion_length": 458.625,
|
| 461 |
+
"epoch": 0.20714285714285716,
|
| 462 |
+
"grad_norm": 0.09729224443435669,
|
| 463 |
+
"kl": 0.03021763078868389,
|
| 464 |
+
"learning_rate": 4.827184371610511e-05,
|
| 465 |
+
"loss": 0.0012,
|
| 466 |
+
"reward": 1.75,
|
| 467 |
+
"reward_std": 1.0350983142852783,
|
| 468 |
+
"rewards/correctness_reward_func": 1.25,
|
| 469 |
+
"rewards/length_reward_func": 0.0,
|
| 470 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 471 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 472 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 473 |
+
"step": 29
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"completion_length": 422.375,
|
| 477 |
+
"epoch": 0.21428571428571427,
|
| 478 |
+
"grad_norm": 0.12195795774459839,
|
| 479 |
+
"kl": 0.04913492873311043,
|
| 480 |
+
"learning_rate": 4.803690529676019e-05,
|
| 481 |
+
"loss": 0.002,
|
| 482 |
+
"reward": 2.25,
|
| 483 |
+
"reward_std": 0.5345224738121033,
|
| 484 |
+
"rewards/correctness_reward_func": 1.75,
|
| 485 |
+
"rewards/length_reward_func": 0.0625,
|
| 486 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 487 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 488 |
+
"rewards/xmlcount_reward_func": 0.4375,
|
| 489 |
+
"step": 30
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"completion_length": 270.875,
|
| 493 |
+
"epoch": 0.22142857142857142,
|
| 494 |
+
"grad_norm": 0.21031659841537476,
|
| 495 |
+
"kl": 0.0698024183511734,
|
| 496 |
+
"learning_rate": 4.778764630779183e-05,
|
| 497 |
+
"loss": 0.0028,
|
| 498 |
+
"reward": 2.21875,
|
| 499 |
+
"reward_std": 0.6999680995941162,
|
| 500 |
+
"rewards/correctness_reward_func": 1.75,
|
| 501 |
+
"rewards/length_reward_func": 0.0,
|
| 502 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 503 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 504 |
+
"rewards/xmlcount_reward_func": 0.46875,
|
| 505 |
+
"step": 31
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"completion_length": 377.5,
|
| 509 |
+
"epoch": 0.22857142857142856,
|
| 510 |
+
"grad_norm": 0.010702410712838173,
|
| 511 |
+
"kl": 0.037429843097925186,
|
| 512 |
+
"learning_rate": 4.752422169756048e-05,
|
| 513 |
+
"loss": 0.0015,
|
| 514 |
+
"reward": 2.5,
|
| 515 |
+
"reward_std": 0.0,
|
| 516 |
+
"rewards/correctness_reward_func": 2.0,
|
| 517 |
+
"rewards/length_reward_func": 0.0,
|
| 518 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 519 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 520 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 521 |
+
"step": 32
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"completion_length": 360.125,
|
| 525 |
+
"epoch": 0.2357142857142857,
|
| 526 |
+
"grad_norm": 0.14157415926456451,
|
| 527 |
+
"kl": 0.08244010806083679,
|
| 528 |
+
"learning_rate": 4.724679522028672e-05,
|
| 529 |
+
"loss": 0.0033,
|
| 530 |
+
"reward": 1.6486248970031738,
|
| 531 |
+
"reward_std": 1.1074291467666626,
|
| 532 |
+
"rewards/correctness_reward_func": 1.25,
|
| 533 |
+
"rewards/length_reward_func": 0.0,
|
| 534 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 535 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 536 |
+
"rewards/xmlcount_reward_func": 0.3986250162124634,
|
| 537 |
+
"step": 33
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"completion_length": 116.125,
|
| 541 |
+
"epoch": 0.24285714285714285,
|
| 542 |
+
"grad_norm": 1.3088772296905518,
|
| 543 |
+
"kl": 0.7027589082717896,
|
| 544 |
+
"learning_rate": 4.6955539334255716e-05,
|
| 545 |
+
"loss": 0.0281,
|
| 546 |
+
"reward": 1.75,
|
| 547 |
+
"reward_std": 1.0350983142852783,
|
| 548 |
+
"rewards/correctness_reward_func": 1.25,
|
| 549 |
+
"rewards/length_reward_func": 0.0,
|
| 550 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 551 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 552 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 553 |
+
"step": 34
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"completion_length": 311.0,
|
| 557 |
+
"epoch": 0.25,
|
| 558 |
+
"grad_norm": 0.7312092781066895,
|
| 559 |
+
"kl": 0.07371841371059418,
|
| 560 |
+
"learning_rate": 4.665063509461097e-05,
|
| 561 |
+
"loss": 0.0029,
|
| 562 |
+
"reward": 0.6698750257492065,
|
| 563 |
+
"reward_std": 0.7472349405288696,
|
| 564 |
+
"rewards/correctness_reward_func": 0.25,
|
| 565 |
+
"rewards/length_reward_func": 0.0,
|
| 566 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 567 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 568 |
+
"rewards/xmlcount_reward_func": 0.41987499594688416,
|
| 569 |
+
"step": 35
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"completion_length": 329.625,
|
| 573 |
+
"epoch": 0.2571428571428571,
|
| 574 |
+
"grad_norm": 0.006386851891875267,
|
| 575 |
+
"kl": 0.039320699870586395,
|
| 576 |
+
"learning_rate": 4.6332272040803895e-05,
|
| 577 |
+
"loss": 0.0016,
|
| 578 |
+
"reward": 2.5,
|
| 579 |
+
"reward_std": 0.0,
|
| 580 |
+
"rewards/correctness_reward_func": 2.0,
|
| 581 |
+
"rewards/length_reward_func": 0.0,
|
| 582 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 583 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 584 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 585 |
+
"step": 36
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"completion_length": 365.875,
|
| 589 |
+
"epoch": 0.2642857142857143,
|
| 590 |
+
"grad_norm": 0.20290029048919678,
|
| 591 |
+
"kl": 0.05483713746070862,
|
| 592 |
+
"learning_rate": 4.600064807876929e-05,
|
| 593 |
+
"loss": 0.0022,
|
| 594 |
+
"reward": 1.6407499313354492,
|
| 595 |
+
"reward_std": 1.0948845148086548,
|
| 596 |
+
"rewards/correctness_reward_func": 1.25,
|
| 597 |
+
"rewards/length_reward_func": 0.0,
|
| 598 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 599 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 600 |
+
"rewards/xmlcount_reward_func": 0.390749990940094,
|
| 601 |
+
"step": 37
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"completion_length": 287.0,
|
| 605 |
+
"epoch": 0.2714285714285714,
|
| 606 |
+
"grad_norm": 0.19799265265464783,
|
| 607 |
+
"kl": 0.13818374276161194,
|
| 608 |
+
"learning_rate": 4.5655969357899874e-05,
