Upload trainer_state.json
Browse files- trainer_state.json +3604 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,3604 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 47.47152290548906,
|
| 5 |
+
"global_step": 46000,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.1,
|
| 12 |
+
"learning_rate": 9.900000000000002e-06,
|
| 13 |
+
"loss": 5.0766,
|
| 14 |
+
"step": 100
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.21,
|
| 18 |
+
"learning_rate": 1.9900000000000003e-05,
|
| 19 |
+
"loss": 2.8354,
|
| 20 |
+
"step": 200
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.31,
|
| 24 |
+
"learning_rate": 2.9900000000000002e-05,
|
| 25 |
+
"loss": 1.7963,
|
| 26 |
+
"step": 300
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.41,
|
| 30 |
+
"learning_rate": 3.99e-05,
|
| 31 |
+
"loss": 1.3003,
|
| 32 |
+
"step": 400
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.52,
|
| 36 |
+
"learning_rate": 4.99e-05,
|
| 37 |
+
"loss": 1.2223,
|
| 38 |
+
"step": 500
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.52,
|
| 42 |
+
"eval_loss": 0.38657668232917786,
|
| 43 |
+
"eval_runtime": 473.1633,
|
| 44 |
+
"eval_samples_per_second": 22.142,
|
| 45 |
+
"eval_steps_per_second": 2.769,
|
| 46 |
+
"eval_wer": 0.3425097378671439,
|
| 47 |
+
"step": 500
|
| 48 |
+
},
|
| 49 |
+
{
|
| 50 |
+
"epoch": 0.62,
|
| 51 |
+
"learning_rate": 4.989676746611053e-05,
|
| 52 |
+
"loss": 1.1977,
|
| 53 |
+
"step": 600
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"epoch": 0.72,
|
| 57 |
+
"learning_rate": 4.979249217935349e-05,
|
| 58 |
+
"loss": 1.1405,
|
| 59 |
+
"step": 700
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 0.83,
|
| 63 |
+
"learning_rate": 4.9688216892596456e-05,
|
| 64 |
+
"loss": 1.1294,
|
| 65 |
+
"step": 800
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.93,
|
| 69 |
+
"learning_rate": 4.958394160583942e-05,
|
| 70 |
+
"loss": 1.108,
|
| 71 |
+
"step": 900
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 1.03,
|
| 75 |
+
"learning_rate": 4.947966631908238e-05,
|
| 76 |
+
"loss": 1.0748,
|
| 77 |
+
"step": 1000
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"epoch": 1.03,
|
| 81 |
+
"eval_loss": 0.25742873549461365,
|
| 82 |
+
"eval_runtime": 470.5268,
|
| 83 |
+
"eval_samples_per_second": 22.267,
|
| 84 |
+
"eval_steps_per_second": 2.784,
|
| 85 |
+
"eval_wer": 0.2169386251184335,
|
| 86 |
+
"step": 1000
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 1.14,
|
| 90 |
+
"learning_rate": 4.937539103232534e-05,
|
| 91 |
+
"loss": 1.0635,
|
| 92 |
+
"step": 1100
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 1.24,
|
| 96 |
+
"learning_rate": 4.9271115745568305e-05,
|
| 97 |
+
"loss": 1.0427,
|
| 98 |
+
"step": 1200
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 1.34,
|
| 102 |
+
"learning_rate": 4.916684045881127e-05,
|
| 103 |
+
"loss": 1.0402,
|
| 104 |
+
"step": 1300
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 1.44,
|
| 108 |
+
"learning_rate": 4.906256517205422e-05,
|
| 109 |
+
"loss": 1.0445,
|
| 110 |
+
"step": 1400
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 1.55,
|
| 114 |
+
"learning_rate": 4.895933263816476e-05,
|
| 115 |
+
"loss": 1.0416,
|
| 116 |
+
"step": 1500
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.55,
|
| 120 |
+
"eval_loss": 0.21774140000343323,
|
| 121 |
+
"eval_runtime": 466.1858,
|
| 122 |
+
"eval_samples_per_second": 22.474,
|
| 123 |
+
"eval_steps_per_second": 2.81,
|
| 124 |
+
"eval_wer": 0.19455732182334984,
|
| 125 |
+
"step": 1500
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"epoch": 1.65,
|
| 129 |
+
"learning_rate": 4.885505735140772e-05,
|
| 130 |
+
"loss": 1.039,
|
| 131 |
+
"step": 1600
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"epoch": 1.75,
|
| 135 |
+
"learning_rate": 4.875078206465068e-05,
|
| 136 |
+
"loss": 1.0274,
|
| 137 |
+
"step": 1700
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.86,
|
| 141 |
+
"learning_rate": 4.864650677789364e-05,
|
| 142 |
+
"loss": 1.0116,
|
| 143 |
+
"step": 1800
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"epoch": 1.96,
|
| 147 |
+
"learning_rate": 4.85422314911366e-05,
|
| 148 |
+
"loss": 1.0057,
|
| 149 |
+
"step": 1900
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 2.06,
|
| 153 |
+
"learning_rate": 4.843795620437956e-05,
|
| 154 |
+
"loss": 0.9951,
|
| 155 |
+
"step": 2000
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"epoch": 2.06,
|
| 159 |
+
"eval_loss": 0.2008274346590042,
|
| 160 |
+
"eval_runtime": 468.5372,
|
| 161 |
+
"eval_samples_per_second": 22.361,
|
| 162 |
+
"eval_steps_per_second": 2.796,
|
| 163 |
+
"eval_wer": 0.17599747341825456,
|
| 164 |
+
"step": 2000
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 2.17,
|
| 168 |
+
"learning_rate": 4.8333680917622525e-05,
|
| 169 |
+
"loss": 0.9822,
|
| 170 |
+
"step": 2100
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 2.27,
|
| 174 |
+
"learning_rate": 4.822940563086549e-05,
|
| 175 |
+
"loss": 0.9847,
|
| 176 |
+
"step": 2200
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 2.37,
|
| 180 |
+
"learning_rate": 4.812513034410845e-05,
|
| 181 |
+
"loss": 0.9909,
|
| 182 |
+
"step": 2300
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 2.48,
|
| 186 |
+
"learning_rate": 4.802085505735141e-05,
|
| 187 |
+
"loss": 0.983,
|
| 188 |
+
"step": 2400
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.58,
|
| 192 |
+
"learning_rate": 4.7916579770594374e-05,
|
| 193 |
+
"loss": 0.975,
|
| 194 |
+
"step": 2500
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 2.58,
|
| 198 |
+
"eval_loss": 0.19611841440200806,
|
| 199 |
+
"eval_runtime": 467.2599,
|
| 200 |
+
"eval_samples_per_second": 22.422,
|
| 201 |
+
"eval_steps_per_second": 2.804,
|
| 202 |
+
"eval_wer": 0.17505000526371198,
|
| 203 |
+
"step": 2500
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"epoch": 2.68,
|
| 207 |
+
"learning_rate": 4.7812304483837336e-05,
|
| 208 |
+
"loss": 0.9699,
|
| 209 |
+
"step": 2600
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.79,
|
| 213 |
+
"learning_rate": 4.77080291970803e-05,
|
| 214 |
+
"loss": 0.9829,
|
| 215 |
+
"step": 2700
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"epoch": 2.89,
|
| 219 |
+
"learning_rate": 4.7603753910323254e-05,
|
| 220 |
+
"loss": 0.9676,
|
| 221 |
+
"step": 2800
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 2.99,
|
| 225 |
+
"learning_rate": 4.7499478623566216e-05,
|
| 226 |
+
"loss": 0.9824,
|
| 227 |
+
"step": 2900
|
| 228 |
+
},
|
| 229 |
+
{
|
| 230 |
+
"epoch": 3.1,
|
| 231 |
+
"learning_rate": 4.739520333680918e-05,
|
| 232 |
+
"loss": 0.9461,
|
| 233 |
+
"step": 3000
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 3.1,
|
| 237 |
+
"eval_loss": 0.19894707202911377,
|
| 238 |
+
"eval_runtime": 467.4994,
|
| 239 |
+
"eval_samples_per_second": 22.411,
|
| 240 |
+
"eval_steps_per_second": 2.802,
|
| 241 |
+
"eval_wer": 0.1782082324455206,
|
| 242 |
+
"step": 3000
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 3.2,
|
| 246 |
+
"learning_rate": 4.729092805005214e-05,
|
| 247 |
+
"loss": 0.9551,
|
| 248 |
+
"step": 3100
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 3.3,
|
| 252 |
+
"learning_rate": 4.71866527632951e-05,
|
| 253 |
+
"loss": 0.961,
|
| 254 |
+
"step": 3200
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 3.41,
|
| 258 |
+
"learning_rate": 4.7082377476538065e-05,
|
| 259 |
+
"loss": 0.9533,
|
| 260 |
+
"step": 3300
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 3.51,
|
| 264 |
+
"learning_rate": 4.697810218978103e-05,
|
| 265 |
+
"loss": 0.9522,
|
| 266 |
+
"step": 3400
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 3.61,
|
| 270 |
+
"learning_rate": 4.687382690302399e-05,
|
| 271 |
+
"loss": 0.9381,
|
| 272 |
+
"step": 3500
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 3.61,
|
| 276 |
+
"eval_loss": 0.1928330808877945,
|
| 277 |
+
"eval_runtime": 470.0863,
|
| 278 |
+
"eval_samples_per_second": 22.287,
|
| 279 |
+
"eval_steps_per_second": 2.787,
|
| 280 |
+
"eval_wer": 0.1699126223813033,
|
| 281 |
+
"step": 3500
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 3.72,
|
| 285 |
+
"learning_rate": 4.676955161626695e-05,
|
| 286 |
+
"loss": 0.9567,
|
| 287 |
+
"step": 3600
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"epoch": 3.82,
|
| 291 |
+
"learning_rate": 4.666527632950991e-05,
|
| 292 |
+
"loss": 0.9431,
|
| 293 |
+
"step": 3700
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 3.92,
|
| 297 |
+
"learning_rate": 4.656100104275287e-05,
|
| 298 |
+
"loss": 0.9537,
|
| 299 |
+
"step": 3800
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 4.02,
|
| 303 |
+
"learning_rate": 4.645672575599583e-05,
|
| 304 |
+
"loss": 0.9193,
|
| 305 |
+
"step": 3900
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 4.13,
|
| 309 |
+
"learning_rate": 4.6352450469238794e-05,
|
| 310 |
+
"loss": 0.934,
|
| 311 |
+
"step": 4000
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 4.13,
|
| 315 |
+
"eval_loss": 0.1922750324010849,
|
| 316 |
+
"eval_runtime": 484.6325,
|
| 317 |
+
"eval_samples_per_second": 21.618,
|
| 318 |
+
"eval_steps_per_second": 2.703,
|
| 319 |
+
"eval_wer": 0.16326981787556585,
|
| 320 |
+
"step": 4000
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 4.23,
|
| 324 |
+
"learning_rate": 4.6248175182481756e-05,
|
| 325 |
+
"loss": 0.9179,
|
| 326 |
+
"step": 4100
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 4.33,
|
| 330 |
+
"learning_rate": 4.614389989572472e-05,
|
| 331 |
+
"loss": 0.9291,
|
| 332 |
+
"step": 4200
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 4.44,
|
| 336 |
+
"learning_rate": 4.6039624608967674e-05,
|
| 337 |
+
"loss": 0.9111,
|
| 338 |
+
"step": 4300
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 4.54,
|
| 342 |
+
"learning_rate": 4.5935349322210636e-05,
|
| 343 |
+
"loss": 0.924,
|
| 344 |
+
"step": 4400
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 4.64,
|
| 348 |
+
"learning_rate": 4.58310740354536e-05,
|
| 349 |
+
"loss": 0.9322,
|
| 350 |
+
"step": 4500
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 4.64,
|
| 354 |
+
"eval_loss": 0.1871235966682434,
|
| 355 |
+
"eval_runtime": 466.8923,
|
| 356 |
+
"eval_samples_per_second": 22.44,
|
| 357 |
+
"eval_steps_per_second": 2.806,
|
| 358 |
+
"eval_wer": 0.1634277292346563,
|
| 359 |
+
"step": 4500
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 4.75,
|
| 363 |
+
"learning_rate": 4.572679874869656e-05,
|
| 364 |
+
"loss": 0.936,
|
| 365 |
+
"step": 4600
|
| 366 |
+
},
|
| 367 |
+
{
|
| 368 |
+
"epoch": 4.85,
|
| 369 |
+
"learning_rate": 4.562252346193952e-05,
|
| 370 |
+
"loss": 0.926,
|
| 371 |
+
"step": 4700
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 4.95,
|
| 375 |
+
"learning_rate": 4.551824817518248e-05,
|
| 376 |
+
"loss": 0.9285,
|
| 377 |
+
"step": 4800
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 5.06,
|
| 381 |
+
"learning_rate": 4.541397288842544e-05,
|
| 382 |
+
"loss": 0.9293,
|
| 383 |
+
"step": 4900
|
| 384 |
+
},
|
| 385 |
+
{
|
| 386 |
+
"epoch": 5.16,
|
| 387 |
+
"learning_rate": 4.53096976016684e-05,
|
| 388 |
+
"loss": 0.9012,
|
| 389 |
+
"step": 5000
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 5.16,
|
| 393 |
+
"eval_loss": 0.1889733076095581,
|
| 394 |
+
"eval_runtime": 464.7576,
|
| 395 |
+
"eval_samples_per_second": 22.543,
|
| 396 |
+
"eval_steps_per_second": 2.819,
|
| 397 |
+
"eval_wer": 0.1701863354037267,
|
| 398 |
+
"step": 5000
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 5.26,
|
| 402 |
+
"learning_rate": 4.5205422314911365e-05,
|
| 403 |
+
"loss": 0.8889,
|
| 404 |
+
"step": 5100
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 5.37,
|
| 408 |
+
"learning_rate": 4.510114702815433e-05,
|
| 409 |
+
"loss": 0.907,
|
| 410 |
+
"step": 5200
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 5.47,
|
| 414 |
+
"learning_rate": 4.499687174139729e-05,
|
| 415 |
+
"loss": 0.9002,
|
| 416 |
+
"step": 5300
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 5.57,
|
| 420 |
+
"learning_rate": 4.489259645464025e-05,
|
| 421 |
+
"loss": 0.8984,
|
| 422 |
+
"step": 5400
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 5.68,
|
| 426 |
+
"learning_rate": 4.4788321167883214e-05,
|
| 427 |
+
"loss": 0.9045,
|
| 428 |
+
"step": 5500
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 5.68,
|
| 432 |
+
"eval_loss": 0.188216432929039,
|
| 433 |
+
"eval_runtime": 466.8913,
|
| 434 |
+
"eval_samples_per_second": 22.44,
|
| 435 |
+
"eval_steps_per_second": 2.806,
|
| 436 |
+
"eval_wer": 0.1739867354458364,
|
| 437 |
+
"step": 5500
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"epoch": 5.78,
|
| 441 |
+
"learning_rate": 4.468404588112618e-05,
|
| 442 |
+
"loss": 0.8995,
|
| 443 |
+
"step": 5600
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 5.88,
|
| 447 |
+
"learning_rate": 4.457977059436914e-05,
|
| 448 |
+
"loss": 0.9113,
|
| 449 |
+
"step": 5700
|
| 450 |
+
},
|
| 451 |
+
{
|
| 452 |
+
"epoch": 5.99,
|
| 453 |
+
"learning_rate": 4.4475495307612094e-05,
|
| 454 |
+
"loss": 0.9044,
|
| 455 |
+
"step": 5800
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 6.09,
|
| 459 |
+
"learning_rate": 4.437122002085506e-05,
|
| 460 |
+
"loss": 0.8822,
|
| 461 |
+