|
| 609 |
+
"loss": 0.0055,
|
| 610 |
+
"reward": 1.875999927520752,
|
| 611 |
+
"reward_std": 0.9384979605674744,
|
| 612 |
+
"rewards/correctness_reward_func": 1.5,
|
| 613 |
+
"rewards/length_reward_func": 0.0,
|
| 614 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 615 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 616 |
+
"rewards/xmlcount_reward_func": 0.37599998712539673,
|
| 617 |
+
"step": 38
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"completion_length": 289.875,
|
| 621 |
+
"epoch": 0.2785714285714286,
|
| 622 |
+
"grad_norm": 0.37082406878471375,
|
| 623 |
+
"kl": 0.14701411128044128,
|
| 624 |
+
"learning_rate": 4.529845014289642e-05,
|
| 625 |
+
"loss": 0.0059,
|
| 626 |
+
"reward": 2.0653750896453857,
|
| 627 |
+
"reward_std": 0.6956271529197693,
|
| 628 |
+
"rewards/correctness_reward_func": 1.75,
|
| 629 |
+
"rewards/length_reward_func": 0.0,
|
| 630 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 631 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 632 |
+
"rewards/xmlcount_reward_func": 0.3153750002384186,
|
| 633 |
+
"step": 39
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"completion_length": 725.125,
|
| 637 |
+
"epoch": 0.2857142857142857,
|
| 638 |
+
"grad_norm": 0.11310902237892151,
|
| 639 |
+
"kl": 0.05320233479142189,
|
| 640 |
+
"learning_rate": 4.4928312680573064e-05,
|
| 641 |
+
"loss": 0.0021,
|
| 642 |
+
"reward": 0.5361250042915344,
|
| 643 |
+
"reward_std": 0.2760908007621765,
|
| 644 |
+
"rewards/correctness_reward_func": 0.0,
|
| 645 |
+
"rewards/length_reward_func": 0.1875,
|
| 646 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 647 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 648 |
+
"rewards/xmlcount_reward_func": 0.3486250042915344,
|
| 649 |
+
"step": 40
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"completion_length": 373.625,
|
| 653 |
+
"epoch": 0.29285714285714287,
|
| 654 |
+
"grad_norm": 0.005955233704298735,
|
| 655 |
+
"kl": 0.09833689779043198,
|
| 656 |
+
"learning_rate": 4.454578706170075e-05,
|
| 657 |
+
"loss": 0.0039,
|
| 658 |
+
"reward": 0.36500000953674316,
|
| 659 |
+
"reward_std": 0.0,
|
| 660 |
+
"rewards/correctness_reward_func": 0.0,
|
| 661 |
+
"rewards/length_reward_func": 0.0,
|
| 662 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 663 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 664 |
+
"rewards/xmlcount_reward_func": 0.36500000953674316,
|
| 665 |
+
"step": 41
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"completion_length": 337.75,
|
| 669 |
+
"epoch": 0.3,
|
| 670 |
+
"grad_norm": 0.18655544519424438,
|
| 671 |
+
"kl": 0.13665920495986938,
|
| 672 |
+
"learning_rate": 4.415111107797445e-05,
|
| 673 |
+
"loss": 0.0055,
|
| 674 |
+
"reward": 1.3464999198913574,
|
| 675 |
+
"reward_std": 1.0498690605163574,
|
| 676 |
+
"rewards/correctness_reward_func": 1.0,
|
| 677 |
+
"rewards/length_reward_func": 0.0,
|
| 678 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 679 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 680 |
+
"rewards/xmlcount_reward_func": 0.3465000092983246,
|
| 681 |
+
"step": 42
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"completion_length": 463.75,
|
| 685 |
+
"epoch": 0.30714285714285716,
|
| 686 |
+
"grad_norm": 0.18999366462230682,
|
| 687 |
+
"kl": 0.1055900901556015,
|
| 688 |
+
"learning_rate": 4.374453007419336e-05,
|
| 689 |
+
"loss": 0.0042,
|
| 690 |
+
"reward": 0.3206250071525574,
|
| 691 |
+
"reward_std": 0.08869842439889908,
|
| 692 |
+
"rewards/correctness_reward_func": 0.0,
|
| 693 |
+
"rewards/length_reward_func": 0.0,
|
| 694 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 695 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 696 |
+
"rewards/xmlcount_reward_func": 0.3206250071525574,
|
| 697 |
+
"step": 43
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"completion_length": 350.25,
|
| 701 |
+
"epoch": 0.3142857142857143,
|
| 702 |
+
"grad_norm": 0.16436553001403809,
|
| 703 |
+
"kl": 0.12820430099964142,
|
| 704 |
+
"learning_rate": 4.332629679574566e-05,
|
| 705 |
+
"loss": 0.0051,
|
| 706 |
+
"reward": 2.3486249446868896,
|
| 707 |
+
"reward_std": 0.04361177235841751,
|
| 708 |
+
"rewards/correctness_reward_func": 2.0,
|
| 709 |
+
"rewards/length_reward_func": 0.0,
|
| 710 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 711 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 712 |
+
"rewards/xmlcount_reward_func": 0.3486250042915344,
|
| 713 |
+
"step": 44
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"completion_length": 396.625,
|
| 717 |
+
"epoch": 0.32142857142857145,
|
| 718 |
+
"grad_norm": 0.12755045294761658,
|
| 719 |
+
"kl": 0.10351184010505676,
|
| 720 |
+
"learning_rate": 4.2896671231492966e-05,
|
| 721 |
+
"loss": 0.0041,
|
| 722 |
+
"reward": 2.1152501106262207,
|
| 723 |
+
"reward_std": 0.707207977771759,
|
| 724 |
+
"rewards/correctness_reward_func": 1.75,
|
| 725 |
+
"rewards/length_reward_func": 0.0,
|
| 726 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 727 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 728 |
+
"rewards/xmlcount_reward_func": 0.36524999141693115,
|
| 729 |
+
"step": 45
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"completion_length": 323.0,
|
| 733 |
+
"epoch": 0.32857142857142857,
|
| 734 |
+
"grad_norm": 0.1433708518743515,
|
| 735 |
+
"kl": 0.12643718719482422,
|
| 736 |
+
"learning_rate": 4.245592045215182e-05,
|
| 737 |
+
"loss": 0.0051,
|
| 738 |
+
"reward": 2.085624933242798,
|
| 739 |
+
"reward_std": 0.6997721791267395,
|
| 740 |
+
"rewards/correctness_reward_func": 1.75,
|
| 741 |
+
"rewards/length_reward_func": 0.0,
|
| 742 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 743 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 744 |
+
"rewards/xmlcount_reward_func": 0.3356249928474426,
|
| 745 |
+
"step": 46
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"completion_length": 465.875,
|
| 749 |
+
"epoch": 0.3357142857142857,
|
| 750 |
+
"grad_norm": 0.14527598023414612,
|
| 751 |
+
"kl": 0.09910832345485687,
|
| 752 |
+
"learning_rate": 4.2004318444272985e-05,
|
| 753 |
+
"loss": 0.004,
|
| 754 |
+
"reward": 1.0987499952316284,
|
| 755 |
+
"reward_std": 1.0445955991744995,
|
| 756 |
+
"rewards/correctness_reward_func": 0.75,
|
| 757 |
+
"rewards/length_reward_func": 0.0,
|
| 758 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 759 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 760 |
+
"rewards/xmlcount_reward_func": 0.3487499952316284,
|
| 761 |
+
"step": 47
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"completion_length": 443.875,
|
| 765 |
+
"epoch": 0.34285714285714286,
|
| 766 |
+
"grad_norm": 0.11678742617368698,
|
| 767 |
+
"kl": 0.09625618904829025,
|
| 768 |
+
"learning_rate": 4.154214593992149e-05,
|
| 769 |
+
"loss": 0.0039,
|
| 770 |
+
"reward": 2.1332499980926514,
|
| 771 |
+
"reward_std": 0.716006875038147,
|
| 772 |
+
"rewards/correctness_reward_func": 1.75,
|
| 773 |
+
"rewards/length_reward_func": 0.0,
|
| 774 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 775 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 776 |
+
"rewards/xmlcount_reward_func": 0.38324999809265137,
|
| 777 |
+
"step": 48
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"completion_length": 298.875,
|
| 781 |
+
"epoch": 0.35,
|
| 782 |
+
"grad_norm": 0.2462267130613327,
|
| 783 |
+
"kl": 0.14918765425682068,
|
| 784 |
+
"learning_rate": 4.1069690242163484e-05,
|
| 785 |
+
"loss": 0.006,
|
| 786 |
+
"reward": 1.3640000820159912,
|
| 787 |
+
"reward_std": 1.0677127838134766,
|
| 788 |
+
"rewards/correctness_reward_func": 1.0,
|
| 789 |
+
"rewards/length_reward_func": 0.0,
|
| 790 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 791 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 792 |
+
"rewards/xmlcount_reward_func": 0.36400002241134644,
|
| 793 |
+
"step": 49
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"completion_length": 604.125,
|
| 797 |
+
"epoch": 0.35714285714285715,
|
| 798 |
+
"grad_norm": 0.11957182735204697,
|
| 799 |
+
"kl": 0.06581288576126099,
|
| 800 |
+
"learning_rate": 4.058724504646834e-05,
|
| 801 |
+
"loss": 0.0026,
|
| 802 |
+
"reward": 0.42887499928474426,
|
| 803 |
+
"reward_std": 0.1762550324201584,
|
| 804 |
+
"rewards/correctness_reward_func": 0.0,
|
| 805 |
+
"rewards/length_reward_func": 0.0625,
|
| 806 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 807 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 808 |
+
"rewards/xmlcount_reward_func": 0.36637499928474426,
|
| 809 |
+
"step": 50
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"completion_length": 630.75,
|
| 813 |
+
"epoch": 0.36428571428571427,
|
| 814 |
+
"grad_norm": 0.20287089049816132,
|
| 815 |
+
"kl": 0.15163259208202362,
|
| 816 |
+
"learning_rate": 4.009511025813694e-05,
|
| 817 |
+
"loss": 0.0061,
|
| 818 |
+
"reward": 0.3474999964237213,
|
| 819 |
+
"reward_std": 0.10380475223064423,
|
| 820 |
+
"rewards/correctness_reward_func": 0.0,
|
| 821 |
+
"rewards/length_reward_func": 0.0,
|
| 822 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 823 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 824 |
+
"rewards/xmlcount_reward_func": 0.3475000262260437,
|
| 825 |
+
"step": 51
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"completion_length": 409.875,
|
| 829 |
+
"epoch": 0.37142857142857144,
|
| 830 |
+
"grad_norm": 0.15633811056613922,
|
| 831 |
+
"kl": 0.12077239155769348,
|
| 832 |
+
"learning_rate": 3.959359180586975e-05,
|
| 833 |
+
"loss": 0.0048,
|
| 834 |
+
"reward": 1.6152499914169312,
|
| 835 |
+
"reward_std": 1.0375232696533203,
|
| 836 |
+
"rewards/correctness_reward_func": 1.25,
|
| 837 |
+
"rewards/length_reward_func": 0.0,
|
| 838 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 839 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 840 |
+
"rewards/xmlcount_reward_func": 0.36524999141693115,
|
| 841 |
+
"step": 52
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"completion_length": 473.875,
|
| 845 |
+
"epoch": 0.37857142857142856,
|
| 846 |
+
"grad_norm": 0.1574457734823227,
|
| 847 |
+
"kl": 0.08710946887731552,
|
| 848 |
+
"learning_rate": 3.908300145159055e-05,
|
| 849 |
+
"loss": 0.0035,
|
| 850 |
+
"reward": 0.8807500004768372,
|
| 851 |
+
"reward_std": 0.9173192381858826,
|
| 852 |
+
"rewards/correctness_reward_func": 0.5,
|
| 853 |
+
"rewards/length_reward_func": 0.0,
|
| 854 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 855 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 856 |
+
"rewards/xmlcount_reward_func": 0.38075000047683716,
|
| 857 |
+
"step": 53
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"completion_length": 321.25,
|
| 861 |
+
"epoch": 0.38571428571428573,
|
| 862 |
+
"grad_norm": 0.19621425867080688,
|
| 863 |
+
"kl": 0.17770174145698547,
|
| 864 |
+
"learning_rate": 3.856365659664399e-05,
|
| 865 |
+
"loss": 0.0071,
|
| 866 |
+
"reward": 2.1021249294281006,
|
| 867 |
+
"reward_std": 0.7031054496765137,
|
| 868 |
+
"rewards/correctness_reward_func": 1.75,
|
| 869 |
+
"rewards/length_reward_func": 0.0,
|
| 870 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 871 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 872 |
+
"rewards/xmlcount_reward_func": 0.35212498903274536,
|
| 873 |
+
"step": 54
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"completion_length": 273.75,
|
| 877 |
+
"epoch": 0.39285714285714285,
|
| 878 |
+
"grad_norm": 0.19145628809928894,
|
| 879 |
+
"kl": 0.16191110014915466,
|
| 880 |
+
"learning_rate": 3.803588008448745e-05,
|
| 881 |
+
"loss": 0.0065,
|
| 882 |
+
"reward": 0.6318750381469727,
|
| 883 |
+
"reward_std": 0.7018797993659973,
|
| 884 |
+
"rewards/correctness_reward_func": 0.25,
|
| 885 |
+
"rewards/length_reward_func": 0.0,
|
| 886 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 887 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 888 |
+
"rewards/xmlcount_reward_func": 0.38187500834465027,
|
| 889 |
+
"step": 55
|
| 890 |
+
},
|
| 891 |
+
{
|
| 892 |
+
"completion_length": 281.25,
|
| 893 |
+
"epoch": 0.4,
|
| 894 |
+
"grad_norm": 0.30190473794937134,
|
| 895 |
+
"kl": 0.1685158759355545,
|
| 896 |
+
"learning_rate": 3.7500000000000003e-05,
|
| 897 |
+
"loss": 0.0067,
|
| 898 |
+
"reward": 0.8665000200271606,
|
| 899 |
+
"reward_std": 0.9249003529548645,