"step": 5900
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 6.19,
|
| 465 |
+
"learning_rate": 4.426694473409802e-05,
|
| 466 |
+
"loss": 0.8826,
|
| 467 |
+
"step": 6000
|
| 468 |
+
},
|
| 469 |
+
{
|
| 470 |
+
"epoch": 6.19,
|
| 471 |
+
"eval_loss": 0.18564490973949432,
|
| 472 |
+
"eval_runtime": 465.7133,
|
| 473 |
+
"eval_samples_per_second": 22.497,
|
| 474 |
+
"eval_steps_per_second": 2.813,
|
| 475 |
+
"eval_wer": 0.15750078955679545,
|
| 476 |
+
"step": 6000
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 6.3,
|
| 480 |
+
"learning_rate": 4.416266944734098e-05,
|
| 481 |
+
"loss": 0.8986,
|
| 482 |
+
"step": 6100
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 6.4,
|
| 486 |
+
"learning_rate": 4.4058394160583944e-05,
|
| 487 |
+
"loss": 0.8936,
|
| 488 |
+
"step": 6200
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 6.5,
|
| 492 |
+
"learning_rate": 4.395516162669447e-05,
|
| 493 |
+
"loss": 0.889,
|
| 494 |
+
"step": 6300
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 6.6,
|
| 498 |
+
"learning_rate": 4.3850886339937434e-05,
|
| 499 |
+
"loss": 0.883,
|
| 500 |
+
"step": 6400
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 6.71,
|
| 504 |
+
"learning_rate": 4.37466110531804e-05,
|
| 505 |
+
"loss": 0.8848,
|
| 506 |
+
"step": 6500
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 6.71,
|
| 510 |
+
"eval_loss": 0.1860707700252533,
|
| 511 |
+
"eval_runtime": 479.2504,
|
| 512 |
+
"eval_samples_per_second": 21.861,
|
| 513 |
+
"eval_steps_per_second": 2.733,
|
| 514 |
+
"eval_wer": 0.16168017686072217,
|
| 515 |
+
"step": 6500
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 6.81,
|
| 519 |
+
"learning_rate": 4.364233576642336e-05,
|
| 520 |
+
"loss": 0.8986,
|
| 521 |
+
"step": 6600
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 6.91,
|
| 525 |
+
"learning_rate": 4.353806047966632e-05,
|
| 526 |
+
"loss": 0.8862,
|
| 527 |
+
"step": 6700
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 7.02,
|
| 531 |
+
"learning_rate": 4.3433785192909284e-05,
|
| 532 |
+
"loss": 0.8704,
|
| 533 |
+
"step": 6800
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 7.12,
|
| 537 |
+
"learning_rate": 4.3329509906152246e-05,
|
| 538 |
+
"loss": 0.8687,
|
| 539 |
+
"step": 6900
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 7.22,
|
| 543 |
+
"learning_rate": 4.322523461939521e-05,
|
| 544 |
+
"loss": 0.8723,
|
| 545 |
+
"step": 7000
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"epoch": 7.22,
|
| 549 |
+
"eval_loss": 0.19270716607570648,
|
| 550 |
+
"eval_runtime": 470.8538,
|
| 551 |
+
"eval_samples_per_second": 22.251,
|
| 552 |
+
"eval_steps_per_second": 2.782,
|
| 553 |
+
"eval_wer": 0.16464891041162227,
|
| 554 |
+
"step": 7000
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 7.33,
|
| 558 |
+
"learning_rate": 4.3120959332638164e-05,
|
| 559 |
+
"loss": 0.8732,
|
| 560 |
+
"step": 7100
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 7.43,
|
| 564 |
+
"learning_rate": 4.3016684045881126e-05,
|
| 565 |
+
"loss": 0.8726,
|
| 566 |
+
"step": 7200
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 7.53,
|
| 570 |
+
"learning_rate": 4.291240875912409e-05,
|
| 571 |
+
"loss": 0.8728,
|
| 572 |
+
"step": 7300
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 7.64,
|
| 576 |
+
"learning_rate": 4.280813347236705e-05,
|
| 577 |
+
"loss": 0.8617,
|
| 578 |
+
"step": 7400
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 7.74,
|
| 582 |
+
"learning_rate": 4.270385818561001e-05,
|
| 583 |
+
"loss": 0.8725,
|
| 584 |
+
"step": 7500
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 7.74,
|
| 588 |
+
"eval_loss": 0.17976288497447968,
|
| 589 |
+
"eval_runtime": 466.2208,
|
| 590 |
+
"eval_samples_per_second": 22.472,
|
| 591 |
+
"eval_steps_per_second": 2.81,
|
| 592 |
+
"eval_wer": 0.15313190862196022,
|
| 593 |
+
"step": 7500
|
| 594 |
+
},
|
| 595 |
+
{
|
| 596 |
+
"epoch": 7.84,
|
| 597 |
+
"learning_rate": 4.2599582898852975e-05,
|
| 598 |
+
"loss": 0.8712,
|
| 599 |
+
"step": 7600
|
| 600 |
+
},
|
| 601 |
+
{
|
| 602 |
+
"epoch": 7.95,
|
| 603 |
+
"learning_rate": 4.249530761209594e-05,
|
| 604 |
+
"loss": 0.8639,
|
| 605 |
+
"step": 7700
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 8.05,
|
| 609 |
+
"learning_rate": 4.23910323253389e-05,
|
| 610 |
+
"loss": 0.8909,
|
| 611 |
+
"step": 7800
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 8.15,
|
| 615 |
+
"learning_rate": 4.228675703858186e-05,
|
| 616 |
+
"loss": 0.866,
|
| 617 |
+
"step": 7900
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 8.26,
|
| 621 |
+
"learning_rate": 4.2182481751824824e-05,
|
| 622 |
+
"loss": 0.8573,
|
| 623 |
+
"step": 8000
|
| 624 |
+
},
|
| 625 |
+
{
|
| 626 |
+
"epoch": 8.26,
|
| 627 |
+
"eval_loss": 0.17806729674339294,
|
| 628 |
+
"eval_runtime": 469.9178,
|
| 629 |
+
"eval_samples_per_second": 22.295,
|
| 630 |
+
"eval_steps_per_second": 2.788,
|
| 631 |
+
"eval_wer": 0.15872197073376146,
|
| 632 |
+
"step": 8000
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 8.36,
|
| 636 |
+
"learning_rate": 4.207820646506778e-05,
|
| 637 |
+
"loss": 0.8568,
|
| 638 |
+
"step": 8100
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 8.46,
|
| 642 |
+
"learning_rate": 4.197393117831074e-05,
|
| 643 |
+
"loss": 0.863,
|
| 644 |
+
"step": 8200
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 8.57,
|
| 648 |
+
"learning_rate": 4.1869655891553704e-05,
|
| 649 |
+
"loss": 0.859,
|
| 650 |
+
"step": 8300
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 8.67,
|
| 654 |
+
"learning_rate": 4.1765380604796666e-05,
|
| 655 |
+
"loss": 0.8374,
|
| 656 |
+
"step": 8400
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 8.77,
|
| 660 |
+
"learning_rate": 4.166110531803963e-05,
|
| 661 |
+
"loss": 0.8633,
|
| 662 |
+
"step": 8500
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 8.77,
|
| 666 |
+
"eval_loss": 0.18520139157772064,
|
| 667 |
+
"eval_runtime": 466.6826,
|
| 668 |
+
"eval_samples_per_second": 22.45,
|
| 669 |
+
"eval_steps_per_second": 2.807,
|
| 670 |
+
"eval_wer": 0.162838193494052,
|
| 671 |
+
"step": 8500
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 8.87,
|
| 675 |
+
"learning_rate": 4.155683003128259e-05,
|
| 676 |
+
"loss": 0.8656,
|
| 677 |
+
"step": 8600
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 8.98,
|
| 681 |
+
"learning_rate": 4.145255474452555e-05,
|
| 682 |
+
"loss": 0.8522,
|
| 683 |
+
"step": 8700
|
| 684 |
+
},
|
| 685 |
+
{
|
| 686 |
+
"epoch": 9.08,
|
| 687 |
+
"learning_rate": 4.1348279457768515e-05,
|
| 688 |
+
"loss": 0.8341,
|
| 689 |
+
"step": 8800
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 9.18,
|
| 693 |
+
"learning_rate": 4.124400417101148e-05,
|
| 694 |
+
"loss": 0.8484,
|
| 695 |
+
"step": 8900
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 9.29,
|
| 699 |
+
"learning_rate": 4.113972888425443e-05,
|
| 700 |
+
"loss": 0.8603,
|
| 701 |
+
"step": 9000
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 9.29,
|
| 705 |
+
"eval_loss": 0.1832893192768097,
|
| 706 |
+
"eval_runtime": 472.9551,
|
| 707 |
+
"eval_samples_per_second": 22.152,
|
| 708 |
+
"eval_steps_per_second": 2.77,
|
| 709 |
+
"eval_wer": 0.16013264554163595,
|
| 710 |
+
"step": 9000
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"epoch": 9.39,
|
| 714 |
+
"learning_rate": 4.1035453597497395e-05,
|
| 715 |
+
"loss": 0.8473,
|
| 716 |
+
"step": 9100
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 9.49,
|
| 720 |
+
"learning_rate": 4.093117831074036e-05,
|
| 721 |
+
"loss": 0.8488,
|
| 722 |
+
"step": 9200
|
| 723 |
+
},
|
| 724 |
+
{
|
| 725 |
+
"epoch": 9.6,
|
| 726 |
+
"learning_rate": 4.082690302398332e-05,
|
| 727 |
+
"loss": 0.8492,
|
| 728 |
+
"step": 9300
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 9.7,
|
| 732 |
+
"learning_rate": 4.072367049009385e-05,
|
| 733 |
+
"loss": 0.8364,
|
| 734 |
+
"step": 9400
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 9.8,
|
| 738 |
+
"learning_rate": 4.061939520333681e-05,
|
| 739 |
+
"loss": 0.8421,
|
| 740 |
+
"step": 9500
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 9.8,
|
| 744 |
+
"eval_loss": 0.17876620590686798,
|
| 745 |
+
"eval_runtime": 467.77,
|
| 746 |
+
"eval_samples_per_second": 22.398,
|
| 747 |
+
"eval_steps_per_second": 2.801,
|
| 748 |
+
"eval_wer": 0.15428992525529003,
|
| 749 |
+
"step": 9500
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 9.91,
|
| 753 |
+
"learning_rate": 4.051511991657977e-05,
|
| 754 |
+
"loss": 0.8468,
|
| 755 |
+
"step": 9600
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 10.01,
|
| 759 |
+
"learning_rate": 4.0410844629822735e-05,
|
| 760 |
+
"loss": 0.8396,
|
| 761 |
+
"step": 9700
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 10.11,
|
| 765 |
+
"learning_rate": 4.03065693430657e-05,
|
| 766 |
+
"loss": 0.8346,
|
| 767 |
+
"step": 9800
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 10.22,
|
| 771 |
+
"learning_rate": 4.020229405630866e-05,
|
| 772 |
+
"loss": 0.8411,
|
| 773 |
+
"step": 9900
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 10.32,
|
| 777 |
+
"learning_rate": 4.009801876955162e-05,
|
| 778 |
+
"loss": 0.8404,
|
| 779 |
+
"step": 10000
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 10.32,
|
| 783 |
+
"eval_loss": 0.184404194355011,
|
| 784 |
+
"eval_runtime": 465.0709,
|
| 785 |
+
"eval_samples_per_second": 22.528,
|
| 786 |
+
"eval_steps_per_second": 2.817,
|
| 787 |
+
"eval_wer": 0.15555321612801348,
|
| 788 |
+
"step": 10000
|
| 789 |
+
},
|
| 790 |
+
{
|
| 791 |
+
"epoch": 10.42,
|
| 792 |
+
"learning_rate": 3.9993743482794584e-05,
|
| 793 |
+
"loss": 0.8382,
|
| 794 |
+
"step": 10100
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 10.53,
|
| 798 |
+
"learning_rate": 3.988946819603754e-05,
|
| 799 |
+
"loss": 0.8454,
|
| 800 |
+
"step": 10200
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 10.63,
|
| 804 |
+
"learning_rate": 3.97851929092805e-05,
|
| 805 |
+
"loss": 0.8472,
|
| 806 |
+
"step": 10300
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 10.73,
|
| 810 |
+
"learning_rate": 3.9680917622523464e-05,
|
| 811 |
+
"loss": 0.8273,
|
| 812 |
+
"step": 10400
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"epoch": 10.84,
|
| 816 |
+
"learning_rate": 3.957664233576642e-05,
|
| 817 |
+
"loss": 0.8342,
|
| 818 |
+
"step": 10500
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 10.84,
|
| 822 |
+
"eval_loss": 0.17697852849960327,
|
| 823 |
+
"eval_runtime": 469.3266,
|
| 824 |
+
"eval_samples_per_second": 22.323,
|
| 825 |
+
"eval_steps_per_second": 2.791,
|
| 826 |
+
"eval_wer": 0.15379513633014003,
|
| 827 |
+
"step": 10500
|
| 828 |
+
},
|
| 829 |
+
{
|
| 830 |
+
"epoch": 10.94,
|
| 831 |
+
"learning_rate": 3.947236704900938e-05,
|
| 832 |
+
"loss": 0.8465,
|
| 833 |
+
"step": 10600
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 11.04,
|
| 837 |
+
"learning_rate": 3.9368091762252344e-05,
|
| 838 |
+
"loss": 0.8563,
|
| 839 |
+
"step": 10700
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 11.15,
|
| 843 |
+
"learning_rate": 3.9263816475495307e-05,
|
| 844 |
+
"loss": 0.823,
|
| 845 |
+
"step": 10800
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 11.25,
|
| 849 |
+
"learning_rate": 3.915954118873827e-05,
|
| 850 |
+
"loss": 0.8325,
|
| 851 |
+
"step": 10900
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 11.35,
|
| 855 |
+
"learning_rate": 3.905526590198123e-05,
|
| 856 |
+
"loss": 0.8161,
|
| 857 |
+
"step": 11000
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 11.35,
|
| 861 |
+
"eval_loss": 0.1820838451385498,
|
| 862 |
+
"eval_runtime": 468.1432,
|
| 863 |
+
"eval_samples_per_second": 22.38,
|
| 864 |
+
"eval_steps_per_second": 2.798,
|
| 865 |
+
"eval_wer": 0.1567112327613433,
|
| 866 |
+
"step": 11000
|
| 867 |
+
},
|
| 868 |
+
{
|
| 869 |
+
"epoch": 11.46,
|
| 870 |
+
"learning_rate": 3.895203336809176e-05,
|
| 871 |
+
"loss": 0.8227,
|
| 872 |
+
"step": 11100
|
| 873 |
+
},
|
| 874 |
+
{
|
| 875 |
+
"epoch": 11.56,
|
| 876 |
+
"learning_rate": 3.884775808133472e-05,
|
| 877 |
+
"loss": 0.8296,
|
| 878 |
+
"step": 11200
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 11.66,
|
| 882 |
+
"learning_rate": 3.8743482794577684e-05,
|
| 883 |
+
"loss": 0.8237,
|
| 884 |
+
"step": 11300
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 11.76,
|
| 888 |
+
"learning_rate": 3.8639207507820647e-05,
|
| 889 |
+
"loss": 0.8171,
|
| 890 |
+
"step": 11400
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 11.87,
|
| 894 |
+
"learning_rate": 3.853493222106361e-05,
|
| 895 |
+
"loss": 0.8371,
|
| 896 |
+
"step": 11500
|
| 897 |
+
},
|
| 898 |
+
{
|
| 899 |
+
"epoch": 11.87,
|
| 900 |
+
"eval_loss": 0.19089022278785706,
|
| 901 |
+
"eval_runtime": 465.3394,
|
| 902 |
+
"eval_samples_per_second": 22.515,
|
| 903 |
+
"eval_steps_per_second": 2.815,
|
| 904 |
+
"eval_wer": 0.16290135803768818,
|
| 905 |
+
"step": 11500
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 11.97,
|
| 909 |
+
"learning_rate": 3.843065693430657e-05,
|
| 910 |
+
"loss": 0.8331,
|
| 911 |
+
"step": 11600
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 12.07,
|
| 915 |
+
"learning_rate": 3.832638164754953e-05,
|
| 916 |
+
"loss": 0.8088,
|
| 917 |