|
| 900 |
+
"rewards/correctness_reward_func": 0.5,
|
| 901 |
+
"rewards/length_reward_func": 0.0,
|
| 902 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 903 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 904 |
+
"rewards/xmlcount_reward_func": 0.36650002002716064,
|
| 905 |
+
"step": 56
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"completion_length": 340.75,
|
| 909 |
+
"epoch": 0.40714285714285714,
|
| 910 |
+
"grad_norm": 0.15627902746200562,
|
| 911 |
+
"kl": 0.12346489727497101,
|
| 912 |
+
"learning_rate": 3.695634946553296e-05,
|
| 913 |
+
"loss": 0.0049,
|
| 914 |
+
"reward": 0.32987499237060547,
|
| 915 |
+
"reward_std": 0.06595548242330551,
|
| 916 |
+
"rewards/correctness_reward_func": 0.0,
|
| 917 |
+
"rewards/length_reward_func": 0.0,
|
| 918 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 919 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 920 |
+
"rewards/xmlcount_reward_func": 0.32987502217292786,
|
| 921 |
+
"step": 57
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"completion_length": 993.25,
|
| 925 |
+
"epoch": 0.4142857142857143,
|
| 926 |
+
"grad_norm": 0.0931173712015152,
|
| 927 |
+
"kl": 0.018140610307455063,
|
| 928 |
+
"learning_rate": 3.6405266433829075e-05,
|
| 929 |
+
"loss": 0.0007,
|
| 930 |
+
"reward": 0.4088750183582306,
|
| 931 |
+
"reward_std": 0.2790803611278534,
|
| 932 |
+
"rewards/correctness_reward_func": 0.0,
|
| 933 |
+
"rewards/length_reward_func": 0.25,
|
| 934 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 935 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 936 |
+
"rewards/xmlcount_reward_func": 0.1588750034570694,
|
| 937 |
+
"step": 58
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"completion_length": 474.875,
|
| 941 |
+
"epoch": 0.42142857142857143,
|
| 942 |
+
"grad_norm": 0.10973533242940903,
|
| 943 |
+
"kl": 0.09137643128633499,
|
| 944 |
+
"learning_rate": 3.5847093477938956e-05,
|
| 945 |
+
"loss": 0.0037,
|
| 946 |
+
"reward": 0.42787498235702515,
|
| 947 |
+
"reward_std": 0.17703060805797577,
|
| 948 |
+
"rewards/correctness_reward_func": 0.0,
|
| 949 |
+
"rewards/length_reward_func": 0.0625,
|
| 950 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 951 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 952 |
+
"rewards/xmlcount_reward_func": 0.36537498235702515,
|
| 953 |
+
"step": 59
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"completion_length": 326.0,
|
| 957 |
+
"epoch": 0.42857142857142855,
|
| 958 |
+
"grad_norm": 0.19174934923648834,
|
| 959 |
+
"kl": 0.13062816858291626,
|
| 960 |
+
"learning_rate": 3.5282177578265296e-05,
|
| 961 |
+
"loss": 0.0052,
|
| 962 |
+
"reward": 0.6150000095367432,
|
| 963 |
+
"reward_std": 0.7071067690849304,
|
| 964 |
+
"rewards/correctness_reward_func": 0.25,
|
| 965 |
+
"rewards/length_reward_func": 0.0,
|
| 966 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 967 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 968 |
+
"rewards/xmlcount_reward_func": 0.36500000953674316,
|
| 969 |
+
"step": 60
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"completion_length": 180.125,
|
| 973 |
+
"epoch": 0.4357142857142857,
|
| 974 |
+
"grad_norm": 0.48899680376052856,
|
| 975 |
+
"kl": 0.4511120319366455,
|
| 976 |
+
"learning_rate": 3.471086990686737e-05,
|
| 977 |
+
"loss": 0.018,
|
| 978 |
+
"reward": 2.324000120162964,
|
| 979 |
+
"reward_std": 0.11677447706460953,
|
| 980 |
+
"rewards/correctness_reward_func": 2.0,
|
| 981 |
+
"rewards/length_reward_func": 0.0,
|
| 982 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 983 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 984 |
+
"rewards/xmlcount_reward_func": 0.3240000009536743,
|
| 985 |
+
"step": 61
|
| 986 |
+
},
|
| 987 |
+
{
|
| 988 |
+
"completion_length": 442.0,
|
| 989 |
+
"epoch": 0.44285714285714284,
|
| 990 |
+
"grad_norm": 0.14630813896656036,
|
| 991 |
+
"kl": 0.11019708961248398,
|
| 992 |
+
"learning_rate": 3.413352560915988e-05,
|
| 993 |
+
"loss": 0.0044,
|
| 994 |
+
"reward": 0.17625001072883606,
|
| 995 |
+
"reward_std": 0.5338656306266785,
|
| 996 |
+
"rewards/correctness_reward_func": 0.0,
|
| 997 |
+
"rewards/length_reward_func": 0.0,
|
| 998 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 999 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1000 |
+
"rewards/xmlcount_reward_func": 0.17625001072883606,
|
| 1001 |
+
"step": 62
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"completion_length": 337.625,
|
| 1005 |
+
"epoch": 0.45,
|
| 1006 |
+
"grad_norm": 0.15681755542755127,
|
| 1007 |
+
"kl": 0.1342012584209442,
|
| 1008 |
+
"learning_rate": 3.355050358314172e-05,
|
| 1009 |
+
"loss": 0.0054,
|
| 1010 |
+
"reward": 1.1151249408721924,
|
| 1011 |
+
"reward_std": 1.0349948406219482,
|
| 1012 |
+
"rewards/correctness_reward_func": 0.75,
|
| 1013 |
+
"rewards/length_reward_func": 0.0,
|
| 1014 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1015 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1016 |
+
"rewards/xmlcount_reward_func": 0.36512500047683716,
|
| 1017 |
+
"step": 63
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"completion_length": 255.375,
|
| 1021 |
+
"epoch": 0.45714285714285713,
|
| 1022 |
+
"grad_norm": 0.15988513827323914,
|
| 1023 |
+
"kl": 0.14537671208381653,
|
| 1024 |
+
"learning_rate": 3.2962166256292113e-05,
|
| 1025 |
+
"loss": 0.0059,
|
| 1026 |
+
"reward": 2.36537504196167,
|
| 1027 |
+
"reward_std": 0.0005174623220227659,
|
| 1028 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1029 |
+
"rewards/length_reward_func": 0.0,
|
| 1030 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1031 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1032 |
+
"rewards/xmlcount_reward_func": 0.36537498235702515,
|
| 1033 |
+
"step": 64
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"completion_length": 303.875,
|
| 1037 |
+
"epoch": 0.4642857142857143,
|
| 1038 |
+
"grad_norm": 0.6239961981773376,
|
| 1039 |
+
"kl": 0.10595186054706573,
|
| 1040 |
+
"learning_rate": 3.2368879360272606e-05,
|
| 1041 |
+
"loss": 0.0042,
|
| 1042 |
+
"reward": 2.1156249046325684,
|
| 1043 |
+
"reward_std": 0.7073594331741333,