+
"step": 11700
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 12.18,
|
| 921 |
+
"learning_rate": 3.8222106360792496e-05,
|
| 922 |
+
"loss": 0.8176,
|
| 923 |
+
"step": 11800
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 12.28,
|
| 927 |
+
"learning_rate": 3.811783107403545e-05,
|
| 928 |
+
"loss": 0.8034,
|
| 929 |
+
"step": 11900
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 12.38,
|
| 933 |
+
"learning_rate": 3.801355578727841e-05,
|
| 934 |
+
"loss": 0.8083,
|
| 935 |
+
"step": 12000
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 12.38,
|
| 939 |
+
"eval_loss": 0.17782799899578094,
|
| 940 |
+
"eval_runtime": 470.9072,
|
| 941 |
+
"eval_samples_per_second": 22.249,
|
| 942 |
+
"eval_steps_per_second": 2.782,
|
| 943 |
+
"eval_wer": 0.14979471523318244,
|
| 944 |
+
"step": 12000
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 12.49,
|
| 948 |
+
"learning_rate": 3.7909280500521376e-05,
|
| 949 |
+
"loss": 0.8128,
|
| 950 |
+
"step": 12100
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 12.59,
|
| 954 |
+
"learning_rate": 3.780500521376434e-05,
|
| 955 |
+
"loss": 0.8234,
|
| 956 |
+
"step": 12200
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 12.69,
|
| 960 |
+
"learning_rate": 3.77007299270073e-05,
|
| 961 |
+
"loss": 0.808,
|
| 962 |
+
"step": 12300
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 12.8,
|
| 966 |
+
"learning_rate": 3.759645464025026e-05,
|
| 967 |
+
"loss": 0.8225,
|
| 968 |
+
"step": 12400
|
| 969 |
+
},
|
| 970 |
+
{
|
| 971 |
+
"epoch": 12.9,
|
| 972 |
+
"learning_rate": 3.7492179353493225e-05,
|
| 973 |
+
"loss": 0.806,
|
| 974 |
+
"step": 12500
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 12.9,
|
| 978 |
+
"eval_loss": 0.18016664683818817,
|
| 979 |
+
"eval_runtime": 471.6501,
|
| 980 |
+
"eval_samples_per_second": 22.214,
|
| 981 |
+
"eval_steps_per_second": 2.777,
|
| 982 |
+
"eval_wer": 0.1547426044846826,
|
| 983 |
+
"step": 12500
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 13.0,
|
| 987 |
+
"learning_rate": 3.738790406673619e-05,
|
| 988 |
+
"loss": 0.8376,
|
| 989 |
+
"step": 12600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 13.11,
|
| 993 |
+
"learning_rate": 3.728362877997915e-05,
|
| 994 |
+
"loss": 0.8087,
|
| 995 |
+
"step": 12700
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 13.21,
|
| 999 |
+
"learning_rate": 3.7179353493222105e-05,
|
| 1000 |
+
"loss": 0.7941,
|
| 1001 |
+
"step": 12800
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 13.31,
|
| 1005 |
+
"learning_rate": 3.707507820646507e-05,
|
| 1006 |
+
"loss": 0.8064,
|
| 1007 |
+
"step": 12900
|
| 1008 |
+
},
|
| 1009 |
+
{
|
| 1010 |
+
"epoch": 13.42,
|
| 1011 |
+
"learning_rate": 3.69718456725756e-05,
|
| 1012 |
+
"loss": 0.8013,
|
| 1013 |
+
"step": 13000
|
| 1014 |
+
},
|
| 1015 |
+
{
|
| 1016 |
+
"epoch": 13.42,
|
| 1017 |
+
"eval_loss": 0.1859445571899414,
|
| 1018 |
+
"eval_runtime": 467.0913,
|
| 1019 |
+
"eval_samples_per_second": 22.43,
|
| 1020 |
+
"eval_steps_per_second": 2.805,
|
| 1021 |
+
"eval_wer": 0.15842720286345932,
|
| 1022 |
+
"step": 13000
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 13.52,
|
| 1026 |
+
"learning_rate": 3.6867570385818565e-05,
|
| 1027 |
+
"loss": 0.8004,
|
| 1028 |
+
"step": 13100
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 13.62,
|
| 1032 |
+
"learning_rate": 3.676329509906152e-05,
|
| 1033 |
+
"loss": 0.8201,
|
| 1034 |
+
"step": 13200
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 13.73,
|
| 1038 |
+
"learning_rate": 3.665901981230448e-05,
|
| 1039 |
+
"loss": 0.8141,
|
| 1040 |
+
"step": 13300
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 13.83,
|
| 1044 |
+
"learning_rate": 3.6554744525547445e-05,
|
| 1045 |
+
"loss": 0.8086,
|
| 1046 |
+
"step": 13400
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 13.93,
|
| 1050 |
+
"learning_rate": 3.645046923879041e-05,
|
| 1051 |
+
"loss": 0.7913,
|
| 1052 |
+
"step": 13500
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 13.93,
|
| 1056 |
+
"eval_loss": 0.18746201694011688,
|
| 1057 |
+
"eval_runtime": 466.4675,
|
| 1058 |
+
"eval_samples_per_second": 22.46,
|
| 1059 |
+
"eval_steps_per_second": 2.808,
|
| 1060 |
+
"eval_wer": 0.15165806927044953,
|
| 1061 |
+
"step": 13500
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 14.04,
|
| 1065 |
+
"learning_rate": 3.634619395203337e-05,
|
| 1066 |
+
"loss": 0.7865,
|
| 1067 |
+
"step": 13600
|
| 1068 |
+
},
|
| 1069 |
+
{
|
| 1070 |
+
"epoch": 14.14,
|
| 1071 |
+
"learning_rate": 3.624191866527633e-05,
|
| 1072 |
+
"loss": 0.7845,
|
| 1073 |
+
"step": 13700
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"epoch": 14.24,
|
| 1077 |
+
"learning_rate": 3.6137643378519294e-05,
|
| 1078 |
+
"loss": 0.8024,
|
| 1079 |
+
"step": 13800
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 14.34,
|
| 1083 |
+
"learning_rate": 3.6033368091762256e-05,
|
| 1084 |
+
"loss": 0.8109,
|
| 1085 |
+
"step": 13900
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 14.45,
|
| 1089 |
+
"learning_rate": 3.592909280500522e-05,
|
| 1090 |
+
"loss": 0.8063,
|
| 1091 |
+
"step": 14000
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 14.45,
|
| 1095 |
+
"eval_loss": 0.17992526292800903,
|
| 1096 |
+
"eval_runtime": 465.4116,
|
| 1097 |
+
"eval_samples_per_second": 22.511,
|
| 1098 |
+
"eval_steps_per_second": 2.815,
|
| 1099 |
+
"eval_wer": 0.15713232971891777,
|
| 1100 |
+
"step": 14000
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 14.55,
|
| 1104 |
+
"learning_rate": 3.582481751824818e-05,
|
| 1105 |
+
"loss": 0.8058,
|
| 1106 |
+
"step": 14100
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 14.65,
|
| 1110 |
+
"learning_rate": 3.5720542231491136e-05,
|
| 1111 |
+
"loss": 0.8093,
|
| 1112 |
+
"step": 14200
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"epoch": 14.76,
|
| 1116 |
+
"learning_rate": 3.56162669447341e-05,
|
| 1117 |
+
"loss": 0.7969,
|
| 1118 |
+
"step": 14300
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 14.86,
|
| 1122 |
+
"learning_rate": 3.551199165797706e-05,
|
| 1123 |
+
"loss": 0.8031,
|
| 1124 |
+
"step": 14400
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 14.96,
|
| 1128 |
+
"learning_rate": 3.540771637122002e-05,
|
| 1129 |
+
"loss": 0.7991,
|
| 1130 |
+
"step": 14500
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 14.96,
|
| 1134 |
+
"eval_loss": 0.17917674779891968,
|
| 1135 |
+
"eval_runtime": 479.1894,
|
| 1136 |
+
"eval_samples_per_second": 21.864,
|
| 1137 |
+
"eval_steps_per_second": 2.734,
|
| 1138 |
+
"eval_wer": 0.15384777344983683,
|
| 1139 |
+
"step": 14500
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 15.07,
|
| 1143 |
+
"learning_rate": 3.5303441084462985e-05,
|
| 1144 |
+
"loss": 0.7973,
|
| 1145 |
+
"step": 14600
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 15.17,
|
| 1149 |
+
"learning_rate": 3.519916579770595e-05,
|
| 1150 |
+
"loss": 0.7975,
|
| 1151 |
+
"step": 14700
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 15.27,
|
| 1155 |
+
"learning_rate": 3.509489051094891e-05,
|
| 1156 |
+
"loss": 0.7785,
|
| 1157 |
+
"step": 14800
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 15.38,
|
| 1161 |
+
"learning_rate": 3.499061522419187e-05,
|
| 1162 |
+
"loss": 0.8042,
|
| 1163 |
+
"step": 14900
|
| 1164 |
+
},
|
| 1165 |
+
{
|
| 1166 |
+
"epoch": 15.48,
|
| 1167 |
+
"learning_rate": 3.4886339937434834e-05,
|
| 1168 |
+
"loss": 0.7843,
|
| 1169 |
+
"step": 15000
|
| 1170 |
+
},
|
| 1171 |
+
{
|
| 1172 |
+
"epoch": 15.48,
|
| 1173 |
+
"eval_loss": 0.17534053325653076,
|
| 1174 |
+
"eval_runtime": 464.1652,
|
| 1175 |
+
"eval_samples_per_second": 22.572,
|
| 1176 |
+
"eval_steps_per_second": 2.822,
|
| 1177 |
+
"eval_wer": 0.14635224760501106,
|
| 1178 |
+
"step": 15000
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"epoch": 15.58,
|
| 1182 |
+
"learning_rate": 3.478206465067779e-05,
|
| 1183 |
+
"loss": 0.7885,
|
| 1184 |
+
"step": 15100
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 15.69,
|
| 1188 |
+
"learning_rate": 3.467778936392075e-05,
|
| 1189 |
+
"loss": 0.7886,
|
| 1190 |
+
"step": 15200
|
| 1191 |
+
},
|
| 1192 |
+
{
|
| 1193 |
+
"epoch": 15.79,
|
| 1194 |
+
"learning_rate": 3.4573514077163714e-05,
|
| 1195 |
+
"loss": 0.7995,
|
| 1196 |
+
"step": 15300
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 15.89,
|
| 1200 |
+
"learning_rate": 3.4469238790406676e-05,
|
| 1201 |
+
"loss": 0.7894,
|
| 1202 |
+
"step": 15400
|
| 1203 |
+
},
|
| 1204 |
+
{
|
| 1205 |
+
"epoch": 16.0,
|
| 1206 |
+
"learning_rate": 3.436496350364964e-05,
|
| 1207 |
+
"loss": 0.7905,
|
| 1208 |
+
"step": 15500
|
| 1209 |
+
},
|
| 1210 |
+
{
|
| 1211 |
+
"epoch": 16.0,
|
| 1212 |
+
"eval_loss": 0.1784074753522873,
|
| 1213 |
+
"eval_runtime": 467.7669,
|
| 1214 |
+
"eval_samples_per_second": 22.398,
|
| 1215 |
+
"eval_steps_per_second": 2.801,
|
| 1216 |
+
"eval_wer": 0.15082640277923992,
|
| 1217 |
+
"step": 15500
|
| 1218 |
+
},
|
| 1219 |
+
{
|
| 1220 |
+
"epoch": 16.1,
|
| 1221 |
+
"learning_rate": 3.42606882168926e-05,
|
| 1222 |
+
"loss": 0.7932,
|
| 1223 |
+
"step": 15600
|
| 1224 |
+
},
|
| 1225 |
+
{
|
| 1226 |
+
"epoch": 16.2,
|
| 1227 |
+
"learning_rate": 3.415641293013556e-05,
|
| 1228 |
+
"loss": 0.7954,
|
| 1229 |
+
"step": 15700
|
| 1230 |
+
},
|
| 1231 |
+
{
|
| 1232 |
+
"epoch": 16.31,
|
| 1233 |
+
"learning_rate": 3.4052137643378525e-05,
|
| 1234 |
+
"loss": 0.778,
|
| 1235 |
+
"step": 15800
|
| 1236 |
+
},
|
| 1237 |
+
{
|
| 1238 |
+
"epoch": 16.41,
|
| 1239 |
+
"learning_rate": 3.394786235662149e-05,
|
| 1240 |
+
"loss": 0.7963,
|
| 1241 |
+
"step": 15900
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 16.51,
|
| 1245 |
+
"learning_rate": 3.384358706986444e-05,
|
| 1246 |
+
"loss": 0.7808,
|
| 1247 |
+
"step": 16000
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 16.51,
|
| 1251 |
+
"eval_loss": 0.17709574103355408,
|
| 1252 |
+
"eval_runtime": 467.7784,
|
| 1253 |
+
"eval_samples_per_second": 22.397,
|
| 1254 |
+
"eval_steps_per_second": 2.8,
|
| 1255 |
+
"eval_wer": 0.148531424360459,
|
| 1256 |
+
"step": 16000
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 16.61,
|
| 1260 |
+
"learning_rate": 3.3739311783107405e-05,
|
| 1261 |
+
"loss": 0.7811,
|
| 1262 |
+
"step": 16100
|
| 1263 |
+
},
|
| 1264 |
+
{
|
| 1265 |
+
"epoch": 16.72,
|
| 1266 |
+
"learning_rate": 3.363503649635036e-05,
|
| 1267 |
+
"loss": 0.7732,
|
| 1268 |
+
"step": 16200
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 16.82,
|
| 1272 |
+
"learning_rate": 3.353076120959332e-05,
|
| 1273 |
+
"loss": 0.7809,
|
| 1274 |
+
"step": 16300
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"epoch": 16.92,
|
| 1278 |
+
"learning_rate": 3.3426485922836285e-05,
|
| 1279 |
+
"loss": 0.7819,
|
| 1280 |
+
"step": 16400
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 17.03,
|
| 1284 |
+
"learning_rate": 3.332221063607925e-05,
|
| 1285 |
+
"loss": 0.7743,
|
| 1286 |
+
"step": 16500
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 17.03,
|
| 1290 |
+
"eval_loss": 0.17948681116104126,
|
| 1291 |
+
"eval_runtime": 462.7998,
|
| 1292 |
+
"eval_samples_per_second": 22.638,
|
| 1293 |
+
"eval_steps_per_second": 2.831,
|
| 1294 |
+
"eval_wer": 0.14913148752500263,
|
| 1295 |
+
"step": 16500
|
| 1296 |
+
},
|
| 1297 |
+
{
|
| 1298 |
+
"epoch": 17.13,
|
| 1299 |
+
"learning_rate": 3.321793534932221e-05,
|
| 1300 |
+
"loss": 0.7669,
|
| 1301 |
+
"step": 16600
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 17.23,
|
| 1305 |
+
"learning_rate": 3.311366006256517e-05,
|
| 1306 |
+
"loss": 0.7713,
|
| 1307 |
+
"step": 16700
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 17.34,
|
| 1311 |
+
"learning_rate": 3.3009384775808134e-05,
|
| 1312 |
+
"loss": 0.7757,
|
| 1313 |
+
"step": 16800
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 17.44,
|
| 1317 |
+
"learning_rate": 3.29051094890511e-05,
|
| 1318 |
+
"loss": 0.7736,
|
| 1319 |
+
"step": 16900
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 17.54,
|
| 1323 |
+
"learning_rate": 3.280083420229406e-05,
|
| 1324 |
+
"loss": 0.7833,
|
| 1325 |
+
"step": 17000
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"epoch": 17.54,
|
| 1329 |
+
"eval_loss": 0.1721695214509964,
|
| 1330 |
+
"eval_runtime": 464.6724,
|
| 1331 |
+
"eval_samples_per_second": 22.547,
|
| 1332 |
+
"eval_steps_per_second": 2.819,
|
| 1333 |
+
"eval_wer": 0.14838404042530792,
|
| 1334 |
+
"step": 17000
|
| 1335 |
+
},
|
| 1336 |
+
{
|
| 1337 |
+
"epoch": 17.65,
|
| 1338 |
+
"learning_rate": 3.2696558915537014e-05,
|
| 1339 |
+
"loss": 0.7752,
|
| 1340 |
+
"step": 17100
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"epoch": 17.75,
|
| 1344 |
+
"learning_rate": 3.259228362877998e-05,
|
| 1345 |
+
"loss": 0.7772,
|
| 1346 |
+
"step": 17200
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 17.85,
|
| 1350 |
+
"learning_rate": 3.248800834202294e-05,
|
| 1351 |
+
"loss": 0.7805,
|
| 1352 |
+
"step": 17300
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 17.96,
|
| 1356 |
+
"learning_rate": 3.23837330552659e-05,
|
| 1357 |
+
"loss": 0.7744,
|
| 1358 |
+
"step": 17400
|
| 1359 |
+
},
|
| 1360 |
+
{
|
| 1361 |
+
"epoch": 18.06,
|
| 1362 |
+
"learning_rate": 3.2279457768508863e-05,