|
| 1044 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1045 |
+
"rewards/length_reward_func": 0.0,
|
| 1046 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1047 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1048 |
+
"rewards/xmlcount_reward_func": 0.3656249940395355,
|
| 1049 |
+
"step": 65
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"completion_length": 554.625,
|
| 1053 |
+
"epoch": 0.4714285714285714,
|
| 1054 |
+
"grad_norm": 0.0981917455792427,
|
| 1055 |
+
"kl": 0.06483708322048187,
|
| 1056 |
+
"learning_rate": 3.177101170357513e-05,
|
| 1057 |
+
"loss": 0.0026,
|
| 1058 |
+
"reward": 0.33550000190734863,
|
| 1059 |
+
"reward_std": 0.08505627512931824,
|
| 1060 |
+
"rewards/correctness_reward_func": 0.0,
|
| 1061 |
+
"rewards/length_reward_func": 0.0,
|
| 1062 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1063 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1064 |
+
"rewards/xmlcount_reward_func": 0.33550000190734863,
|
| 1065 |
+
"step": 66
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"completion_length": 164.0,
|
| 1069 |
+
"epoch": 0.4785714285714286,
|
| 1070 |
+
"grad_norm": 0.2880488634109497,
|
| 1071 |
+
"kl": 0.24433362483978271,
|
| 1072 |
+
"learning_rate": 3.116893494225734e-05,
|
| 1073 |
+
"loss": 0.0098,
|
| 1074 |
+
"reward": 0.41474997997283936,
|
| 1075 |
+
"reward_std": 0.3112049102783203,
|
| 1076 |
+
"rewards/correctness_reward_func": 0.0,
|
| 1077 |
+
"rewards/length_reward_func": 0.0,
|
| 1078 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1079 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1080 |
+
"rewards/xmlcount_reward_func": 0.35224997997283936,
|
| 1081 |
+
"step": 67
|
| 1082 |
+
},
|
| 1083 |
+
{
|
| 1084 |
+
"completion_length": 327.375,
|
| 1085 |
+
"epoch": 0.4857142857142857,
|
| 1086 |
+
"grad_norm": 0.15951593220233917,
|
| 1087 |
+
"kl": 0.09316325187683105,
|
| 1088 |
+
"learning_rate": 3.056302334890786e-05,
|
| 1089 |
+
"loss": 0.0037,
|
| 1090 |
+
"reward": 1.013374924659729,
|
| 1091 |
+
"reward_std": 0.8874584436416626,
|
| 1092 |
+
"rewards/correctness_reward_func": 0.75,
|
| 1093 |
+
"rewards/length_reward_func": 0.0,
|
| 1094 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1095 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1096 |
+
"rewards/xmlcount_reward_func": 0.2633749842643738,
|
| 1097 |
+
"step": 68
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"completion_length": 183.625,
|
| 1101 |
+
"epoch": 0.4928571428571429,
|
| 1102 |
+
"grad_norm": 0.19407892227172852,
|
| 1103 |
+
"kl": 0.2086183726787567,
|
| 1104 |
+
"learning_rate": 2.9953653579984942e-05,
|
| 1105 |
+
"loss": 0.0083,
|
| 1106 |
+
"reward": 2.365499973297119,
|
| 1107 |
+
"reward_std": 0.0005344200180843472,
|
| 1108 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1109 |
+
"rewards/length_reward_func": 0.0,
|
| 1110 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1111 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1112 |
+
"rewards/xmlcount_reward_func": 0.36549997329711914,
|
| 1113 |
+
"step": 69
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"completion_length": 260.375,
|
| 1117 |
+
"epoch": 0.5,
|
| 1118 |
+
"grad_norm": 0.2512143850326538,
|
| 1119 |
+
"kl": 0.14164677262306213,
|
| 1120 |
+
"learning_rate": 2.9341204441673266e-05,
|
| 1121 |
+
"loss": 0.0057,
|
| 1122 |
+
"reward": 2.020250082015991,
|
| 1123 |
+
"reward_std": 0.7143486142158508,
|
| 1124 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1125 |
+
"rewards/length_reward_func": 0.0,
|
| 1126 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1127 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1128 |
+
"rewards/xmlcount_reward_func": 0.27024999260902405,
|
| 1129 |
+
"step": 70
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"completion_length": 191.125,
|
| 1133 |
+
"epoch": 0.5071428571428571,
|
| 1134 |
+
"grad_norm": 0.3029208779335022,
|
| 1135 |
+
"kl": 0.11700859665870667,
|
| 1136 |
+
"learning_rate": 2.872605665440436e-05,
|
| 1137 |
+
"loss": 0.0047,
|
| 1138 |
+
"reward": 2.012125015258789,
|
| 1139 |
+
"reward_std": 0.7928214073181152,
|
| 1140 |
+
"rewards/correctness_reward_func": 1.5,
|
| 1141 |
+
"rewards/length_reward_func": 0.0,
|
| 1142 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1143 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1144 |
+
"rewards/xmlcount_reward_func": 0.44962501525878906,
|
| 1145 |
+
"step": 71
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"completion_length": 500.5,
|
| 1149 |
+
"epoch": 0.5142857142857142,
|
| 1150 |
+
"grad_norm": 0.12518590688705444,
|
| 1151 |
+
"kl": 0.02341982163488865,
|
| 1152 |
+
"learning_rate": 2.8108592616187133e-05,
|
| 1153 |
+
"loss": 0.0009,
|
| 1154 |
+
"reward": -0.19187498092651367,
|
| 1155 |
+
"reward_std": 0.7031666040420532,
|
| 1156 |
+
"rewards/correctness_reward_func": 0.0,
|
| 1157 |
+
"rewards/length_reward_func": 0.0,
|
| 1158 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1159 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1160 |
+
"rewards/xmlcount_reward_func": -0.19187499582767487,
|
| 1161 |
+
"step": 72
|
| 1162 |
+
},
|
| 1163 |
+
{
|
| 1164 |
+
"completion_length": 710.25,
|
| 1165 |
+
"epoch": 0.5214285714285715,
|
| 1166 |
+
"grad_norm": 0.11030412465333939,
|
| 1167 |
+
"kl": 0.01801430992782116,
|
| 1168 |
+
"learning_rate": 2.748919616489542e-05,
|
| 1169 |
+
"loss": 0.0007,
|
| 1170 |
+
"reward": -0.19962501525878906,
|
| 1171 |
+
"reward_std": 1.141786813735962,
|
| 1172 |
+
"rewards/correctness_reward_func": 0.0,
|
| 1173 |
+
"rewards/length_reward_func": 0.125,
|
| 1174 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1175 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1176 |
+
"rewards/xmlcount_reward_func": -0.3246249854564667,
|
| 1177 |
+
"step": 73
|
| 1178 |
+
},
|
| 1179 |
+
{
|
| 1180 |
+
"completion_length": 395.0,
|
| 1181 |
+
"epoch": 0.5285714285714286,
|
| 1182 |
+
"grad_norm": 0.08873239159584045,
|
| 1183 |
+
"kl": 0.02234470844268799,
|
| 1184 |
+
"learning_rate": 2.686825233966061e-05,
|
| 1185 |
+
"loss": 0.0009,
|
| 1186 |
+
"reward": 1.4856250286102295,