|
| 1363 |
+
"loss": 0.7763,
|
| 1364 |
+
"step": 17500
|
| 1365 |
+
},
|
| 1366 |
+
{
|
| 1367 |
+
"epoch": 18.06,
|
| 1368 |
+
"eval_loss": 0.1766623854637146,
|
| 1369 |
+
"eval_runtime": 465.5186,
|
| 1370 |
+
"eval_samples_per_second": 22.506,
|
| 1371 |
+
"eval_steps_per_second": 2.814,
|
| 1372 |
+
"eval_wer": 0.15183703547741867,
|
| 1373 |
+
"step": 17500
|
| 1374 |
+
},
|
| 1375 |
+
{
|
| 1376 |
+
"epoch": 18.16,
|
| 1377 |
+
"learning_rate": 3.2175182481751826e-05,
|
| 1378 |
+
"loss": 0.7705,
|
| 1379 |
+
"step": 17600
|
| 1380 |
+
},
|
| 1381 |
+
{
|
| 1382 |
+
"epoch": 18.27,
|
| 1383 |
+
"learning_rate": 3.207090719499479e-05,
|
| 1384 |
+
"loss": 0.7576,
|
| 1385 |
+
"step": 17700
|
| 1386 |
+
},
|
| 1387 |
+
{
|
| 1388 |
+
"epoch": 18.37,
|
| 1389 |
+
"learning_rate": 3.196663190823775e-05,
|
| 1390 |
+
"loss": 0.7605,
|
| 1391 |
+
"step": 17800
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 18.47,
|
| 1395 |
+
"learning_rate": 3.186235662148071e-05,
|
| 1396 |
+
"loss": 0.7672,
|
| 1397 |
+
"step": 17900
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 18.58,
|
| 1401 |
+
"learning_rate": 3.1758081334723675e-05,
|
| 1402 |
+
"loss": 0.7698,
|
| 1403 |
+
"step": 18000
|
| 1404 |
+
},
|
| 1405 |
+
{
|
| 1406 |
+
"epoch": 18.58,
|
| 1407 |
+
"eval_loss": 0.17199581861495972,
|
| 1408 |
+
"eval_runtime": 465.8946,
|
| 1409 |
+
"eval_samples_per_second": 22.488,
|
| 1410 |
+
"eval_steps_per_second": 2.812,
|
| 1411 |
+
"eval_wer": 0.14599431519107275,
|
| 1412 |
+
"step": 18000
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 18.68,
|
| 1416 |
+
"learning_rate": 3.1654848800834204e-05,
|
| 1417 |
+
"loss": 0.7677,
|
| 1418 |
+
"step": 18100
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 18.78,
|
| 1422 |
+
"learning_rate": 3.1550573514077166e-05,
|
| 1423 |
+
"loss": 0.7732,
|
| 1424 |
+
"step": 18200
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 18.89,
|
| 1428 |
+
"learning_rate": 3.144629822732013e-05,
|
| 1429 |
+
"loss": 0.7687,
|
| 1430 |
+
"step": 18300
|
| 1431 |
+
},
|
| 1432 |
+
{
|
| 1433 |
+
"epoch": 18.99,
|
| 1434 |
+
"learning_rate": 3.134202294056309e-05,
|
| 1435 |
+
"loss": 0.7687,
|
| 1436 |
+
"step": 18400
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 19.09,
|
| 1440 |
+
"learning_rate": 3.123879040667362e-05,
|
| 1441 |
+
"loss": 0.7571,
|
| 1442 |
+
"step": 18500
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 19.09,
|
| 1446 |
+
"eval_loss": 0.1734933853149414,
|
| 1447 |
+
"eval_runtime": 481.2248,
|
| 1448 |
+
"eval_samples_per_second": 21.772,
|
| 1449 |
+
"eval_steps_per_second": 2.722,
|
| 1450 |
+
"eval_wer": 0.14779450468470365,
|
| 1451 |
+
"step": 18500
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 19.2,
|
| 1455 |
+
"learning_rate": 3.113451511991658e-05,
|
| 1456 |
+
"loss": 0.7649,
|
| 1457 |
+
"step": 18600
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 19.3,
|
| 1461 |
+
"learning_rate": 3.1030239833159544e-05,
|
| 1462 |
+
"loss": 0.7691,
|
| 1463 |
+
"step": 18700
|
| 1464 |
+
},
|
| 1465 |
+
{
|
| 1466 |
+
"epoch": 19.4,
|
| 1467 |
+
"learning_rate": 3.0925964546402506e-05,
|
| 1468 |
+
"loss": 0.7709,
|
| 1469 |
+
"step": 18800
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 19.5,
|
| 1473 |
+
"learning_rate": 3.082168925964546e-05,
|
| 1474 |
+
"loss": 0.7758,
|
| 1475 |
+
"step": 18900
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"epoch": 19.61,
|
| 1479 |
+
"learning_rate": 3.0717413972888424e-05,
|
| 1480 |
+
"loss": 0.7673,
|
| 1481 |
+
"step": 19000
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"epoch": 19.61,
|
| 1485 |
+
"eval_loss": 0.18169456720352173,
|
| 1486 |
+
"eval_runtime": 475.7022,
|
| 1487 |
+
"eval_samples_per_second": 22.024,
|
| 1488 |
+
"eval_steps_per_second": 2.754,
|
| 1489 |
+
"eval_wer": 0.15106853352984526,
|
| 1490 |
+
"step": 19000
|
| 1491 |
+
},
|
| 1492 |
+
{
|
| 1493 |
+
"epoch": 19.71,
|
| 1494 |
+
"learning_rate": 3.0613138686131386e-05,
|
| 1495 |
+
"loss": 0.7466,
|
| 1496 |
+
"step": 19100
|
| 1497 |
+
},
|
| 1498 |
+
{
|
| 1499 |
+
"epoch": 19.81,
|
| 1500 |
+
"learning_rate": 3.0508863399374348e-05,
|
| 1501 |
+
"loss": 0.7556,
|
| 1502 |
+
"step": 19200
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 19.92,
|
| 1506 |
+
"learning_rate": 3.040458811261731e-05,
|
| 1507 |
+
"loss": 0.7424,
|
| 1508 |
+
"step": 19300
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 20.02,
|
| 1512 |
+
"learning_rate": 3.0300312825860273e-05,
|
| 1513 |
+
"loss": 0.7455,
|
| 1514 |
+
"step": 19400
|
| 1515 |
+
},
|
| 1516 |
+
{
|
| 1517 |
+
"epoch": 20.12,
|
| 1518 |
+
"learning_rate": 3.0196037539103235e-05,
|
| 1519 |
+
"loss": 0.7415,
|
| 1520 |
+
"step": 19500
|
| 1521 |
+
},
|
| 1522 |
+
{
|
| 1523 |
+
"epoch": 20.12,
|
| 1524 |
+
"eval_loss": 0.17630091309547424,
|
| 1525 |
+
"eval_runtime": 467.192,
|
| 1526 |
+
"eval_samples_per_second": 22.425,
|
| 1527 |
+
"eval_steps_per_second": 2.804,
|
| 1528 |
+
"eval_wer": 0.14813138225076325,
|
| 1529 |
+
"step": 19500
|
| 1530 |
+
},
|
| 1531 |
+
{
|
| 1532 |
+
"epoch": 20.23,
|
| 1533 |
+
"learning_rate": 3.0091762252346194e-05,
|
| 1534 |
+
"loss": 0.7479,
|
| 1535 |
+
"step": 19600
|
| 1536 |
+
},
|
| 1537 |
+
{
|
| 1538 |
+
"epoch": 20.33,
|
| 1539 |
+
"learning_rate": 2.9987486965589156e-05,
|
| 1540 |
+
"loss": 0.7506,
|
| 1541 |
+
"step": 19700
|
| 1542 |
+
},
|
| 1543 |
+
{
|
| 1544 |
+
"epoch": 20.43,
|
| 1545 |
+
"learning_rate": 2.9883211678832118e-05,
|
| 1546 |
+
"loss": 0.7523,
|
| 1547 |
+
"step": 19800
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 20.54,
|
| 1551 |
+
"learning_rate": 2.977893639207508e-05,
|
| 1552 |
+
"loss": 0.7646,
|
| 1553 |
+
"step": 19900
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 20.64,
|
| 1557 |
+
"learning_rate": 2.9674661105318043e-05,
|
| 1558 |
+
"loss": 0.751,
|
| 1559 |
+
"step": 20000
|
| 1560 |
+
},
|
| 1561 |
+
{
|
| 1562 |
+
"epoch": 20.64,
|
| 1563 |
+
"eval_loss": 0.1741507649421692,
|
| 1564 |
+
"eval_runtime": 472.297,
|
| 1565 |
+
"eval_samples_per_second": 22.183,
|
| 1566 |
+
"eval_steps_per_second": 2.774,
|
| 1567 |
+
"eval_wer": 0.1483629855774292,
|
| 1568 |
+
"step": 20000
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 20.74,
|
| 1572 |
+
"learning_rate": 2.9570385818561e-05,
|
| 1573 |
+
"loss": 0.7533,
|
| 1574 |
+
"step": 20100
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 20.85,
|
| 1578 |
+
"learning_rate": 2.9466110531803964e-05,
|
| 1579 |
+
"loss": 0.7505,
|
| 1580 |
+
"step": 20200
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 20.95,
|
| 1584 |
+
"learning_rate": 2.9361835245046926e-05,
|
| 1585 |
+
"loss": 0.7481,
|
| 1586 |
+
"step": 20300
|
| 1587 |
+
},
|
| 1588 |
+
{
|
| 1589 |
+
"epoch": 21.05,
|
| 1590 |
+
"learning_rate": 2.925755995828989e-05,
|
| 1591 |
+
"loss": 0.7741,
|
| 1592 |
+
"step": 20400
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 21.16,
|
| 1596 |
+
"learning_rate": 2.9153284671532847e-05,
|
| 1597 |
+
"loss": 0.7563,
|
| 1598 |
+
"step": 20500
|
| 1599 |
+
},
|
| 1600 |
+
{
|
| 1601 |
+
"epoch": 21.16,
|
| 1602 |
+
"eval_loss": 0.18097689747810364,
|
| 1603 |
+
"eval_runtime": 466.3942,
|
| 1604 |
+
"eval_samples_per_second": 22.464,
|
| 1605 |
+
"eval_steps_per_second": 2.809,
|
| 1606 |
+
"eval_wer": 0.16113275081587536,
|
| 1607 |
+
"step": 20500
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 21.26,
|
| 1611 |
+
"learning_rate": 2.904900938477581e-05,
|
| 1612 |
+
"loss": 0.755,
|
| 1613 |
+
"step": 20600
|
| 1614 |
+
},
|
| 1615 |
+
{
|
| 1616 |
+
"epoch": 21.36,
|
| 1617 |
+
"learning_rate": 2.8944734098018772e-05,
|
| 1618 |
+
"loss": 0.7452,
|
| 1619 |
+
"step": 20700
|
| 1620 |
+
},
|
| 1621 |
+
{
|
| 1622 |
+
"epoch": 21.47,
|
| 1623 |
+
"learning_rate": 2.8840458811261734e-05,
|
| 1624 |
+
"loss": 0.7559,
|
| 1625 |
+
"step": 20800
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 21.57,
|
| 1629 |
+
"learning_rate": 2.8736183524504696e-05,
|
| 1630 |
+
"loss": 0.7326,
|
| 1631 |
+
"step": 20900
|
| 1632 |
+
},
|
| 1633 |
+
{
|
| 1634 |
+
"epoch": 21.67,
|
| 1635 |
+
"learning_rate": 2.8631908237747655e-05,
|
| 1636 |
+
"loss": 0.7423,
|
| 1637 |
+
"step": 21000
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 21.67,
|
| 1641 |
+
"eval_loss": 0.1817416250705719,
|
| 1642 |
+
"eval_runtime": 467.2447,
|
| 1643 |
+
"eval_samples_per_second": 22.423,
|
| 1644 |
+
"eval_steps_per_second": 2.804,
|
| 1645 |
+
"eval_wer": 0.15571112748710392,
|
| 1646 |
+
"step": 21000
|
| 1647 |
+
},
|
| 1648 |
+
{
|
| 1649 |
+
"epoch": 21.77,
|
| 1650 |
+
"learning_rate": 2.8527632950990617e-05,
|
| 1651 |
+
"loss": 0.744,
|
| 1652 |
+
"step": 21100
|
| 1653 |
+
},
|
| 1654 |
+
{
|
| 1655 |
+
"epoch": 21.88,
|
| 1656 |
+
"learning_rate": 2.842335766423358e-05,
|
| 1657 |
+
"loss": 0.7539,
|
| 1658 |
+
"step": 21200
|
| 1659 |
+
},
|
| 1660 |
+
{
|
| 1661 |
+
"epoch": 21.98,
|
| 1662 |
+
"learning_rate": 2.8319082377476542e-05,
|
| 1663 |
+
"loss": 0.7502,
|
| 1664 |
+
"step": 21300
|
| 1665 |
+
},
|
| 1666 |
+
{
|
| 1667 |
+
"epoch": 22.08,
|
| 1668 |
+
"learning_rate": 2.8214807090719504e-05,
|
| 1669 |
+
"loss": 0.7374,
|
| 1670 |
+
"step": 21400
|
| 1671 |
+
},
|
| 1672 |
+
{
|
| 1673 |
+
"epoch": 22.19,
|
| 1674 |
+
"learning_rate": 2.8110531803962463e-05,
|
| 1675 |
+
"loss": 0.7242,
|
| 1676 |
+
"step": 21500
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"epoch": 22.19,
|
| 1680 |
+
"eval_loss": 0.1689848154783249,
|
| 1681 |
+
"eval_runtime": 467.3401,
|
| 1682 |
+
"eval_samples_per_second": 22.418,
|
| 1683 |
+
"eval_steps_per_second": 2.803,
|
| 1684 |
+
"eval_wer": 0.14457311295925887,
|
| 1685 |
+
"step": 21500
|
| 1686 |
+
},
|
| 1687 |
+
{
|
| 1688 |
+
"epoch": 22.29,
|
| 1689 |
+
"learning_rate": 2.8006256517205425e-05,
|
| 1690 |
+
"loss": 0.7461,
|
| 1691 |
+
"step": 21600
|
| 1692 |
+
},
|
| 1693 |
+
{
|
| 1694 |
+
"epoch": 22.39,
|
| 1695 |
+
"learning_rate": 2.7901981230448388e-05,
|
| 1696 |
+
"loss": 0.7269,
|
| 1697 |
+
"step": 21700
|
| 1698 |
+
},
|
| 1699 |
+
{
|
| 1700 |
+
"epoch": 22.5,
|
| 1701 |
+
"learning_rate": 2.779770594369135e-05,
|
| 1702 |
+
"loss": 0.744,
|
| 1703 |
+
"step": 21800
|
| 1704 |
+
},
|
| 1705 |
+
{
|
| 1706 |
+
"epoch": 22.6,
|
| 1707 |
+
"learning_rate": 2.7693430656934312e-05,
|
| 1708 |
+
"loss": 0.7437,
|
| 1709 |
+
"step": 21900
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 22.7,
|
| 1713 |
+
"learning_rate": 2.759019812304484e-05,
|
| 1714 |
+
"loss": 0.7251,
|
| 1715 |
+
"step": 22000
|
| 1716 |
+
},
|
| 1717 |
+
{
|
| 1718 |
+
"epoch": 22.7,
|
| 1719 |
+
"eval_loss": 0.16843970119953156,
|
| 1720 |
+
"eval_runtime": 466.906,
|
| 1721 |
+
"eval_samples_per_second": 22.439,
|
| 1722 |
+
"eval_steps_per_second": 2.806,
|
| 1723 |
+
"eval_wer": 0.14457311295925887,
|
| 1724 |
+
"step": 22000
|
| 1725 |
+
},
|
| 1726 |
+
{
|
| 1727 |
+
"epoch": 22.81,
|
| 1728 |
+
"learning_rate": 2.7485922836287803e-05,
|
| 1729 |
+
"loss": 0.7233,
|
| 1730 |
+
"step": 22100
|
| 1731 |
+
},
|
| 1732 |
+
{
|
| 1733 |
+
"epoch": 22.91,
|
| 1734 |
+
"learning_rate": 2.7381647549530765e-05,
|
| 1735 |
+
"loss": 0.7169,
|
| 1736 |
+
"step": 22200
|
| 1737 |
+
},
|
| 1738 |
+
{
|
| 1739 |
+
"epoch": 23.01,
|
| 1740 |
+
"learning_rate": 2.7277372262773728e-05,
|
| 1741 |
+
"loss": 0.7374,
|
| 1742 |
+
"step": 22300
|
| 1743 |
+
},
|
| 1744 |
+
{
|
| 1745 |
+
"epoch": 23.12,
|
| 1746 |
+
"learning_rate": 2.7173096976016687e-05,
|
| 1747 |
+
"loss": 0.7289,
|
| 1748 |
+
"step": 22400
|
| 1749 |
+
},
|
| 1750 |
+
{
|
| 1751 |
+
"epoch": 23.22,
|
| 1752 |
+
"learning_rate": 2.706882168925965e-05,
|
| 1753 |
+
"loss": 0.7302,
|
| 1754 |
+
"step": 22500
|
| 1755 |
+
},
|
| 1756 |
+
{
|
| 1757 |
+
"epoch": 23.22,
|
| 1758 |
+
"eval_loss": 0.17354781925678253,
|
| 1759 |
+
"eval_runtime": 469.7694,
|
| 1760 |
+
"eval_samples_per_second": 22.302,
|
| 1761 |
+
"eval_steps_per_second": 2.789,
|
| 1762 |
+
"eval_wer": 0.14295188967259712,
|
| 1763 |
+
"step": 22500
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 23.32,
|
| 1767 |
+
"learning_rate": 2.696454640250261e-05,
|
| 1768 |
+
"loss": 0.7307,
|
| 1769 |
+
"step": 22600
|
| 1770 |
+
},
|
| 1771 |
+
{
|
| 1772 |
+
"epoch": 23.43,
|
| 1773 |
+
"learning_rate": 2.6860271115745573e-05,
|
| 1774 |
+
"loss": 0.7343,
|
| 1775 |
+
"step": 22700
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 23.53,
|
| 1779 |
+
"learning_rate": 2.6755995828988532e-05,
|
| 1780 |
+
"loss": 0.7346,
|
| 1781 |
+
"step": 22800
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 23.63,
|
| 1785 |
+
"learning_rate": 2.665172054223149e-05,
|
| 1786 |
+
"loss": 0.7349,
|
| 1787 |
+
"step": 22900
|
| 1788 |
+
},
|
| 1789 |
+
{
|
| 1790 |
+
"epoch": 23.74,
|
| 1791 |
+
"learning_rate": 2.654744525547445e-05,
|
| 1792 |
+
"loss": 0.733,
|
| 1793 |
+
"step": 23000
|
| 1794 |
+
},
|
| 1795 |
+
{
|
| 1796 |
+
"epoch": 23.74,
|
| 1797 |
+
"eval_loss": 0.17200088500976562,
|
| 1798 |
+
"eval_runtime": 479.405,
|
| 1799 |
+
"eval_samples_per_second": 21.854,
|
| 1800 |
+
"eval_steps_per_second": 2.733,
|
| 1801 |
+
"eval_wer": 0.14537319717865038,
|
| 1802 |
+
"step": 23000
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"epoch": 23.84,
|
| 1806 |