|
| 1187 |
+
"reward_std": 0.9701153039932251,
|
| 1188 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1189 |
+
"rewards/length_reward_func": 0.0,
|
| 1190 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1191 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1192 |
+
"rewards/xmlcount_reward_func": -0.2643750011920929,
|
| 1193 |
+
"step": 74
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"completion_length": 386.125,
|
| 1197 |
+
"epoch": 0.5357142857142857,
|
| 1198 |
+
"grad_norm": 0.09125195443630219,
|
| 1199 |
+
"kl": 0.029970454052090645,
|
| 1200 |
+
"learning_rate": 2.624614714151743e-05,
|
| 1201 |
+
"loss": 0.0012,
|
| 1202 |
+
"reward": 1.2940000295639038,
|
| 1203 |
+
"reward_std": 0.5996603965759277,
|
| 1204 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1205 |
+
"rewards/length_reward_func": 0.0,
|
| 1206 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1207 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1208 |
+
"rewards/xmlcount_reward_func": -0.45600003004074097,
|
| 1209 |
+
"step": 75
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"completion_length": 239.375,
|
| 1213 |
+
"epoch": 0.5428571428571428,
|
| 1214 |
+
"grad_norm": 0.1535108983516693,
|
| 1215 |
+
"kl": 0.05824340879917145,
|
| 1216 |
+
"learning_rate": 2.5623267293451826e-05,
|
| 1217 |
+
"loss": 0.0023,
|
| 1218 |
+
"reward": 1.394374966621399,
|
| 1219 |
+
"reward_std": 0.8498432040214539,
|
| 1220 |
+
"rewards/correctness_reward_func": 1.5,
|
| 1221 |
+
"rewards/length_reward_func": 0.0,
|
| 1222 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1223 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1224 |
+
"rewards/xmlcount_reward_func": -0.10562500357627869,
|
| 1225 |
+
"step": 76
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"completion_length": 351.625,
|
| 1229 |
+
"epoch": 0.55,
|
| 1230 |
+
"grad_norm": 0.15809431672096252,
|
| 1231 |
+
"kl": 0.02210851013660431,
|
| 1232 |
+
"learning_rate": 2.5e-05,
|
| 1233 |
+
"loss": 0.0009,
|
| 1234 |
+
"reward": 1.4718749523162842,
|
| 1235 |
+
"reward_std": 1.3246859312057495,
|
| 1236 |
+
"rewards/correctness_reward_func": 1.5,
|
| 1237 |
+
"rewards/length_reward_func": 0.0,
|
| 1238 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1239 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1240 |
+
"rewards/xmlcount_reward_func": -0.02812500298023224,
|
| 1241 |
+
"step": 77
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"completion_length": 199.375,
|
| 1245 |
+
"epoch": 0.5571428571428572,
|
| 1246 |
+
"grad_norm": 0.20733776688575745,
|
| 1247 |
+
"kl": 0.06919591128826141,
|
| 1248 |
+
"learning_rate": 2.4376732706548183e-05,
|
| 1249 |
+
"loss": 0.0028,
|
| 1250 |
+
"reward": 2.0334999561309814,
|
| 1251 |
+
"reward_std": 0.9354116320610046,
|
| 1252 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1253 |
+
"rewards/length_reward_func": 0.0,
|
| 1254 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1255 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1256 |
+
"rewards/xmlcount_reward_func": 0.22100000083446503,
|
| 1257 |
+
"step": 78
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"completion_length": 490.625,
|
| 1261 |
+
"epoch": 0.5642857142857143,
|
| 1262 |
+
"grad_norm": 0.3020256459712982,
|
| 1263 |
+
"kl": 0.020964641124010086,
|
| 1264 |
+
"learning_rate": 2.375385285848257e-05,
|
| 1265 |
+
"loss": 0.0008,
|
| 1266 |
+
"reward": 0.2606250047683716,
|
| 1267 |
+
"reward_std": 1.5538617372512817,
|
| 1268 |
+
"rewards/correctness_reward_func": 0.75,
|
| 1269 |
+
"rewards/length_reward_func": 0.125,
|
| 1270 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1271 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1272 |
+
"rewards/xmlcount_reward_func": -0.6143749952316284,
|
| 1273 |
+
"step": 79
|
| 1274 |
+
},
|
| 1275 |
+
{
|
| 1276 |
+
"completion_length": 257.875,
|
| 1277 |
+
"epoch": 0.5714285714285714,
|
| 1278 |
+
"grad_norm": 0.17333918809890747,
|
| 1279 |
+
"kl": 0.040599275380373,
|
| 1280 |
+
"learning_rate": 2.3131747660339394e-05,
|
| 1281 |
+
"loss": 0.0016,
|
| 1282 |
+
"reward": 1.8927500247955322,
|
| 1283 |
+
"reward_std": 0.6703715324401855,
|
| 1284 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1285 |
+
"rewards/length_reward_func": 0.0,
|
| 1286 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1287 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1288 |
+
"rewards/xmlcount_reward_func": 0.14274999499320984,
|
| 1289 |
+
"step": 80
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"completion_length": 292.625,
|
| 1293 |
+
"epoch": 0.5785714285714286,
|
| 1294 |
+
"grad_norm": 0.14176879823207855,
|
| 1295 |
+
"kl": 0.03724910691380501,
|
| 1296 |
+
"learning_rate": 2.251080383510459e-05,
|
| 1297 |
+
"loss": 0.0015,
|
| 1298 |
+
"reward": 2.164875030517578,
|
| 1299 |
+
"reward_std": 0.35948193073272705,
|
| 1300 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1301 |
+
"rewards/length_reward_func": 0.0,
|
| 1302 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1303 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1304 |
+
"rewards/xmlcount_reward_func": 0.16487500071525574,
|
| 1305 |
+
"step": 81
|
| 1306 |
+
},
|
| 1307 |
+
{
|
| 1308 |
+
"completion_length": 500.25,
|
| 1309 |
+
"epoch": 0.5857142857142857,
|
| 1310 |
+
"grad_norm": 0.1182783842086792,
|
| 1311 |
+
"kl": 0.023212479427456856,
|
| 1312 |
+
"learning_rate": 2.189140738381288e-05,
|
| 1313 |
+
"loss": 0.0009,
|
| 1314 |
+
"reward": 0.7456250190734863,
|
| 1315 |
+
"reward_std": 1.171677589416504,
|
| 1316 |
+
"rewards/correctness_reward_func": 0.5,
|
| 1317 |
+
"rewards/length_reward_func": 0.0,
|
| 1318 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1319 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1320 |
+
"rewards/xmlcount_reward_func": 0.24562500417232513,
|
| 1321 |
+
"step": 82
|
| 1322 |
+
},
|
| 1323 |
+
{
|
| 1324 |
+
"completion_length": 487.125,
|
| 1325 |
+
"epoch": 0.5928571428571429,
|
| 1326 |
+
"grad_norm": 0.12226809561252594,
|
| 1327 |
+
"kl": 0.019420940428972244,
|
| 1328 |
+
"learning_rate": 2.1273943345595637e-05,
|
| 1329 |