+
"learning_rate": 2.6443169968717412e-05,
|
| 1807 |
+
"loss": 0.7285,
|
| 1808 |
+
"step": 23100
|
| 1809 |
+
},
|
| 1810 |
+
{
|
| 1811 |
+
"epoch": 23.94,
|
| 1812 |
+
"learning_rate": 2.6338894681960374e-05,
|
| 1813 |
+
"loss": 0.7238,
|
| 1814 |
+
"step": 23200
|
| 1815 |
+
},
|
| 1816 |
+
{
|
| 1817 |
+
"epoch": 24.05,
|
| 1818 |
+
"learning_rate": 2.6234619395203337e-05,
|
| 1819 |
+
"loss": 0.7355,
|
| 1820 |
+
"step": 23300
|
| 1821 |
+
},
|
| 1822 |
+
{
|
| 1823 |
+
"epoch": 24.15,
|
| 1824 |
+
"learning_rate": 2.61303441084463e-05,
|
| 1825 |
+
"loss": 0.725,
|
| 1826 |
+
"step": 23400
|
| 1827 |
+
},
|
| 1828 |
+
{
|
| 1829 |
+
"epoch": 24.25,
|
| 1830 |
+
"learning_rate": 2.6026068821689258e-05,
|
| 1831 |
+
"loss": 0.7128,
|
| 1832 |
+
"step": 23500
|
| 1833 |
+
},
|
| 1834 |
+
{
|
| 1835 |
+
"epoch": 24.25,
|
| 1836 |
+
"eval_loss": 0.16681526601314545,
|
| 1837 |
+
"eval_runtime": 465.5348,
|
| 1838 |
+
"eval_samples_per_second": 22.505,
|
| 1839 |
+
"eval_steps_per_second": 2.814,
|
| 1840 |
+
"eval_wer": 0.1382987682913991,
|
| 1841 |
+
"step": 23500
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 24.35,
|
| 1845 |
+
"learning_rate": 2.592179353493222e-05,
|
| 1846 |
+
"loss": 0.7104,
|
| 1847 |
+
"step": 23600
|
| 1848 |
+
},
|
| 1849 |
+
{
|
| 1850 |
+
"epoch": 24.46,
|
| 1851 |
+
"learning_rate": 2.5817518248175182e-05,
|
| 1852 |
+
"loss": 0.7139,
|
| 1853 |
+
"step": 23700
|
| 1854 |
+
},
|
| 1855 |
+
{
|
| 1856 |
+
"epoch": 24.56,
|
| 1857 |
+
"learning_rate": 2.5713242961418145e-05,
|
| 1858 |
+
"loss": 0.7086,
|
| 1859 |
+
"step": 23800
|
| 1860 |
+
},
|
| 1861 |
+
{
|
| 1862 |
+
"epoch": 24.66,
|
| 1863 |
+
"learning_rate": 2.5608967674661103e-05,
|
| 1864 |
+
"loss": 0.7124,
|
| 1865 |
+
"step": 23900
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 24.77,
|
| 1869 |
+
"learning_rate": 2.5504692387904066e-05,
|
| 1870 |
+
"loss": 0.7184,
|
| 1871 |
+
"step": 24000
|
| 1872 |
+
},
|
| 1873 |
+
{
|
| 1874 |
+
"epoch": 24.77,
|
| 1875 |
+
"eval_loss": 0.16353647410869598,
|
| 1876 |
+
"eval_runtime": 467.6672,
|
| 1877 |
+
"eval_samples_per_second": 22.403,
|
| 1878 |
+
"eval_steps_per_second": 2.801,
|
| 1879 |
+
"eval_wer": 0.1377408148226129,
|
| 1880 |
+
"step": 24000
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 24.87,
|
| 1884 |
+
"learning_rate": 2.5400417101147028e-05,
|
| 1885 |
+
"loss": 0.7174,
|
| 1886 |
+
"step": 24100
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"epoch": 24.97,
|
| 1890 |
+
"learning_rate": 2.529614181438999e-05,
|
| 1891 |
+
"loss": 0.7164,
|
| 1892 |
+
"step": 24200
|
| 1893 |
+
},
|
| 1894 |
+
{
|
| 1895 |
+
"epoch": 25.08,
|
| 1896 |
+
"learning_rate": 2.5191866527632953e-05,
|
| 1897 |
+
"loss": 0.6949,
|
| 1898 |
+
"step": 24300
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"epoch": 25.18,
|
| 1902 |
+
"learning_rate": 2.508759124087591e-05,
|
| 1903 |
+
"loss": 0.7121,
|
| 1904 |
+
"step": 24400
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 25.28,
|
| 1908 |
+
"learning_rate": 2.4983315954118874e-05,
|
| 1909 |
+
"loss": 0.7015,
|
| 1910 |
+
"step": 24500
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"epoch": 25.28,
|
| 1914 |
+
"eval_loss": 0.1646386831998825,
|
| 1915 |
+
"eval_runtime": 470.6383,
|
| 1916 |
+
"eval_samples_per_second": 22.261,
|
| 1917 |
+
"eval_steps_per_second": 2.783,
|
| 1918 |
+
"eval_wer": 0.1389409411517002,
|
| 1919 |
+
"step": 24500
|
| 1920 |
+
},
|
| 1921 |
+
{
|
| 1922 |
+
"epoch": 25.39,
|
| 1923 |
+
"learning_rate": 2.4879040667361836e-05,
|
| 1924 |
+
"loss": 0.7104,
|
| 1925 |
+
"step": 24600
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 25.49,
|
| 1929 |
+
"learning_rate": 2.4774765380604798e-05,
|
| 1930 |
+
"loss": 0.7083,
|
| 1931 |
+
"step": 24700
|
| 1932 |
+
},
|
| 1933 |
+
{
|
| 1934 |
+
"epoch": 25.59,
|
| 1935 |
+
"learning_rate": 2.467049009384776e-05,
|
| 1936 |
+
"loss": 0.7085,
|
| 1937 |
+
"step": 24800
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 25.7,
|
| 1941 |
+
"learning_rate": 2.456621480709072e-05,
|
| 1942 |
+
"loss": 0.7215,
|
| 1943 |
+
"step": 24900
|
| 1944 |
+
},
|
| 1945 |
+
{
|
| 1946 |
+
"epoch": 25.8,
|
| 1947 |
+
"learning_rate": 2.446193952033368e-05,
|
| 1948 |
+
"loss": 0.7198,
|
| 1949 |
+
"step": 25000
|
| 1950 |
+
},
|
| 1951 |
+
{
|
| 1952 |
+
"epoch": 25.8,
|
| 1953 |
+
"eval_loss": 0.17749331891536713,
|
| 1954 |
+
"eval_runtime": 481.0207,
|
| 1955 |
+
"eval_samples_per_second": 21.781,
|
| 1956 |
+
"eval_steps_per_second": 2.723,
|
| 1957 |
+
"eval_wer": 0.14620486366986,
|
| 1958 |
+
"step": 25000
|
| 1959 |
+
},
|
| 1960 |
+
{
|
| 1961 |
+
"epoch": 25.9,
|
| 1962 |
+
"learning_rate": 2.4357664233576644e-05,
|
| 1963 |
+
"loss": 0.7108,
|
| 1964 |
+
"step": 25100
|
| 1965 |
+
},
|
| 1966 |
+
{
|
| 1967 |
+
"epoch": 26.01,
|
| 1968 |
+
"learning_rate": 2.4253388946819606e-05,
|
| 1969 |
+
"loss": 0.7197,
|
| 1970 |
+
"step": 25200
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"epoch": 26.11,
|
| 1974 |
+
"learning_rate": 2.414911366006257e-05,
|
| 1975 |
+
"loss": 0.6984,
|
| 1976 |
+
"step": 25300
|
| 1977 |
+
},
|
| 1978 |
+
{
|
| 1979 |
+
"epoch": 26.21,
|
| 1980 |
+
"learning_rate": 2.4044838373305527e-05,
|
| 1981 |
+
"loss": 0.6976,
|
| 1982 |
+
"step": 25400
|
| 1983 |
+
},
|
| 1984 |
+
{
|
| 1985 |
+
"epoch": 26.32,
|
| 1986 |
+
"learning_rate": 2.394056308654849e-05,
|
| 1987 |
+
"loss": 0.7178,
|
| 1988 |
+
"step": 25500
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 26.32,
|
| 1992 |
+
"eval_loss": 0.17049279808998108,
|
| 1993 |
+
"eval_runtime": 463.0645,
|
| 1994 |
+
"eval_samples_per_second": 22.625,
|
| 1995 |
+
"eval_steps_per_second": 2.829,
|
| 1996 |
+
"eval_wer": 0.1418780924307822,
|
| 1997 |
+
"step": 25500
|
| 1998 |
+
},
|
| 1999 |
+
{
|
| 2000 |
+
"epoch": 26.42,
|
| 2001 |
+
"learning_rate": 2.3836287799791452e-05,
|
| 2002 |
+
"loss": 0.7164,
|
| 2003 |
+
"step": 25600
|
| 2004 |
+
},
|
| 2005 |
+
{
|
| 2006 |
+
"epoch": 26.52,
|
| 2007 |
+
"learning_rate": 2.3732012513034414e-05,
|
| 2008 |
+
"loss": 0.7322,
|
| 2009 |
+
"step": 25700
|
| 2010 |
+
},
|
| 2011 |
+
{
|
| 2012 |
+
"epoch": 26.63,
|
| 2013 |
+
"learning_rate": 2.3627737226277373e-05,
|
| 2014 |
+
"loss": 0.7158,
|
| 2015 |
+
"step": 25800
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 26.73,
|
| 2019 |
+
"learning_rate": 2.3523461939520335e-05,
|
| 2020 |
+
"loss": 0.7146,
|
| 2021 |
+
"step": 25900
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 26.83,
|
| 2025 |
+
"learning_rate": 2.3419186652763297e-05,
|
| 2026 |
+
"loss": 0.7199,
|
| 2027 |
+
"step": 26000
|
| 2028 |
+
},
|
| 2029 |
+
{
|
| 2030 |
+
"epoch": 26.83,
|
| 2031 |
+
"eval_loss": 0.16493017971515656,
|
| 2032 |
+
"eval_runtime": 466.7445,
|
| 2033 |
+
"eval_samples_per_second": 22.447,
|
| 2034 |
+
"eval_steps_per_second": 2.807,
|
| 2035 |
+
"eval_wer": 0.14164648910411623,
|
| 2036 |
+
"step": 26000
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 26.93,
|
| 2040 |
+
"learning_rate": 2.331595411887383e-05,
|
| 2041 |
+
"loss": 0.7001,
|
| 2042 |
+
"step": 26100
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 27.04,
|
| 2046 |
+
"learning_rate": 2.321167883211679e-05,
|
| 2047 |
+
"loss": 0.6918,
|
| 2048 |
+
"step": 26200
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 27.14,
|
| 2052 |
+
"learning_rate": 2.310740354535975e-05,
|
| 2053 |
+
"loss": 0.6864,
|
| 2054 |
+
"step": 26300
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 27.24,
|
| 2058 |
+
"learning_rate": 2.3003128258602713e-05,
|
| 2059 |
+
"loss": 0.7008,
|
| 2060 |
+
"step": 26400
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 27.35,
|
| 2064 |
+
"learning_rate": 2.2898852971845675e-05,
|
| 2065 |
+
"loss": 0.6981,
|
| 2066 |
+
"step": 26500
|
| 2067 |
+
},
|
| 2068 |
+
{
|
| 2069 |
+
"epoch": 27.35,
|
| 2070 |
+
"eval_loss": 0.1723548024892807,
|
| 2071 |
+
"eval_runtime": 467.3195,
|
| 2072 |
+
"eval_samples_per_second": 22.419,
|
| 2073 |
+
"eval_steps_per_second": 2.803,
|
| 2074 |
+
"eval_wer": 0.14182545531108537,
|
| 2075 |
+
"step": 26500
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 27.45,
|
| 2079 |
+
"learning_rate": 2.2794577685088637e-05,
|
| 2080 |
+
"loss": 0.7041,
|
| 2081 |
+
"step": 26600
|
| 2082 |
+
},
|
| 2083 |
+
{
|
| 2084 |
+
"epoch": 27.55,
|
| 2085 |
+
"learning_rate": 2.2690302398331596e-05,
|
| 2086 |
+
"loss": 0.6864,
|
| 2087 |
+
"step": 26700
|
| 2088 |
+
},
|
| 2089 |
+
{
|
| 2090 |
+
"epoch": 27.66,
|
| 2091 |
+
"learning_rate": 2.258602711157456e-05,
|
| 2092 |
+
"loss": 0.6989,
|
| 2093 |
+
"step": 26800
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 27.76,
|
| 2097 |
+
"learning_rate": 2.2481751824817517e-05,
|
| 2098 |
+
"loss": 0.687,
|
| 2099 |
+
"step": 26900
|
| 2100 |
+
},
|
| 2101 |
+
{
|
| 2102 |
+
"epoch": 27.86,
|
| 2103 |
+
"learning_rate": 2.237747653806048e-05,
|
| 2104 |
+
"loss": 0.6886,
|
| 2105 |
+
"step": 27000
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 27.86,
|
| 2109 |
+
"eval_loss": 0.16325753927230835,
|
| 2110 |
+
"eval_runtime": 468.8775,
|
| 2111 |
+
"eval_samples_per_second": 22.345,
|
| 2112 |
+
"eval_steps_per_second": 2.794,
|
| 2113 |
+
"eval_wer": 0.13823560374776292,
|
| 2114 |
+
"step": 27000
|
| 2115 |
+
},
|
| 2116 |
+
{
|
| 2117 |
+
"epoch": 27.97,
|
| 2118 |
+
"learning_rate": 2.2273201251303442e-05,
|
| 2119 |
+
"loss": 0.7044,
|
| 2120 |
+
"step": 27100
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 28.07,
|
| 2124 |
+
"learning_rate": 2.2168925964546404e-05,
|
| 2125 |
+
"loss": 0.6713,
|
| 2126 |
+
"step": 27200
|
| 2127 |
+
},
|
| 2128 |
+
{
|
| 2129 |
+
"epoch": 28.17,
|
| 2130 |
+
"learning_rate": 2.2064650677789363e-05,
|
| 2131 |
+
"loss": 0.6889,
|
| 2132 |
+
"step": 27300
|
| 2133 |
+
},
|
| 2134 |
+
{
|
| 2135 |
+
"epoch": 28.28,
|
| 2136 |
+
"learning_rate": 2.1960375391032325e-05,
|
| 2137 |
+
"loss": 0.6769,
|
| 2138 |
+
"step": 27400
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"epoch": 28.38,
|
| 2142 |
+
"learning_rate": 2.185714285714286e-05,
|
| 2143 |
+
"loss": 0.6922,
|
| 2144 |
+
"step": 27500
|
| 2145 |
+
},
|
| 2146 |
+
{
|
| 2147 |
+
"epoch": 28.38,
|
| 2148 |
+
"eval_loss": 0.16981077194213867,
|
| 2149 |
+
"eval_runtime": 479.6579,
|
| 2150 |
+
"eval_samples_per_second": 21.843,
|
| 2151 |
+
"eval_steps_per_second": 2.731,
|
| 2152 |
+
"eval_wer": 0.14196231182229707,
|
| 2153 |
+
"step": 27500
|
| 2154 |
+
},
|
| 2155 |
+
{
|
| 2156 |
+
"epoch": 28.48,
|
| 2157 |
+
"learning_rate": 2.175286757038582e-05,
|
| 2158 |
+
"loss": 0.6962,
|
| 2159 |
+
"step": 27600
|
| 2160 |
+
},
|
| 2161 |
+
{
|
| 2162 |
+
"epoch": 28.59,
|
| 2163 |
+
"learning_rate": 2.164859228362878e-05,
|
| 2164 |
+
"loss": 0.6907,
|
| 2165 |
+
"step": 27700
|
| 2166 |
+
},
|
| 2167 |
+
{
|
| 2168 |
+
"epoch": 28.69,
|
| 2169 |
+
"learning_rate": 2.154431699687174e-05,
|
| 2170 |
+
"loss": 0.6967,
|
| 2171 |
+
"step": 27800
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 28.79,
|
| 2175 |
+
"learning_rate": 2.1440041710114703e-05,
|
| 2176 |
+
"loss": 0.6699,
|
| 2177 |
+
"step": 27900
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 28.9,
|
| 2181 |
+
"learning_rate": 2.1335766423357665e-05,
|
| 2182 |
+
"loss": 0.6833,
|
| 2183 |
+
"step": 28000
|
| 2184 |
+
},
|
| 2185 |
+
{
|
| 2186 |
+
"epoch": 28.9,
|
| 2187 |
+
"eval_loss": 0.1610790491104126,
|
| 2188 |
+
"eval_runtime": 471.6243,
|
| 2189 |
+
"eval_samples_per_second": 22.215,
|
| 2190 |
+
"eval_steps_per_second": 2.778,
|
| 2191 |
+
"eval_wer": 0.13509843141383304,
|
| 2192 |
+
"step": 28000
|
| 2193 |
+
},
|
| 2194 |
+
{
|
| 2195 |
+
"epoch": 29.0,
|
| 2196 |
+
"learning_rate": 2.1231491136600624e-05,
|
| 2197 |
+
"loss": 0.694,
|
| 2198 |
+
"step": 28100
|
| 2199 |
+
},
|
| 2200 |
+
{
|
| 2201 |
+
"epoch": 29.1,
|
| 2202 |
+
"learning_rate": 2.1127215849843587e-05,
|
| 2203 |
+
"loss": 0.6839,
|
| 2204 |
+
"step": 28200
|
| 2205 |
+
},
|
| 2206 |
+
{
|
| 2207 |
+
"epoch": 29.21,
|
| 2208 |
+
"learning_rate": 2.102294056308655e-05,
|
| 2209 |
+
"loss": 0.6791,
|
| 2210 |
+
"step": 28300
|
| 2211 |
+
},
|
| 2212 |
+
{
|
| 2213 |
+
"epoch": 29.31,
|
| 2214 |
+
"learning_rate": 2.091866527632951e-05,
|
| 2215 |
+
"loss": 0.6855,
|
| 2216 |
+
"step": 28400
|
| 2217 |
+
},
|
| 2218 |
+
{
|
| 2219 |
+
"epoch": 29.41,
|
| 2220 |
+
"learning_rate": 2.0814389989572473e-05,
|
| 2221 |
+
"loss": 0.6798,
|
| 2222 |
+
"step": 28500
|
| 2223 |
+
},
|
| 2224 |
+
{
|
| 2225 |
+
"epoch": 29.41,
|
| 2226 |
+
"eval_loss": 0.16391155123710632,
|
| 2227 |
+
"eval_runtime": 478.5296,
|
| 2228 |
+
"eval_samples_per_second": 21.894,
|
| 2229 |
+
"eval_steps_per_second": 2.738,
|
| 2230 |
+
"eval_wer": 0.1364669965259501,
|
| 2231 |
+
"step": 28500
|
| 2232 |
+
},
|
| 2233 |
+
{
|
| 2234 |
+
"epoch": 29.51,
|
| 2235 |
+
"learning_rate": 2.0710114702815432e-05,
|
| 2236 |
+
"loss": 0.6755,
|
| 2237 |
+
"step": 28600
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"epoch": 29.62,
|
| 2241 |
+
"learning_rate": 2.0605839416058394e-05,
|
| 2242 |
+
"loss": 0.6892,
|
| 2243 |
+
"step": 28700
|
| 2244 |
+
},
|
| 2245 |
+
{
|
| 2246 |
+
"epoch": 29.72,
|
| 2247 |
+
"learning_rate": 2.0501564129301357e-05,
|
| 2248 |
+
"loss": 0.6883,
|
| 2249 |
+