+
"loss": 0.0008,
|
| 1330 |
+
"reward": 0.3148750066757202,
|
| 1331 |
+
"reward_std": 1.1947816610336304,
|
| 1332 |
+
"rewards/correctness_reward_func": 0.25,
|
| 1333 |
+
"rewards/length_reward_func": 0.0,
|
| 1334 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1335 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1336 |
+
"rewards/xmlcount_reward_func": 0.06487500667572021,
|
| 1337 |
+
"step": 83
|
| 1338 |
+
},
|
| 1339 |
+
{
|
| 1340 |
+
"completion_length": 152.875,
|
| 1341 |
+
"epoch": 0.6,
|
| 1342 |
+
"grad_norm": 0.27960774302482605,
|
| 1343 |
+
"kl": 0.07769262790679932,
|
| 1344 |
+
"learning_rate": 2.0658795558326743e-05,
|
| 1345 |
+
"loss": 0.0031,
|
| 1346 |
+
"reward": 2.5625,
|
| 1347 |
+
"reward_std": 0.1767766922712326,
|
| 1348 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1349 |
+
"rewards/length_reward_func": 0.0,
|
| 1350 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1351 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1352 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1353 |
+
"step": 84
|
| 1354 |
+
},
|
| 1355 |
+
{
|
| 1356 |
+
"completion_length": 235.75,
|
| 1357 |
+
"epoch": 0.6071428571428571,
|
| 1358 |
+
"grad_norm": 0.2650700807571411,
|
| 1359 |
+
"kl": 0.0834224745631218,
|
| 1360 |
+
"learning_rate": 2.0046346420015067e-05,
|
| 1361 |
+
"loss": 0.0033,
|
| 1362 |
+
"reward": 2.243499994277954,
|
| 1363 |
+
"reward_std": 0.5382997393608093,
|
| 1364 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1365 |
+
"rewards/length_reward_func": 0.0,
|
| 1366 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1367 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1368 |
+
"rewards/xmlcount_reward_func": 0.4309999942779541,
|
| 1369 |
+
"step": 85
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"completion_length": 324.0,
|
| 1373 |
+
"epoch": 0.6142857142857143,
|
| 1374 |
+
"grad_norm": 0.00965797994285822,
|
| 1375 |
+
"kl": 0.04011977091431618,
|
| 1376 |
+
"learning_rate": 1.9436976651092144e-05,
|
| 1377 |
+
"loss": 0.0016,
|
| 1378 |
+
"reward": 0.5,
|
| 1379 |
+
"reward_std": 0.0,
|
| 1380 |
+
"rewards/correctness_reward_func": 0.0,
|
| 1381 |
+
"rewards/length_reward_func": 0.0,
|
| 1382 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1383 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1384 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1385 |
+
"step": 86
|
| 1386 |
+
},
|
| 1387 |
+
{
|
| 1388 |
+
"completion_length": 227.0,
|
| 1389 |
+
"epoch": 0.6214285714285714,
|
| 1390 |
+
"grad_norm": 0.19126032292842865,
|
| 1391 |
+
"kl": 0.12643316388130188,
|
| 1392 |
+
"learning_rate": 1.8831065057742657e-05,
|
| 1393 |
+
"loss": 0.0051,
|
| 1394 |
+
"reward": 2.5,
|
| 1395 |
+
"reward_std": 0.0,
|
| 1396 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1397 |
+
"rewards/length_reward_func": 0.0,
|
| 1398 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1399 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1400 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1401 |
+
"step": 87
|
| 1402 |
+
},
|
| 1403 |
+
{
|
| 1404 |
+
"completion_length": 510.75,
|
| 1405 |
+
"epoch": 0.6285714285714286,
|
| 1406 |
+
"grad_norm": 0.003844304708763957,
|
| 1407 |
+
"kl": 0.016257166862487793,
|
| 1408 |
+
"learning_rate": 1.8228988296424877e-05,
|
| 1409 |
+
"loss": 0.0007,
|
| 1410 |
+
"reward": 0.5,
|
| 1411 |
+
"reward_std": 0.0,
|
| 1412 |
+
"rewards/correctness_reward_func": 0.0,
|
| 1413 |
+
"rewards/length_reward_func": 0.0,
|
| 1414 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1415 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1416 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1417 |
+
"step": 88
|
| 1418 |
+
},
|
| 1419 |
+
{
|
| 1420 |
+
"completion_length": 243.0,
|
| 1421 |
+
"epoch": 0.6357142857142857,
|
| 1422 |
+
"grad_norm": 0.008268176577985287,
|
| 1423 |
+
"kl": 0.045144014060497284,
|
| 1424 |
+
"learning_rate": 1.7631120639727393e-05,
|
| 1425 |
+
"loss": 0.0018,
|
| 1426 |
+
"reward": 2.5,
|
| 1427 |
+
"reward_std": 0.0,
|
| 1428 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1429 |
+
"rewards/length_reward_func": 0.0,
|
| 1430 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1431 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1432 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1433 |
+
"step": 89
|
| 1434 |
+
},
|
| 1435 |
+
{
|
| 1436 |
+
"completion_length": 235.25,
|
| 1437 |
+
"epoch": 0.6428571428571429,
|
| 1438 |
+
"grad_norm": 0.14653581380844116,
|
| 1439 |
+
"kl": 0.038456518203020096,
|
| 1440 |
+
"learning_rate": 1.7037833743707892e-05,
|
| 1441 |
+
"loss": 0.0015,
|
| 1442 |
+
"reward": 2.375,
|
| 1443 |
+
"reward_std": 0.5824823379516602,
|
| 1444 |
+
"rewards/correctness_reward_func": 1.75,
|
| 1445 |
+
"rewards/length_reward_func": 0.0,
|
| 1446 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1447 |
+
"rewards/strict_format_reward_func": 0.125,
|
| 1448 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1449 |
+
"step": 90
|
| 1450 |
+
},
|
| 1451 |
+
{
|
| 1452 |
+
"completion_length": 220.125,
|
| 1453 |
+
"epoch": 0.65,
|
| 1454 |
+
"grad_norm": 0.14805400371551514,
|
| 1455 |
+
"kl": 0.07326260954141617,
|
| 1456 |
+
"learning_rate": 1.6449496416858284e-05,
|
| 1457 |
+
"loss": 0.0029,
|
| 1458 |
+
"reward": 1.3125,
|
| 1459 |
+
"reward_std": 1.1319231986999512,
|
| 1460 |
+
"rewards/correctness_reward_func": 0.75,
|
| 1461 |
+
"rewards/length_reward_func": 0.0,
|
| 1462 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1463 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1464 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1465 |
+
"step": 91
|
| 1466 |
+
},
|
| 1467 |
+
{
|
| 1468 |
+
"completion_length": 449.0,
|
| 1469 |
+
"epoch": 0.6571428571428571,
|
| 1470 |
+
"grad_norm": 0.11858955770730972,
|
| 1471 |
+
"kl": 0.027846258133649826,
|
| 1472 |
+
"learning_rate": 1.5866474390840125e-05,
|
| 1473 |
+
"loss": 0.0011,
|
| 1474 |
+
"reward": 0.75,
|
| 1475 |
+
"reward_std": 0.7071067690849304,
|
| 1476 |
+
"rewards/correctness_reward_func": 0.25,
|
| 1477 |
+
"rewards/length_reward_func": 0.0,
|