"step": 28800
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 29.82,
|
| 2253 |
+
"learning_rate": 2.039728884254432e-05,
|
| 2254 |
+
"loss": 0.685,
|
| 2255 |
+
"step": 28900
|
| 2256 |
+
},
|
| 2257 |
+
{
|
| 2258 |
+
"epoch": 29.93,
|
| 2259 |
+
"learning_rate": 2.029301355578728e-05,
|
| 2260 |
+
"loss": 0.6711,
|
| 2261 |
+
"step": 29000
|
| 2262 |
+
},
|
| 2263 |
+
{
|
| 2264 |
+
"epoch": 29.93,
|
| 2265 |
+
"eval_loss": 0.16675817966461182,
|
| 2266 |
+
"eval_runtime": 467.8784,
|
| 2267 |
+
"eval_samples_per_second": 22.393,
|
| 2268 |
+
"eval_steps_per_second": 2.8,
|
| 2269 |
+
"eval_wer": 0.13579324139383092,
|
| 2270 |
+
"step": 29000
|
| 2271 |
+
},
|
| 2272 |
+
{
|
| 2273 |
+
"epoch": 30.03,
|
| 2274 |
+
"learning_rate": 2.018873826903024e-05,
|
| 2275 |
+
"loss": 0.6585,
|
| 2276 |
+
"step": 29100
|
| 2277 |
+
},
|
| 2278 |
+
{
|
| 2279 |
+
"epoch": 30.13,
|
| 2280 |
+
"learning_rate": 2.0084462982273202e-05,
|
| 2281 |
+
"loss": 0.6604,
|
| 2282 |
+
"step": 29200
|
| 2283 |
+
},
|
| 2284 |
+
{
|
| 2285 |
+
"epoch": 30.24,
|
| 2286 |
+
"learning_rate": 1.9980187695516165e-05,
|
| 2287 |
+
"loss": 0.6662,
|
| 2288 |
+
"step": 29300
|
| 2289 |
+
},
|
| 2290 |
+
{
|
| 2291 |
+
"epoch": 30.34,
|
| 2292 |
+
"learning_rate": 1.9875912408759127e-05,
|
| 2293 |
+
"loss": 0.6639,
|
| 2294 |
+
"step": 29400
|
| 2295 |
+
},
|
| 2296 |
+
{
|
| 2297 |
+
"epoch": 30.44,
|
| 2298 |
+
"learning_rate": 1.9771637122002086e-05,
|
| 2299 |
+
"loss": 0.6762,
|
| 2300 |
+
"step": 29500
|
| 2301 |
+
},
|
| 2302 |
+
{
|
| 2303 |
+
"epoch": 30.44,
|
| 2304 |
+
"eval_loss": 0.1681944876909256,
|
| 2305 |
+
"eval_runtime": 470.2202,
|
| 2306 |
+
"eval_samples_per_second": 22.281,
|
| 2307 |
+
"eval_steps_per_second": 2.786,
|
| 2308 |
+
"eval_wer": 0.13550900094746815,
|
| 2309 |
+
"step": 29500
|
| 2310 |
+
},
|
| 2311 |
+
{
|
| 2312 |
+
"epoch": 30.55,
|
| 2313 |
+
"learning_rate": 1.9667361835245048e-05,
|
| 2314 |
+
"loss": 0.6732,
|
| 2315 |
+
"step": 29600
|
| 2316 |
+
},
|
| 2317 |
+
{
|
| 2318 |
+
"epoch": 30.65,
|
| 2319 |
+
"learning_rate": 1.956308654848801e-05,
|
| 2320 |
+
"loss": 0.6729,
|
| 2321 |
+
"step": 29700
|
| 2322 |
+
},
|
| 2323 |
+
{
|
| 2324 |
+
"epoch": 30.75,
|
| 2325 |
+
"learning_rate": 1.9458811261730973e-05,
|
| 2326 |
+
"loss": 0.6591,
|
| 2327 |
+
"step": 29800
|
| 2328 |
+
},
|
| 2329 |
+
{
|
| 2330 |
+
"epoch": 30.86,
|
| 2331 |
+
"learning_rate": 1.935453597497393e-05,
|
| 2332 |
+
"loss": 0.6612,
|
| 2333 |
+
"step": 29900
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 30.96,
|
| 2337 |
+
"learning_rate": 1.9250260688216894e-05,
|
| 2338 |
+
"loss": 0.6594,
|
| 2339 |
+
"step": 30000
|
| 2340 |
+
},
|
| 2341 |
+
{
|
| 2342 |
+
"epoch": 30.96,
|
| 2343 |
+
"eval_loss": 0.16289828717708588,
|
| 2344 |
+
"eval_runtime": 468.3015,
|
| 2345 |
+
"eval_samples_per_second": 22.372,
|
| 2346 |
+
"eval_steps_per_second": 2.797,
|
| 2347 |
+
"eval_wer": 0.13452995052110747,
|
| 2348 |
+
"step": 30000
|
| 2349 |
+
},
|
| 2350 |
+
{
|
| 2351 |
+
"epoch": 31.06,
|
| 2352 |
+
"learning_rate": 1.9145985401459853e-05,
|
| 2353 |
+
"loss": 0.6624,
|
| 2354 |
+
"step": 30100
|
| 2355 |
+
},
|
| 2356 |
+
{
|
| 2357 |
+
"epoch": 31.17,
|
| 2358 |
+
"learning_rate": 1.9041710114702815e-05,
|
| 2359 |
+
"loss": 0.6696,
|
| 2360 |
+
"step": 30200
|
| 2361 |
+
},
|
| 2362 |
+
{
|
| 2363 |
+
"epoch": 31.27,
|
| 2364 |
+
"learning_rate": 1.8937434827945777e-05,
|
| 2365 |
+
"loss": 0.6672,
|
| 2366 |
+
"step": 30300
|
| 2367 |
+
},
|
| 2368 |
+
{
|
| 2369 |
+
"epoch": 31.37,
|
| 2370 |
+
"learning_rate": 1.883315954118874e-05,
|
| 2371 |
+
"loss": 0.666,
|
| 2372 |
+
"step": 30400
|
| 2373 |
+
},
|
| 2374 |
+
{
|
| 2375 |
+
"epoch": 31.48,
|
| 2376 |
+
"learning_rate": 1.87288842544317e-05,
|
| 2377 |
+
"loss": 0.6664,
|
| 2378 |
+
"step": 30500
|
| 2379 |
+
},
|
| 2380 |
+
{
|
| 2381 |
+
"epoch": 31.48,
|
| 2382 |
+
"eval_loss": 0.16248419880867004,
|
| 2383 |
+
"eval_runtime": 467.0084,
|
| 2384 |
+
"eval_samples_per_second": 22.434,
|
| 2385 |
+
"eval_steps_per_second": 2.805,
|
| 2386 |
+
"eval_wer": 0.1321402252868723,
|
| 2387 |
+
"step": 30500
|
| 2388 |
+
},
|
| 2389 |
+
{
|
| 2390 |
+
"epoch": 31.58,
|
| 2391 |
+
"learning_rate": 1.862460896767466e-05,
|
| 2392 |
+
"loss": 0.6811,
|
| 2393 |
+
"step": 30600
|
| 2394 |
+
},
|
| 2395 |
+
{
|
| 2396 |
+
"epoch": 31.68,
|
| 2397 |
+
"learning_rate": 1.8520333680917623e-05,
|
| 2398 |
+
"loss": 0.653,
|
| 2399 |
+
"step": 30700
|
| 2400 |
+
},
|
| 2401 |
+
{
|
| 2402 |
+
"epoch": 31.79,
|
| 2403 |
+
"learning_rate": 1.8416058394160585e-05,
|
| 2404 |
+
"loss": 0.6717,
|
| 2405 |
+
"step": 30800
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 31.89,
|
| 2409 |
+
"learning_rate": 1.8311783107403547e-05,
|
| 2410 |
+
"loss": 0.6654,
|
| 2411 |
+
"step": 30900
|
| 2412 |
+
},
|
| 2413 |
+
{
|
| 2414 |
+
"epoch": 31.99,
|
| 2415 |
+
"learning_rate": 1.8207507820646506e-05,
|
| 2416 |
+
"loss": 0.6838,
|
| 2417 |
+
"step": 31000
|
| 2418 |
+
},
|
| 2419 |
+
{
|
| 2420 |
+
"epoch": 31.99,
|
| 2421 |
+
"eval_loss": 0.15968269109725952,
|
| 2422 |
+
"eval_runtime": 469.2514,
|
| 2423 |
+
"eval_samples_per_second": 22.327,
|
| 2424 |
+
"eval_steps_per_second": 2.792,
|
| 2425 |
+
"eval_wer": 0.13722497104958417,
|
| 2426 |
+
"step": 31000
|
| 2427 |
+
},
|
| 2428 |
+
{
|
| 2429 |
+
"epoch": 32.09,
|
| 2430 |
+
"learning_rate": 1.810323253388947e-05,
|
| 2431 |
+
"loss": 0.6643,
|
| 2432 |
+
"step": 31100
|
| 2433 |
+
},
|
| 2434 |
+
{
|
| 2435 |
+
"epoch": 32.2,
|
| 2436 |
+
"learning_rate": 1.799895724713243e-05,
|
| 2437 |
+
"loss": 0.6728,
|
| 2438 |
+
"step": 31200
|
| 2439 |
+
},
|
| 2440 |
+
{
|
| 2441 |
+
"epoch": 32.3,
|
| 2442 |
+
"learning_rate": 1.7894681960375393e-05,
|
| 2443 |
+
"loss": 0.6599,
|
| 2444 |
+
"step": 31300
|
| 2445 |
+
},
|
| 2446 |
+
{
|
| 2447 |
+
"epoch": 32.4,
|
| 2448 |
+
"learning_rate": 1.7790406673618355e-05,
|
| 2449 |
+
"loss": 0.6645,
|
| 2450 |
+
"step": 31400
|
| 2451 |
+
},
|
| 2452 |
+
{
|
| 2453 |
+
"epoch": 32.51,
|
| 2454 |
+
"learning_rate": 1.7686131386861314e-05,
|
| 2455 |
+
"loss": 0.6603,
|
| 2456 |
+
"step": 31500
|
| 2457 |
+
},
|
| 2458 |
+
{
|
| 2459 |
+
"epoch": 32.51,
|
| 2460 |
+
"eval_loss": 0.1583111584186554,
|
| 2461 |
+
"eval_runtime": 484.3394,
|
| 2462 |
+
"eval_samples_per_second": 21.632,
|
| 2463 |
+
"eval_steps_per_second": 2.705,
|
| 2464 |
+
"eval_wer": 0.13024528897778714,
|
| 2465 |
+
"step": 31500
|
| 2466 |
+
},
|
| 2467 |
+
{
|
| 2468 |
+
"epoch": 32.61,
|
| 2469 |
+
"learning_rate": 1.7581856100104276e-05,
|
| 2470 |
+
"loss": 0.6659,
|
| 2471 |
+
"step": 31600
|
| 2472 |
+
},
|
| 2473 |
+
{
|
| 2474 |
+
"epoch": 32.71,
|
| 2475 |
+
"learning_rate": 1.747758081334724e-05,
|
| 2476 |
+
"loss": 0.6551,
|
| 2477 |
+
"step": 31700
|
| 2478 |
+
},
|
| 2479 |
+
{
|
| 2480 |
+
"epoch": 32.82,
|
| 2481 |
+
"learning_rate": 1.73733055265902e-05,
|
| 2482 |
+
"loss": 0.6581,
|
| 2483 |
+
"step": 31800
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 32.92,
|
| 2487 |
+
"learning_rate": 1.7269030239833163e-05,
|
| 2488 |
+
"loss": 0.6526,
|
| 2489 |
+
"step": 31900
|
| 2490 |
+
},
|
| 2491 |
+
{
|
| 2492 |
+
"epoch": 33.02,
|
| 2493 |
+
"learning_rate": 1.7164754953076122e-05,
|
| 2494 |
+
"loss": 0.6468,
|
| 2495 |
+
"step": 32000
|
| 2496 |
+
},
|
| 2497 |
+
{
|
| 2498 |
+
"epoch": 33.02,
|
| 2499 |
+
"eval_loss": 0.15948732197284698,
|
| 2500 |
+
"eval_runtime": 467.9806,
|
| 2501 |
+
"eval_samples_per_second": 22.388,
|
| 2502 |
+
"eval_steps_per_second": 2.799,
|
| 2503 |
+
"eval_wer": 0.13224549952626594,
|
| 2504 |
+
"step": 32000
|
| 2505 |
+
},
|
| 2506 |
+
{
|
| 2507 |
+
"epoch": 33.13,
|
| 2508 |
+
"learning_rate": 1.706047966631908e-05,
|
| 2509 |
+
"loss": 0.642,
|
| 2510 |
+
"step": 32100
|
| 2511 |
+
},
|
| 2512 |
+
{
|
| 2513 |
+
"epoch": 33.23,
|
| 2514 |
+
"learning_rate": 1.6956204379562043e-05,
|
| 2515 |
+
"loss": 0.6546,
|
| 2516 |
+
"step": 32200
|
| 2517 |
+
},
|
| 2518 |
+
{
|
| 2519 |
+
"epoch": 33.33,
|
| 2520 |
+
"learning_rate": 1.685297184567258e-05,
|
| 2521 |
+
"loss": 0.6438,
|
| 2522 |
+
"step": 32300
|
| 2523 |
+
},
|
| 2524 |
+
{
|
| 2525 |
+
"epoch": 33.44,
|
| 2526 |
+
"learning_rate": 1.6748696558915537e-05,
|
| 2527 |
+
"loss": 0.6466,
|
| 2528 |
+
"step": 32400
|
| 2529 |
+
},
|
| 2530 |
+
{
|
| 2531 |
+
"epoch": 33.54,
|
| 2532 |
+
"learning_rate": 1.66444212721585e-05,
|
| 2533 |
+
"loss": 0.6464,
|
| 2534 |
+
"step": 32500
|
| 2535 |
+
},
|
| 2536 |
+
{
|
| 2537 |
+
"epoch": 33.54,
|
| 2538 |
+
"eval_loss": 0.16093206405639648,
|
| 2539 |
+
"eval_runtime": 468.2076,
|
| 2540 |
+
"eval_samples_per_second": 22.377,
|
| 2541 |
+
"eval_steps_per_second": 2.798,
|
| 2542 |
+
"eval_wer": 0.13148752500263186,
|
| 2543 |
+
"step": 32500
|
| 2544 |
+
},
|
| 2545 |
+
{
|
| 2546 |
+
"epoch": 33.64,
|
| 2547 |
+
"learning_rate": 1.6540145985401462e-05,
|
| 2548 |
+
"loss": 0.6528,
|
| 2549 |
+
"step": 32600
|
| 2550 |
+
},
|
| 2551 |
+
{
|
| 2552 |
+
"epoch": 33.75,
|
| 2553 |
+
"learning_rate": 1.6435870698644424e-05,
|
| 2554 |
+
"loss": 0.641,
|
| 2555 |
+
"step": 32700
|
| 2556 |
+
},
|
| 2557 |
+
{
|
| 2558 |
+
"epoch": 33.85,
|
| 2559 |
+
"learning_rate": 1.6331595411887386e-05,
|
| 2560 |
+
"loss": 0.6416,
|
| 2561 |
+
"step": 32800
|
| 2562 |
+
},
|
| 2563 |
+
{
|
| 2564 |
+
"epoch": 33.95,
|
| 2565 |
+
"learning_rate": 1.6227320125130342e-05,
|
| 2566 |
+
"loss": 0.6398,
|
| 2567 |
+
"step": 32900
|
| 2568 |
+
},
|
| 2569 |
+
{
|
| 2570 |
+
"epoch": 34.06,
|
| 2571 |
+
"learning_rate": 1.6123044838373304e-05,
|
| 2572 |
+
"loss": 0.6623,
|
| 2573 |
+
"step": 33000
|
| 2574 |
+
},
|
| 2575 |
+
{
|
| 2576 |
+
"epoch": 34.06,
|
| 2577 |
+
"eval_loss": 0.16215749084949493,
|
| 2578 |
+
"eval_runtime": 466.7215,
|
| 2579 |
+
"eval_samples_per_second": 22.448,
|
| 2580 |
+
"eval_steps_per_second": 2.807,
|
| 2581 |
+
"eval_wer": 0.13661438046110116,
|
| 2582 |
+
"step": 33000
|
| 2583 |
+
},
|
| 2584 |
+
{
|
| 2585 |
+
"epoch": 34.16,
|
| 2586 |
+
"learning_rate": 1.6018769551616266e-05,
|
| 2587 |
+
"loss": 0.6433,
|
| 2588 |
+
"step": 33100
|
| 2589 |
+
},
|
| 2590 |
+
{
|
| 2591 |
+
"epoch": 34.26,
|
| 2592 |
+
"learning_rate": 1.591449426485923e-05,
|
| 2593 |
+
"loss": 0.6484,
|
| 2594 |
+
"step": 33200
|
| 2595 |
+
},
|
| 2596 |
+
{
|
| 2597 |
+
"epoch": 34.37,
|
| 2598 |
+
"learning_rate": 1.581021897810219e-05,
|
| 2599 |
+
"loss": 0.6483,
|
| 2600 |
+
"step": 33300
|
| 2601 |
+
},
|
| 2602 |
+
{
|
| 2603 |
+
"epoch": 34.47,
|
| 2604 |
+
"learning_rate": 1.570594369134515e-05,
|
| 2605 |
+
"loss": 0.6439,
|
| 2606 |
+
"step": 33400
|
| 2607 |
+
},
|
| 2608 |
+
{
|
| 2609 |
+
"epoch": 34.57,
|
| 2610 |
+
"learning_rate": 1.5601668404588112e-05,
|
| 2611 |
+
"loss": 0.6414,
|
| 2612 |
+
"step": 33500
|
| 2613 |
+
},
|
| 2614 |
+
{
|
| 2615 |
+
"epoch": 34.57,
|
| 2616 |
+
"eval_loss": 0.15873852372169495,
|
| 2617 |
+
"eval_runtime": 469.8368,
|
| 2618 |
+
"eval_samples_per_second": 22.299,
|
| 2619 |
+
"eval_steps_per_second": 2.788,
|
| 2620 |
+
"eval_wer": 0.13302452889777872,
|
| 2621 |
+
"step": 33500
|
| 2622 |
+
},
|
| 2623 |
+
{
|
| 2624 |
+
"epoch": 34.67,
|
| 2625 |
+
"learning_rate": 1.5497393117831074e-05,
|
| 2626 |
+
"loss": 0.6458,
|
| 2627 |
+
"step": 33600
|
| 2628 |
+
},
|
| 2629 |
+
{
|
| 2630 |
+
"epoch": 34.78,
|
| 2631 |
+
"learning_rate": 1.5393117831074037e-05,
|
| 2632 |
+
"loss": 0.6423,
|
| 2633 |
+
"step": 33700
|
| 2634 |
+
},
|
| 2635 |
+
{
|
| 2636 |
+
"epoch": 34.88,
|
| 2637 |
+
"learning_rate": 1.5288842544317e-05,
|
| 2638 |
+
"loss": 0.6416,
|
| 2639 |
+
"step": 33800
|
| 2640 |
+
},
|
| 2641 |
+
{
|
| 2642 |
+
"epoch": 34.98,
|
| 2643 |
+
"learning_rate": 1.518456725755996e-05,
|
| 2644 |
+
"loss": 0.6481,
|
| 2645 |
+
"step": 33900
|
| 2646 |
+
},
|
| 2647 |
+
{
|
| 2648 |
+
"epoch": 35.09,
|
| 2649 |
+
"learning_rate": 1.508029197080292e-05,
|
| 2650 |
+
"loss": 0.6242,
|
| 2651 |
+
"step": 34000
|
| 2652 |
+
},
|
| 2653 |
+
{
|
| 2654 |
+
"epoch": 35.09,
|
| 2655 |
+
"eval_loss": 0.16140839457511902,
|
| 2656 |
+
"eval_runtime": 481.1715,
|
| 2657 |
+
"eval_samples_per_second": 21.774,
|
| 2658 |
+
"eval_steps_per_second": 2.723,
|
| 2659 |
+
"eval_wer": 0.1336982840298979,
|
| 2660 |
+
"step": 34000
|
| 2661 |
+
},
|
| 2662 |
+
{
|
| 2663 |
+
"epoch": 35.19,
|
| 2664 |
+
"learning_rate": 1.4976016684045882e-05,
|
| 2665 |
+
"loss": 0.6335,
|
| 2666 |
+
"step": 34100
|
| 2667 |
+
},
|
| 2668 |
+
{
|
| 2669 |
+
"epoch": 35.29,
|
| 2670 |
+
"learning_rate": 1.4871741397288843e-05,
|
| 2671 |
+
"loss": 0.6365,
|
| 2672 |
+
"step": 34200
|
| 2673 |
+
},
|
| 2674 |
+
{
|
| 2675 |
+
"epoch": 35.4,
|
| 2676 |
+
"learning_rate": 1.4767466110531805e-05,
|
| 2677 |
+
"loss": 0.6392,
|
| 2678 |
+
"step": 34300
|
| 2679 |
+
},
|
| 2680 |
+
{
|
| 2681 |
+
"epoch": 35.5,
|
| 2682 |
+
"learning_rate": 1.4663190823774767e-05,
|
| 2683 |
+
"loss": 0.6355,
|
| 2684 |
+
"step": 34400
|
| 2685 |
+
},
|
| 2686 |
+
{
|
| 2687 |
+
"epoch": 35.6,
|
| 2688 |
+
"learning_rate": 1.4558915537017728e-05,
|
| 2689 |
+
"loss": 0.632,
|
| 2690 |
+
"step": 34500
|
| 2691 |
+
},
|
| 2692 |
+
{
|
| 2693 |
+
"epoch": 35.6,
|
| 2694 |
+
"eval_loss": 0.15684087574481964,
|
| 2695 |
+
"eval_runtime": 469.3002,