| 1478 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1479 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1480 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1481 |
+
"step": 92
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"completion_length": 219.5,
|
| 1485 |
+
"epoch": 0.6642857142857143,
|
| 1486 |
+
"grad_norm": 0.21257686614990234,
|
| 1487 |
+
"kl": 0.058622974902391434,
|
| 1488 |
+
"learning_rate": 1.5289130093132632e-05,
|
| 1489 |
+
"loss": 0.0023,
|
| 1490 |
+
"reward": 1.046875,
|
| 1491 |
+
"reward_std": 0.9159731268882751,
|
| 1492 |
+
"rewards/correctness_reward_func": 0.5,
|
| 1493 |
+
"rewards/length_reward_func": 0.0,
|
| 1494 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1495 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1496 |
+
"rewards/xmlcount_reward_func": 0.484375,
|
| 1497 |
+
"step": 93
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"completion_length": 191.125,
|
| 1501 |
+
"epoch": 0.6714285714285714,
|
| 1502 |
+
"grad_norm": 0.017704889178276062,
|
| 1503 |
+
"kl": 0.06602377444505692,
|
| 1504 |
+
"learning_rate": 1.4717822421734718e-05,
|
| 1505 |
+
"loss": 0.0026,
|
| 1506 |
+
"reward": 2.5,
|
| 1507 |
+
"reward_std": 0.0,
|
| 1508 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1509 |
+
"rewards/length_reward_func": 0.0,
|
| 1510 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1511 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1512 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1513 |
+
"step": 94
|
| 1514 |
+
},
|
| 1515 |
+
{
|
| 1516 |
+
"completion_length": 727.75,
|
| 1517 |
+
"epoch": 0.6785714285714286,
|
| 1518 |
+
"grad_norm": 0.07640424370765686,
|
| 1519 |
+
"kl": 0.015565132722258568,
|
| 1520 |
+
"learning_rate": 1.4152906522061048e-05,
|
| 1521 |
+
"loss": 0.0006,
|
| 1522 |
+
"reward": 1.875,
|
| 1523 |
+
"reward_std": 1.0264363288879395,
|
| 1524 |
+
"rewards/correctness_reward_func": 1.25,
|
| 1525 |
+
"rewards/length_reward_func": 0.125,
|
| 1526 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1527 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1528 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1529 |
+
"step": 95
|
| 1530 |
+
},
|
| 1531 |
+
{
|
| 1532 |
+
"completion_length": 333.625,
|
| 1533 |
+
"epoch": 0.6857142857142857,
|
| 1534 |
+
"grad_norm": 0.011164901778101921,
|
| 1535 |
+
"kl": 0.03891594707965851,
|
| 1536 |
+
"learning_rate": 1.3594733566170926e-05,
|
| 1537 |
+
"loss": 0.0016,
|
| 1538 |
+
"reward": 2.5,
|
| 1539 |
+
"reward_std": 0.0,
|
| 1540 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1541 |
+
"rewards/length_reward_func": 0.0,
|
| 1542 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1543 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1544 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1545 |
+
"step": 96
|
| 1546 |
+
},
|
| 1547 |
+
{
|
| 1548 |
+
"completion_length": 795.5,
|
| 1549 |
+
"epoch": 0.6928571428571428,
|
| 1550 |
+
"grad_norm": 0.07001640647649765,
|
| 1551 |
+
"kl": 0.014953548088669777,
|
| 1552 |
+
"learning_rate": 1.3043650534467053e-05,
|
| 1553 |
+
"loss": 0.0006,
|
| 1554 |
+
"reward": 1.203125,
|
| 1555 |
+
"reward_std": 0.9863747954368591,
|
| 1556 |
+
"rewards/correctness_reward_func": 0.5,
|
| 1557 |
+
"rewards/length_reward_func": 0.25,
|
| 1558 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1559 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1560 |
+
"rewards/xmlcount_reward_func": 0.453125,
|
| 1561 |
+
"step": 97
|
| 1562 |
+
},
|
| 1563 |
+
{
|
| 1564 |
+
"completion_length": 342.5,
|
| 1565 |
+
"epoch": 0.7,
|
| 1566 |
+
"grad_norm": 0.015015755780041218,
|
| 1567 |
+
"kl": 0.03935668244957924,
|
| 1568 |
+
"learning_rate": 1.2500000000000006e-05,
|
| 1569 |
+
"loss": 0.0016,
|
| 1570 |
+
"reward": 2.5,
|
| 1571 |
+
"reward_std": 0.0,
|
| 1572 |
+
"rewards/correctness_reward_func": 2.0,
|
| 1573 |
+
"rewards/length_reward_func": 0.0,
|
| 1574 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1575 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1576 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1577 |
+
"step": 98
|
| 1578 |
+
},
|
| 1579 |
+
{
|
| 1580 |
+
"completion_length": 550.25,
|
| 1581 |
+
"epoch": 0.7071428571428572,
|
| 1582 |
+
"grad_norm": 0.10194458067417145,
|
| 1583 |
+
"kl": 0.022631347179412842,
|
| 1584 |
+
"learning_rate": 1.196411991551255e-05,
|
| 1585 |
+
"loss": 0.0009,
|
| 1586 |
+
"reward": 1.5,
|
| 1587 |
+
"reward_std": 1.0690449476242065,
|
| 1588 |
+
"rewards/correctness_reward_func": 1.0,
|
| 1589 |
+
"rewards/length_reward_func": 0.0,
|
| 1590 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1591 |
+
"rewards/strict_format_reward_func": 0.0,
|
| 1592 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1593 |
+
"step": 99
|
| 1594 |
+
},
|
| 1595 |
+
{
|
| 1596 |
+
"completion_length": 247.375,
|
| 1597 |
+
"epoch": 0.7142857142857143,
|
| 1598 |
+
"grad_norm": 0.15261337161064148,
|
| 1599 |
+
"kl": 0.08087805658578873,
|
| 1600 |
+
"learning_rate": 1.1436343403356017e-05,
|
| 1601 |
+
"loss": 0.0032,
|
| 1602 |
+
"reward": 2.0625,
|
| 1603 |
+
"reward_std": 0.979704737663269,
|
| 1604 |
+
"rewards/correctness_reward_func": 1.5,
|
| 1605 |
+
"rewards/length_reward_func": 0.0,
|
| 1606 |
+
"rewards/soft_format_reward_func": 0.0,
|
| 1607 |
+
"rewards/strict_format_reward_func": 0.0625,
|
| 1608 |
+
"rewards/xmlcount_reward_func": 0.5,
|
| 1609 |
+
"step": 100
|
| 1610 |
+
}
|
| 1611 |
+
],
|
| 1612 |
+
"logging_steps": 1,
|
| 1613 |
+
"max_steps": 140,
|
| 1614 |
+
"num_input_tokens_seen": 0,
|
| 1615 |
+
"num_train_epochs": 1,
|
| 1616 |
+
"save_steps": 100,
|
| 1617 |
+
"stateful_callbacks": {
|
| 1618 |
+
"TrainerControl": {
|
| 1619 |
+
"args": {
|
| 1620 |
+
"should_epoch_stop": false,
|
| 1621 |
+
"should_evaluate": false,
|
| 1622 |
+
"should_log": false,
|
| 1623 |
+
"should_save": true,
|
| 1624 |
+
"should_training_stop": false
|
| 1625 |
+
},
|
| 1626 |
+
"attributes": {}
|
| 1627 |
+
}
|
| 1628 |
+
},
|
| 1629 |
+
"total_flos": 0.0,
|
| 1630 |
+
"train_batch_size": 8,
|
| 1631 |
+
"trial_name": null,
|
| 1632 |
+
"trial_params": null
|
| 1633 |
+
}
|