|
| 2696 |
+
"eval_samples_per_second": 22.325,
|
| 2697 |
+
"eval_steps_per_second": 2.791,
|
| 2698 |
+
"eval_wer": 0.12724497315506894,
|
| 2699 |
+
"step": 34500
|
| 2700 |
+
},
|
| 2701 |
+
{
|
| 2702 |
+
"epoch": 35.71,
|
| 2703 |
+
"learning_rate": 1.4455683003128258e-05,
|
| 2704 |
+
"loss": 0.6337,
|
| 2705 |
+
"step": 34600
|
| 2706 |
+
},
|
| 2707 |
+
{
|
| 2708 |
+
"epoch": 35.81,
|
| 2709 |
+
"learning_rate": 1.435140771637122e-05,
|
| 2710 |
+
"loss": 0.6304,
|
| 2711 |
+
"step": 34700
|
| 2712 |
+
},
|
| 2713 |
+
{
|
| 2714 |
+
"epoch": 35.91,
|
| 2715 |
+
"learning_rate": 1.4247132429614183e-05,
|
| 2716 |
+
"loss": 0.6398,
|
| 2717 |
+
"step": 34800
|
| 2718 |
+
},
|
| 2719 |
+
{
|
| 2720 |
+
"epoch": 36.02,
|
| 2721 |
+
"learning_rate": 1.4142857142857143e-05,
|
| 2722 |
+
"loss": 0.6341,
|
| 2723 |
+
"step": 34900
|
| 2724 |
+
},
|
| 2725 |
+
{
|
| 2726 |
+
"epoch": 36.12,
|
| 2727 |
+
"learning_rate": 1.4038581856100106e-05,
|
| 2728 |
+
"loss": 0.6346,
|
| 2729 |
+
"step": 35000
|
| 2730 |
+
},
|
| 2731 |
+
{
|
| 2732 |
+
"epoch": 36.12,
|
| 2733 |
+
"eval_loss": 0.15833768248558044,
|
| 2734 |
+
"eval_runtime": 469.6695,
|
| 2735 |
+
"eval_samples_per_second": 22.307,
|
| 2736 |
+
"eval_steps_per_second": 2.789,
|
| 2737 |
+
"eval_wer": 0.12739235709022,
|
| 2738 |
+
"step": 35000
|
| 2739 |
+
},
|
| 2740 |
+
{
|
| 2741 |
+
"epoch": 36.22,
|
| 2742 |
+
"learning_rate": 1.3934306569343066e-05,
|
| 2743 |
+
"loss": 0.6286,
|
| 2744 |
+
"step": 35100
|
| 2745 |
+
},
|
| 2746 |
+
{
|
| 2747 |
+
"epoch": 36.33,
|
| 2748 |
+
"learning_rate": 1.3830031282586029e-05,
|
| 2749 |
+
"loss": 0.6331,
|
| 2750 |
+
"step": 35200
|
| 2751 |
+
},
|
| 2752 |
+
{
|
| 2753 |
+
"epoch": 36.43,
|
| 2754 |
+
"learning_rate": 1.372575599582899e-05,
|
| 2755 |
+
"loss": 0.6256,
|
| 2756 |
+
"step": 35300
|
| 2757 |
+
},
|
| 2758 |
+
{
|
| 2759 |
+
"epoch": 36.53,
|
| 2760 |
+
"learning_rate": 1.3621480709071951e-05,
|
| 2761 |
+
"loss": 0.6289,
|
| 2762 |
+
"step": 35400
|
| 2763 |
+
},
|
| 2764 |
+
{
|
| 2765 |
+
"epoch": 36.64,
|
| 2766 |
+
"learning_rate": 1.3517205422314914e-05,
|
| 2767 |
+
"loss": 0.6143,
|
| 2768 |
+
"step": 35500
|
| 2769 |
+
},
|
| 2770 |
+
{
|
| 2771 |
+
"epoch": 36.64,
|
| 2772 |
+
"eval_loss": 0.15764719247817993,
|
| 2773 |
+
"eval_runtime": 468.9196,
|
| 2774 |
+
"eval_samples_per_second": 22.343,
|
| 2775 |
+
"eval_steps_per_second": 2.794,
|
| 2776 |
+
"eval_wer": 0.12639225181598063,
|
| 2777 |
+
"step": 35500
|
| 2778 |
+
},
|
| 2779 |
+
{
|
| 2780 |
+
"epoch": 36.74,
|
| 2781 |
+
"learning_rate": 1.3412930135557874e-05,
|
| 2782 |
+
"loss": 0.6414,
|
| 2783 |
+
"step": 35600
|
| 2784 |
+
},
|
| 2785 |
+
{
|
| 2786 |
+
"epoch": 36.84,
|
| 2787 |
+
"learning_rate": 1.3308654848800836e-05,
|
| 2788 |
+
"loss": 0.6184,
|
| 2789 |
+
"step": 35700
|
| 2790 |
+
},
|
| 2791 |
+
{
|
| 2792 |
+
"epoch": 36.95,
|
| 2793 |
+
"learning_rate": 1.3204379562043795e-05,
|
| 2794 |
+
"loss": 0.6227,
|
| 2795 |
+
"step": 35800
|
| 2796 |
+
},
|
| 2797 |
+
{
|
| 2798 |
+
"epoch": 37.05,
|
| 2799 |
+
"learning_rate": 1.3100104275286756e-05,
|
| 2800 |
+
"loss": 0.6488,
|
| 2801 |
+
"step": 35900
|
| 2802 |
+
},
|
| 2803 |
+
{
|
| 2804 |
+
"epoch": 37.15,
|
| 2805 |
+
"learning_rate": 1.2995828988529718e-05,
|
| 2806 |
+
"loss": 0.6208,
|
| 2807 |
+
"step": 36000
|
| 2808 |
+
},
|
| 2809 |
+
{
|
| 2810 |
+
"epoch": 37.15,
|
| 2811 |
+
"eval_loss": 0.16206438839435577,
|
| 2812 |
+
"eval_runtime": 468.3673,
|
| 2813 |
+
"eval_samples_per_second": 22.369,
|
| 2814 |
+
"eval_steps_per_second": 2.797,
|
| 2815 |
+
"eval_wer": 0.12628697757658702,
|
| 2816 |
+
"step": 36000
|
| 2817 |
+
},
|
| 2818 |
+
{
|
| 2819 |
+
"epoch": 37.25,
|
| 2820 |
+
"learning_rate": 1.2891553701772679e-05,
|
| 2821 |
+
"loss": 0.6335,
|
| 2822 |
+
"step": 36100
|
| 2823 |
+
},
|
| 2824 |
+
{
|
| 2825 |
+
"epoch": 37.36,
|
| 2826 |
+
"learning_rate": 1.2787278415015641e-05,
|
| 2827 |
+
"loss": 0.6227,
|
| 2828 |
+
"step": 36200
|
| 2829 |
+
},
|
| 2830 |
+
{
|
| 2831 |
+
"epoch": 37.46,
|
| 2832 |
+
"learning_rate": 1.2683003128258603e-05,
|
| 2833 |
+
"loss": 0.6291,
|
| 2834 |
+
"step": 36300
|
| 2835 |
+
},
|
| 2836 |
+
{
|
| 2837 |
+
"epoch": 37.56,
|
| 2838 |
+
"learning_rate": 1.2578727841501564e-05,
|
| 2839 |
+
"loss": 0.6203,
|
| 2840 |
+
"step": 36400
|
| 2841 |
+
},
|
| 2842 |
+
{
|
| 2843 |
+
"epoch": 37.67,
|
| 2844 |
+
"learning_rate": 1.2474452554744526e-05,
|
| 2845 |
+
"loss": 0.6185,
|
| 2846 |
+
"step": 36500
|
| 2847 |
+
},
|
| 2848 |
+
{
|
| 2849 |
+
"epoch": 37.67,
|
| 2850 |
+
"eval_loss": 0.16232742369174957,
|
| 2851 |
+
"eval_runtime": 467.766,
|
| 2852 |
+
"eval_samples_per_second": 22.398,
|
| 2853 |
+
"eval_steps_per_second": 2.801,
|
| 2854 |
+
"eval_wer": 0.12702389725234234,
|
| 2855 |
+
"step": 36500
|
| 2856 |
+
},
|
| 2857 |
+
{
|
| 2858 |
+
"epoch": 37.77,
|
| 2859 |
+
"learning_rate": 1.2370177267987487e-05,
|
| 2860 |
+
"loss": 0.6137,
|
| 2861 |
+
"step": 36600
|
| 2862 |
+
},
|
| 2863 |
+
{
|
| 2864 |
+
"epoch": 37.87,
|
| 2865 |
+
"learning_rate": 1.2265901981230449e-05,
|
| 2866 |
+
"loss": 0.6131,
|
| 2867 |
+
"step": 36700
|
| 2868 |
+
},
|
| 2869 |
+
{
|
| 2870 |
+
"epoch": 37.98,
|
| 2871 |
+
"learning_rate": 1.216162669447341e-05,
|
| 2872 |
+
"loss": 0.6076,
|
| 2873 |
+
"step": 36800
|
| 2874 |
+
},
|
| 2875 |
+
{
|
| 2876 |
+
"epoch": 38.08,
|
| 2877 |
+
"learning_rate": 1.2057351407716372e-05,
|
| 2878 |
+
"loss": 0.6058,
|
| 2879 |
+
"step": 36900
|
| 2880 |
+
},
|
| 2881 |
+
{
|
| 2882 |
+
"epoch": 38.18,
|
| 2883 |
+
"learning_rate": 1.1953076120959334e-05,
|
| 2884 |
+
"loss": 0.6128,
|
| 2885 |
+
"step": 37000
|
| 2886 |
+
},
|
| 2887 |
+
{
|
| 2888 |
+
"epoch": 38.18,
|
| 2889 |
+
"eval_loss": 0.16044454276561737,
|
| 2890 |
+
"eval_runtime": 479.5033,
|
| 2891 |
+
"eval_samples_per_second": 21.85,
|
| 2892 |
+
"eval_steps_per_second": 2.732,
|
| 2893 |
+
"eval_wer": 0.12678176650173703,
|
| 2894 |
+
"step": 37000
|
| 2895 |
+
},
|
| 2896 |
+
{
|
| 2897 |
+
"epoch": 38.29,
|
| 2898 |
+
"learning_rate": 1.1848800834202295e-05,
|
| 2899 |
+
"loss": 0.6095,
|
| 2900 |
+
"step": 37100
|
| 2901 |
+
},
|
| 2902 |
+
{
|
| 2903 |
+
"epoch": 38.39,
|
| 2904 |
+
"learning_rate": 1.1746611053180396e-05,
|
| 2905 |
+
"loss": 0.62,
|
| 2906 |
+
"step": 37200
|
| 2907 |
+
},
|
| 2908 |
+
{
|
| 2909 |
+
"epoch": 38.49,
|
| 2910 |
+
"learning_rate": 1.1642335766423359e-05,
|
| 2911 |
+
"loss": 0.607,
|
| 2912 |
+
"step": 37300
|
| 2913 |
+
},
|
| 2914 |
+
{
|
| 2915 |
+
"epoch": 38.6,
|
| 2916 |
+
"learning_rate": 1.153806047966632e-05,
|
| 2917 |
+
"loss": 0.62,
|
| 2918 |
+
"step": 37400
|
| 2919 |
+
},
|
| 2920 |
+
{
|
| 2921 |
+
"epoch": 38.7,
|
| 2922 |
+
"learning_rate": 1.1433785192909282e-05,
|
| 2923 |
+
"loss": 0.6151,
|
| 2924 |
+
"step": 37500
|
| 2925 |
+
},
|
| 2926 |
+
{
|
| 2927 |
+
"epoch": 38.7,
|
| 2928 |
+
"eval_loss": 0.15926694869995117,
|
| 2929 |
+
"eval_runtime": 481.5864,
|
| 2930 |
+
"eval_samples_per_second": 21.755,
|
| 2931 |
+
"eval_steps_per_second": 2.72,
|
| 2932 |
+
"eval_wer": 0.12461311717022845,
|
| 2933 |
+
"step": 37500
|
| 2934 |
+
},
|
| 2935 |
+
{
|
| 2936 |
+
"epoch": 38.8,
|
| 2937 |
+
"learning_rate": 1.1329509906152242e-05,
|
| 2938 |
+
"loss": 0.6178,
|
| 2939 |
+
"step": 37600
|
| 2940 |
+
},
|
| 2941 |
+
{
|
| 2942 |
+
"epoch": 38.91,
|
| 2943 |
+
"learning_rate": 1.1225234619395204e-05,
|
| 2944 |
+
"loss": 0.615,
|
| 2945 |
+
"step": 37700
|
| 2946 |
+
},
|
| 2947 |
+
{
|
| 2948 |
+
"epoch": 39.01,
|
| 2949 |
+
"learning_rate": 1.1120959332638167e-05,
|
| 2950 |
+
"loss": 0.6096,
|
| 2951 |
+
"step": 37800
|
| 2952 |
+
},
|
| 2953 |
+
{
|
| 2954 |
+
"epoch": 39.11,
|
| 2955 |
+
"learning_rate": 1.1016684045881126e-05,
|
| 2956 |
+
"loss": 0.6134,
|
| 2957 |
+
"step": 37900
|
| 2958 |
+
},
|
| 2959 |
+
{
|
| 2960 |
+
"epoch": 39.22,
|
| 2961 |
+
"learning_rate": 1.0912408759124088e-05,
|
| 2962 |
+
"loss": 0.6082,
|
| 2963 |
+
"step": 38000
|
| 2964 |
+
},
|
| 2965 |
+
{
|
| 2966 |
+
"epoch": 39.22,
|
| 2967 |
+
"eval_loss": 0.15319351851940155,
|
| 2968 |
+
"eval_runtime": 480.0879,
|
| 2969 |
+
"eval_samples_per_second": 21.823,
|
| 2970 |
+
"eval_steps_per_second": 2.729,
|
| 2971 |
+
"eval_wer": 0.12383408779871566,
|
| 2972 |
+
"step": 38000
|
| 2973 |
+
},
|
| 2974 |
+
{
|
| 2975 |
+
"epoch": 39.32,
|
| 2976 |
+
"learning_rate": 1.0808133472367048e-05,
|
| 2977 |
+
"loss": 0.6099,
|
| 2978 |
+
"step": 38100
|
| 2979 |
+
},
|
| 2980 |
+
{
|
| 2981 |
+
"epoch": 39.42,
|
| 2982 |
+
"learning_rate": 1.070385818561001e-05,
|
| 2983 |
+
"loss": 0.606,
|
| 2984 |
+
"step": 38200
|
| 2985 |
+
},
|
| 2986 |
+
{
|
| 2987 |
+
"epoch": 39.53,
|
| 2988 |
+
"learning_rate": 1.0599582898852973e-05,
|
| 2989 |
+
"loss": 0.6115,
|
| 2990 |
+
"step": 38300
|
| 2991 |
+
},
|
| 2992 |
+
{
|
| 2993 |
+
"epoch": 39.63,
|
| 2994 |
+
"learning_rate": 1.0495307612095933e-05,
|
| 2995 |
+
"loss": 0.6105,
|
| 2996 |
+
"step": 38400
|
| 2997 |
+
},
|
| 2998 |
+
{
|
| 2999 |
+
"epoch": 39.73,
|
| 3000 |
+
"learning_rate": 1.0391032325338896e-05,
|
| 3001 |
+
"loss": 0.6,
|
| 3002 |
+
"step": 38500
|
| 3003 |
+
},
|
| 3004 |
+
{
|
| 3005 |
+
"epoch": 39.73,
|
| 3006 |
+
"eval_loss": 0.15240420401096344,
|
| 3007 |
+
"eval_runtime": 471.9993,
|
| 3008 |
+
"eval_samples_per_second": 22.197,
|
| 3009 |
+
"eval_steps_per_second": 2.775,
|
| 3010 |
+
"eval_wer": 0.1224339404147805,
|
| 3011 |
+
"step": 38500
|
| 3012 |
+
},
|
| 3013 |
+
{
|
| 3014 |
+
"epoch": 39.83,
|
| 3015 |
+
"learning_rate": 1.0286757038581856e-05,
|
| 3016 |
+
"loss": 0.5997,
|
| 3017 |
+
"step": 38600
|
| 3018 |
+
},
|
| 3019 |
+
{
|
| 3020 |
+
"epoch": 39.94,
|
| 3021 |
+
"learning_rate": 1.0182481751824819e-05,
|
| 3022 |
+
"loss": 0.5957,
|
| 3023 |
+
"step": 38700
|
| 3024 |
+
},
|
| 3025 |
+
{
|
| 3026 |
+
"epoch": 40.04,
|
| 3027 |
+
"learning_rate": 1.0078206465067779e-05,
|
| 3028 |
+
"loss": 0.613,
|
| 3029 |
+
"step": 38800
|
| 3030 |
+
},
|
| 3031 |
+
{
|
| 3032 |
+
"epoch": 40.14,
|
| 3033 |
+
"learning_rate": 9.973931178310741e-06,
|
| 3034 |
+
"loss": 0.5961,
|
| 3035 |
+
"step": 38900
|
| 3036 |
+
},
|
| 3037 |
+
{
|
| 3038 |
+
"epoch": 40.25,
|
| 3039 |
+
"learning_rate": 9.869655891553702e-06,
|
| 3040 |
+
"loss": 0.6032,
|
| 3041 |
+
"step": 39000
|
| 3042 |
+
},
|
| 3043 |
+
{
|
| 3044 |
+
"epoch": 40.25,
|
| 3045 |
+
"eval_loss": 0.1521444320678711,
|
| 3046 |
+
"eval_runtime": 467.5416,
|
| 3047 |
+
"eval_samples_per_second": 22.409,
|
| 3048 |
+
"eval_steps_per_second": 2.802,
|
| 3049 |
+
"eval_wer": 0.12116012211811769,
|
| 3050 |
+
"step": 39000
|
| 3051 |
+
},
|
| 3052 |
+
{
|
| 3053 |
+
"epoch": 40.35,
|
| 3054 |
+
"learning_rate": 9.765380604796663e-06,
|
| 3055 |
+
"loss": 0.5952,
|
| 3056 |
+
"step": 39100
|
| 3057 |
+
},
|
| 3058 |
+
{
|
| 3059 |
+
"epoch": 40.45,
|
| 3060 |
+
"learning_rate": 9.661105318039625e-06,
|
| 3061 |
+
"loss": 0.6034,
|
| 3062 |
+
"step": 39200
|
| 3063 |
+
},
|
| 3064 |
+
{
|
| 3065 |
+
"epoch": 40.56,
|
| 3066 |
+
"learning_rate": 9.556830031282587e-06,
|
| 3067 |
+
"loss": 0.5917,
|
| 3068 |
+
"step": 39300
|
| 3069 |
+
},
|
| 3070 |
+
{
|
| 3071 |
+
"epoch": 40.66,
|
| 3072 |
+
"learning_rate": 9.452554744525548e-06,
|
| 3073 |
+
"loss": 0.602,
|
| 3074 |
+
"step": 39400
|
| 3075 |
+
},
|
| 3076 |
+
{
|
| 3077 |
+
"epoch": 40.76,
|
| 3078 |
+
"learning_rate": 9.34827945776851e-06,
|
| 3079 |
+
"loss": 0.6016,
|
| 3080 |
+
"step": 39500
|
| 3081 |
+
},
|
| 3082 |
+
{
|
| 3083 |
+
"epoch": 40.76,
|
| 3084 |
+
"eval_loss": 0.15509599447250366,
|
| 3085 |
+
"eval_runtime": 467.848,
|
| 3086 |
+
"eval_samples_per_second": 22.394,
|
| 3087 |
+
"eval_steps_per_second": 2.8,
|
| 3088 |
+
"eval_wer": 0.12151805453205601,
|
| 3089 |
+
"step": 39500
|
| 3090 |
+
},
|
| 3091 |
+
{
|
| 3092 |
+
"epoch": 40.87,
|
| 3093 |
+
"learning_rate": 9.24400417101147e-06,
|
| 3094 |
+
"loss": 0.6019,
|
| 3095 |
+
"step": 39600
|
| 3096 |
+
},
|
| 3097 |
+
{
|
| 3098 |
+
"epoch": 40.97,
|
| 3099 |
+
"learning_rate": 9.139728884254433e-06,
|
| 3100 |
+
"loss": 0.5999,
|
| 3101 |
+
"step": 39700
|
| 3102 |
+
},
|
| 3103 |
+
{
|
| 3104 |
+
"epoch": 41.07,
|
| 3105 |
+
"learning_rate": 9.035453597497393e-06,
|
| 3106 |
+
"loss": 0.5809,
|
| 3107 |
+
"step": 39800
|
| 3108 |
+
},
|
| 3109 |
+
{
|
| 3110 |
+
"epoch": 41.18,
|
| 3111 |
+
"learning_rate": 8.931178310740356e-06,
|
| 3112 |
+
"loss": 0.5907,
|
| 3113 |
+
"step": 39900
|
| 3114 |
+
},
|
| 3115 |
+
{
|
| 3116 |
+
"epoch": 41.28,
|
| 3117 |
+
"learning_rate": 8.826903023983318e-06,
|
| 3118 |
+
"loss": 0.6009,
|
| 3119 |
+
"step": 40000
|
| 3120 |
+
},
|
| 3121 |
+
{
|
| 3122 |
+
"epoch": 41.28,
|
| 3123 |
+
"eval_loss": 0.1522541046142578,
|
| 3124 |
+
"eval_runtime": 472.1828,
|
| 3125 |
+
"eval_samples_per_second": 22.188,
|
| 3126 |
+
"eval_steps_per_second": 2.774,
|
| 3127 |
+
"eval_wer": 0.12153910937993473,
|
| 3128 |
+
"step": 40000
|
| 3129 |
+
},
|
| 3130 |
+
{
|
| 3131 |
+
"epoch": 41.38,
|
| 3132 |
+
"learning_rate": 8.722627737226277e-06,
|
| 3133 |
+
"loss": 0.5898,
|
| 3134 |
+
"step": 40100
|
| 3135 |
+
},
|
| 3136 |
+
{
|
| 3137 |
+
"epoch": 41.49,
|
| 3138 |
+
"learning_rate": 8.618352450469239e-06,
|
| 3139 |
+
"loss": 0.6018,
|
| 3140 |
+
"step": 40200
|
| 3141 |
+
},
|
| 3142 |
+
{
|
| 3143 |
+
"epoch": 41.59,
|
| 3144 |
+
"learning_rate": 8.5140771637122e-06,
|
| 3145 |
+
"loss": 0.6043,
|
| 3146 |
+
"step": 40300
|
| 3147 |
+
},
|
| 3148 |
+
{
|
| 3149 |
+
"epoch": 41.69,
|
| 3150 |
+
"learning_rate": 8.409801876955162e-06,
|
| 3151 |
+
"loss": 0.585,
|
| 3152 |
+
"step": 40400
|
| 3153 |
+
},
|
| 3154 |
+
{
|
| 3155 |
+
"epoch": 41.8,
|
| 3156 |
+
"learning_rate": 8.305526590198124e-06,
|
| 3157 |
+
"loss": 0.5875,
|
| 3158 |
+
"step": 40500
|
| 3159 |
+
},
|
| 3160 |
+
{
|
| 3161 |
+
"epoch": 41.8,
|
| 3162 |
+
"eval_loss": 0.1540576070547104,
|
| 3163 |
+
"eval_runtime": 466.9539,
|
| 3164 |
+
"eval_samples_per_second": 22.437,
|
| 3165 |
+
"eval_steps_per_second": 2.805,
|
| 3166 |
+
"eval_wer": 0.1216022739235709,
|
| 3167 |
+
"step": 40500
|
| 3168 |
+
},
|
| 3169 |
+
{
|
| 3170 |
+
"epoch": 41.9,
|
| 3171 |
+
"learning_rate": 8.201251303441085e-06,
|
| 3172 |
+
"loss": 0.5941,
|
| 3173 |
+
"step": 40600
|
| 3174 |
+
},
|
| 3175 |
+
{
|
| 3176 |
+
"epoch": 42.0,
|
| 3177 |
+
"learning_rate": 8.096976016684047e-06,
|
| 3178 |
+
"loss": 0.6136,
|
| 3179 |
+
"step": 40700
|
| 3180 |
+
},
|
| 3181 |
+
{
|
| 3182 |
+
"epoch": 42.11,
|
| 3183 |
+
"learning_rate": 7.992700729927007e-06,
|
| 3184 |
+
"loss": 0.5969,
|
| 3185 |
+
"step": 40800
|
| 3186 |
+
},
|
| 3187 |
+
{
|
| 3188 |
+
"epoch": 42.21,
|
| 3189 |
+
"learning_rate": 7.88842544316997e-06,
|
| 3190 |
+
"loss": 0.5959,
|
| 3191 |
+
"step": 40900
|
| 3192 |
+
},
|
| 3193 |
+
{
|
| 3194 |
+
"epoch": 42.31,
|
| 3195 |
+
"learning_rate": 7.78415015641293e-06,
|
| 3196 |
+
"loss": 0.608,
|
| 3197 |
+
"step": 41000
|
| 3198 |
+
},
|
| 3199 |
+
{
|
| 3200 |
+
"epoch": 42.31,
|
| 3201 |
+
"eval_loss": 0.1535973846912384,
|
| 3202 |
+
"eval_runtime": 466.2489,
|
| 3203 |
+
"eval_samples_per_second": 22.471,
|
| 3204 |
+
"eval_steps_per_second": 2.81,
|
| 3205 |
+
"eval_wer": 0.12093904621539109,
|
| 3206 |
+
"step": 41000
|
| 3207 |
+
},
|
| 3208 |
+
{
|
| 3209 |
+
"epoch": 42.41,
|
| 3210 |
+
"learning_rate": 7.679874869655892e-06,
|
| 3211 |
+
"loss": 0.5965,
|
| 3212 |
+
"step": 41100
|
| 3213 |
+
},
|
| 3214 |
+
{
|
| 3215 |
+
"epoch": 42.52,
|
| 3216 |
+
"learning_rate": 7.575599582898853e-06,
|
| 3217 |
+
"loss": 0.5773,
|
| 3218 |
+
"step": 41200
|
| 3219 |
+
},
|
| 3220 |
+
{
|
| 3221 |
+
"epoch": 42.62,
|
| 3222 |
+
"learning_rate": 7.472367049009385e-06,
|
| 3223 |
+
"loss": 0.5844,
|
| 3224 |
+
"step": 41300
|
| 3225 |
+
},
|
| 3226 |
+
{
|
| 3227 |
+
"epoch": 42.72,
|
| 3228 |
+
"learning_rate": 7.3680917622523466e-06,
|
| 3229 |
+
"loss": 0.5906,
|
| 3230 |
+
"step": 41400
|
| 3231 |
+
},
|
| 3232 |
+
{
|
| 3233 |
+
"epoch": 42.83,
|
| 3234 |
+
"learning_rate": 7.263816475495308e-06,
|
| 3235 |
+
"loss": 0.5876,
|
| 3236 |
+
"step": 41500
|
| 3237 |
+
},
|
| 3238 |
+
{
|
| 3239 |
+
"epoch": 42.83,
|
| 3240 |
+
"eval_loss": 0.15671317279338837,
|
| 3241 |
+
"eval_runtime": 479.4414,
|
| 3242 |
+
"eval_samples_per_second": 21.853,
|
| 3243 |
+
"eval_steps_per_second": 2.732,
|
| 3244 |
+
"eval_wer": 0.12110748499842089,
|
| 3245 |
+
"step": 41500
|
| 3246 |
+
},
|
| 3247 |
+
{
|
| 3248 |
+
"epoch": 42.93,
|
| 3249 |
+
"learning_rate": 7.159541188738269e-06,
|
| 3250 |
+
"loss": 0.5746,
|
| 3251 |
+
"step": 41600
|
| 3252 |
+
},
|
| 3253 |
+
{
|
| 3254 |
+
"epoch": 43.03,
|
| 3255 |
+
"learning_rate": 7.055265901981231e-06,
|
| 3256 |
+
"loss": 0.5863,
|
| 3257 |
+
"step": 41700
|
| 3258 |
+
},
|
| 3259 |
+
{
|
| 3260 |
+
"epoch": 43.14,
|
| 3261 |
+
"learning_rate": 6.950990615224193e-06,
|
| 3262 |
+
"loss": 0.5768,
|
| 3263 |
+
"step": 41800
|
| 3264 |
+
},
|
| 3265 |
+
{
|
| 3266 |
+
"epoch": 43.24,
|
| 3267 |
+
"learning_rate": 6.8467153284671545e-06,
|
| 3268 |
+
"loss": 0.5852,
|
| 3269 |
+
"step": 41900
|
| 3270 |
+
},
|
| 3271 |
+
{
|
| 3272 |
+
"epoch": 43.34,
|
| 3273 |
+
"learning_rate": 6.742440041710114e-06,
|
| 3274 |
+
"loss": 0.5714,
|
| 3275 |
+
"step": 42000
|
| 3276 |
+
},
|
| 3277 |
+
{
|
| 3278 |
+
"epoch": 43.34,
|
| 3279 |
+
"eval_loss": 0.15320713818073273,
|
| 3280 |
+
"eval_runtime": 486.9148,
|
| 3281 |
+
"eval_samples_per_second": 21.517,
|
| 3282 |
+
"eval_steps_per_second": 2.69,
|
| 3283 |
+
"eval_wer": 0.12170754816296453,
|
| 3284 |
+
"step": 42000
|
| 3285 |
+
},
|
| 3286 |
+
{
|
| 3287 |
+
"epoch": 43.45,
|
| 3288 |
+
"learning_rate": 6.638164754953076e-06,
|
| 3289 |
+
"loss": 0.5857,
|
| 3290 |
+
"step": 42100
|
| 3291 |
+
},
|
| 3292 |
+
{
|
| 3293 |
+
"epoch": 43.55,
|
| 3294 |
+
"learning_rate": 6.5349322210636086e-06,
|
| 3295 |
+
"loss": 0.592,
|
| 3296 |
+
"step": 42200
|
| 3297 |
+
},
|
| 3298 |
+
{
|
| 3299 |
+
"epoch": 43.65,
|
| 3300 |
+
"learning_rate": 6.43065693430657e-06,
|
| 3301 |
+
"loss": 0.5853,
|
| 3302 |
+
"step": 42300
|
| 3303 |
+
},
|
| 3304 |
+
{
|
| 3305 |
+
"epoch": 43.76,
|
| 3306 |
+
"learning_rate": 6.326381647549531e-06,
|
| 3307 |
+
"loss": 0.5834,
|
| 3308 |
+
"step": 42400
|
| 3309 |
+
},
|
| 3310 |
+
{
|
| 3311 |
+
"epoch": 43.86,
|
| 3312 |
+
"learning_rate": 6.222106360792492e-06,
|
| 3313 |
+
"loss": 0.5756,
|
| 3314 |
+
"step": 42500
|
| 3315 |
+
},
|
| 3316 |
+
{
|
| 3317 |
+
"epoch": 43.86,
|
| 3318 |
+
"eval_loss": 0.15160232782363892,
|
| 3319 |
+
"eval_runtime": 477.2403,
|
| 3320 |
+
"eval_samples_per_second": 21.953,
|
| 3321 |
+
"eval_steps_per_second": 2.745,
|
| 3322 |
+
"eval_wer": 0.11955995367933467,
|
| 3323 |
+
"step": 42500
|
| 3324 |
+
},
|
| 3325 |
+
{
|
| 3326 |
+
"epoch": 43.96,
|
| 3327 |
+
"learning_rate": 6.117831074035453e-06,
|
| 3328 |
+
"loss": 0.5835,
|
| 3329 |
+
"step": 42600
|
| 3330 |
+
},
|
| 3331 |
+
{
|
| 3332 |
+
"epoch": 44.07,
|
| 3333 |
+
"learning_rate": 6.013555787278416e-06,
|
| 3334 |
+
"loss": 0.5761,
|
| 3335 |
+
"step": 42700
|
| 3336 |
+
},
|
| 3337 |
+
{
|
| 3338 |
+
"epoch": 44.17,
|
| 3339 |
+
"learning_rate": 5.909280500521377e-06,
|
| 3340 |
+
"loss": 0.5703,
|
| 3341 |
+
"step": 42800
|
| 3342 |
+
},
|
| 3343 |
+
{
|
| 3344 |
+
"epoch": 44.27,
|
| 3345 |
+
"learning_rate": 5.8050052137643385e-06,
|
| 3346 |
+
"loss": 0.5838,
|
| 3347 |
+
"step": 42900
|
| 3348 |
+
},
|
| 3349 |
+
{
|
| 3350 |
+
"epoch": 44.38,
|
| 3351 |
+
"learning_rate": 5.7007299270073e-06,
|
| 3352 |
+
"loss": 0.5719,
|
| 3353 |
+
"step": 43000
|
| 3354 |
+
},
|
| 3355 |
+
{
|
| 3356 |
+
"epoch": 44.38,
|
| 3357 |
+
"eval_loss": 0.14908714592456818,
|
| 3358 |
+
"eval_runtime": 478.7642,
|
| 3359 |
+
"eval_samples_per_second": 21.883,
|
| 3360 |
+
"eval_steps_per_second": 2.736,
|
| 3361 |
+
"eval_wer": 0.11914938414569955,
|
| 3362 |
+
"step": 43000
|
| 3363 |
+
},
|
| 3364 |
+
{
|
| 3365 |
+
"epoch": 44.48,
|
| 3366 |
+
"learning_rate": 5.5964546402502604e-06,
|
| 3367 |
+
"loss": 0.5849,
|
| 3368 |
+
"step": 43100
|
| 3369 |
+
},
|
| 3370 |
+
{
|
| 3371 |
+
"epoch": 44.58,
|
| 3372 |
+
"learning_rate": 5.492179353493222e-06,
|
| 3373 |
+
"loss": 0.5731,
|
| 3374 |
+
"step": 43200
|
| 3375 |
+
},
|
| 3376 |
+
{
|
| 3377 |
+
"epoch": 44.69,
|
| 3378 |
+
"learning_rate": 5.387904066736184e-06,
|
| 3379 |
+
"loss": 0.5722,
|
| 3380 |
+
"step": 43300
|
| 3381 |
+
},
|
| 3382 |
+
{
|
| 3383 |
+
"epoch": 44.79,
|
| 3384 |
+
"learning_rate": 5.2836287799791455e-06,
|
| 3385 |
+
"loss": 0.578,
|
| 3386 |
+
"step": 43400
|
| 3387 |
+
},
|
| 3388 |
+
{
|
| 3389 |
+
"epoch": 44.89,
|
| 3390 |
+
"learning_rate": 5.179353493222107e-06,
|
| 3391 |
+
"loss": 0.5829,
|
| 3392 |
+
"step": 43500
|
| 3393 |
+
},
|
| 3394 |
+
{
|
| 3395 |
+
"epoch": 44.89,
|
| 3396 |
+
"eval_loss": 0.1497020274400711,
|
| 3397 |
+
"eval_runtime": 463.7767,
|
| 3398 |
+
"eval_samples_per_second": 22.591,
|
| 3399 |
+
"eval_steps_per_second": 2.825,
|
| 3400 |
+
"eval_wer": 0.11930729550478998,
|
| 3401 |
+
"step": 43500
|
| 3402 |
+
},
|
| 3403 |
+
{
|
| 3404 |
+
"epoch": 44.99,
|
| 3405 |
+
"learning_rate": 5.0750782064650675e-06,
|
| 3406 |
+
"loss": 0.5936,
|
| 3407 |
+
"step": 43600
|
| 3408 |
+
},
|
| 3409 |
+
{
|
| 3410 |
+
"epoch": 45.1,
|
| 3411 |
+
"learning_rate": 4.970802919708029e-06,
|
| 3412 |
+
"loss": 0.5714,
|
| 3413 |
+
"step": 43700
|
| 3414 |
+
},
|
| 3415 |
+
{
|
| 3416 |
+
"epoch": 45.2,
|
| 3417 |
+
"learning_rate": 4.866527632950991e-06,
|
| 3418 |
+
"loss": 0.5593,
|
| 3419 |
+
"step": 43800
|
| 3420 |
+
},
|
| 3421 |
+
{
|
| 3422 |
+
"epoch": 45.3,
|
| 3423 |
+
"learning_rate": 4.762252346193953e-06,
|
| 3424 |
+
"loss": 0.5766,
|
| 3425 |
+
"step": 43900
|
| 3426 |
+
},
|
| 3427 |
+
{
|
| 3428 |
+
"epoch": 45.41,
|
| 3429 |
+
"learning_rate": 4.657977059436914e-06,
|
| 3430 |
+
"loss": 0.5664,
|
| 3431 |
+
"step": 44000
|
| 3432 |
+
},
|
| 3433 |
+
{
|
| 3434 |
+
"epoch": 45.41,
|
| 3435 |
+
"eval_loss": 0.14872093498706818,
|
| 3436 |
+
"eval_runtime": 471.3612,
|
| 3437 |
+
"eval_samples_per_second": 22.227,
|
| 3438 |
+
"eval_steps_per_second": 2.779,
|
| 3439 |
+
"eval_wer": 0.11728603010843247,
|
| 3440 |
+
"step": 44000
|
| 3441 |
+
},
|
| 3442 |
+
{
|
| 3443 |
+
"epoch": 45.51,
|
| 3444 |
+
"learning_rate": 4.553701772679875e-06,
|
| 3445 |
+
"loss": 0.5697,
|
| 3446 |
+
"step": 44100
|
| 3447 |
+
},
|
| 3448 |
+
{
|
| 3449 |
+
"epoch": 45.61,
|
| 3450 |
+
"learning_rate": 4.449426485922836e-06,
|
| 3451 |
+
"loss": 0.5718,
|
| 3452 |
+
"step": 44200
|
| 3453 |
+
},
|
| 3454 |
+
{
|
| 3455 |
+
"epoch": 45.72,
|
| 3456 |
+
"learning_rate": 4.345151199165797e-06,
|
| 3457 |
+
"loss": 0.5744,
|
| 3458 |
+
"step": 44300
|
| 3459 |
+
},
|
| 3460 |
+
{
|
| 3461 |
+
"epoch": 45.82,
|
| 3462 |
+
"learning_rate": 4.24087591240876e-06,
|
| 3463 |
+
"loss": 0.5672,
|
| 3464 |
+
"step": 44400
|
| 3465 |
+
},
|
| 3466 |
+
{
|
| 3467 |
+
"epoch": 45.92,
|
| 3468 |
+
"learning_rate": 4.137643378519291e-06,
|
| 3469 |
+
"loss": 0.5707,
|
| 3470 |
+
"step": 44500
|
| 3471 |
+
},
|
| 3472 |
+
{
|
| 3473 |
+
"epoch": 45.92,
|
| 3474 |
+
"eval_loss": 0.1470126211643219,
|
| 3475 |
+
"eval_runtime": 478.6316,
|
| 3476 |
+
"eval_samples_per_second": 21.889,
|
| 3477 |
+
"eval_steps_per_second": 2.737,
|
| 3478 |
+
"eval_wer": 0.11641225392146541,
|
| 3479 |
+
"step": 44500
|
| 3480 |
+
},
|
| 3481 |
+
{
|
| 3482 |
+
"epoch": 46.03,
|
| 3483 |
+
"learning_rate": 4.033368091762252e-06,
|
| 3484 |
+
"loss": 0.5555,
|
| 3485 |
+
"step": 44600
|
| 3486 |
+
},
|
| 3487 |
+
{
|
| 3488 |
+
"epoch": 46.13,
|
| 3489 |
+
"learning_rate": 3.929092805005214e-06,
|
| 3490 |
+
"loss": 0.5699,
|
| 3491 |
+
"step": 44700
|
| 3492 |
+
},
|
| 3493 |
+
{
|
| 3494 |
+
"epoch": 46.23,
|
| 3495 |
+
"learning_rate": 3.824817518248176e-06,
|
| 3496 |
+
"loss": 0.573,
|
| 3497 |
+
"step": 44800
|
| 3498 |
+
},
|
| 3499 |
+
{
|
| 3500 |
+
"epoch": 46.34,
|
| 3501 |
+
"learning_rate": 3.7205422314911366e-06,
|
| 3502 |
+
"loss": 0.5697,
|
| 3503 |
+
"step": 44900
|
| 3504 |
+
},
|
| 3505 |
+
{
|
| 3506 |
+
"epoch": 46.44,
|
| 3507 |
+
"learning_rate": 3.616266944734098e-06,
|
| 3508 |
+
"loss": 0.5696,
|
| 3509 |
+
"step": 45000
|
| 3510 |
+
},
|
| 3511 |
+
{
|
| 3512 |
+
"epoch": 46.44,
|
| 3513 |
+
"eval_loss": 0.14786458015441895,
|
| 3514 |
+
"eval_runtime": 467.8513,
|
| 3515 |
+
"eval_samples_per_second": 22.394,
|
| 3516 |
+
"eval_steps_per_second": 2.8,
|
| 3517 |
+
"eval_wer": 0.11606484893146647,
|
| 3518 |
+
"step": 45000
|
| 3519 |
+
},
|
| 3520 |
+
{
|
| 3521 |
+
"epoch": 46.54,
|
| 3522 |
+
"learning_rate": 3.5119916579770594e-06,
|
| 3523 |
+
"loss": 0.5634,
|
| 3524 |
+
"step": 45100
|
| 3525 |
+
},
|
| 3526 |
+
{
|
| 3527 |
+
"epoch": 46.65,
|
| 3528 |
+
"learning_rate": 3.4077163712200212e-06,
|
| 3529 |
+
"loss": 0.5742,
|
| 3530 |
+
"step": 45200
|
| 3531 |
+
},
|
| 3532 |
+
{
|
| 3533 |
+
"epoch": 46.75,
|
| 3534 |
+
"learning_rate": 3.3044838373305525e-06,
|
| 3535 |
+
"loss": 0.5612,
|
| 3536 |
+
"step": 45300
|
| 3537 |
+
},
|
| 3538 |
+
{
|
| 3539 |
+
"epoch": 46.85,
|
| 3540 |
+
"learning_rate": 3.2002085505735143e-06,
|
| 3541 |
+
"loss": 0.5548,
|
| 3542 |
+
"step": 45400
|
| 3543 |
+
},
|
| 3544 |
+
{
|
| 3545 |
+
"epoch": 46.96,
|
| 3546 |
+
"learning_rate": 3.0959332638164757e-06,
|
| 3547 |
+
"loss": 0.5767,
|
| 3548 |
+
"step": 45500
|
| 3549 |
+
},
|
| 3550 |
+
{
|
| 3551 |
+
"epoch": 46.96,
|
| 3552 |
+
"eval_loss": 0.1492297500371933,
|
| 3553 |
+
"eval_runtime": 465.8572,
|
| 3554 |
+
"eval_samples_per_second": 22.49,
|
| 3555 |
+
"eval_steps_per_second": 2.812,
|
| 3556 |
+
"eval_wer": 0.11754921570691652,
|
| 3557 |
+
"step": 45500
|
| 3558 |
+
},
|
| 3559 |
+
{
|
| 3560 |
+
"epoch": 47.06,
|
| 3561 |
+
"learning_rate": 2.991657977059437e-06,
|
| 3562 |
+
"loss": 0.5831,
|
| 3563 |
+
"step": 45600
|
| 3564 |
+
},
|
| 3565 |
+
{
|
| 3566 |
+
"epoch": 47.16,
|
| 3567 |
+
"learning_rate": 2.8873826903023986e-06,
|
| 3568 |
+
"loss": 0.5653,
|
| 3569 |
+
"step": 45700
|
| 3570 |
+
},
|
| 3571 |
+
{
|
| 3572 |
+
"epoch": 47.27,
|
| 3573 |
+
"learning_rate": 2.78310740354536e-06,
|
| 3574 |
+
"loss": 0.5744,
|
| 3575 |
+
"step": 45800
|
| 3576 |
+
},
|
| 3577 |
+
{
|
| 3578 |
+
"epoch": 47.37,
|
| 3579 |
+
"learning_rate": 2.6788321167883214e-06,
|
| 3580 |
+
"loss": 0.5587,
|
| 3581 |
+
"step": 45900
|
| 3582 |
+
},
|
| 3583 |
+
{
|
| 3584 |
+
"epoch": 47.47,
|
| 3585 |
+
"learning_rate": 2.574556830031283e-06,
|
| 3586 |
+
"loss": 0.5573,
|
| 3587 |
+
"step": 46000
|
| 3588 |
+
},
|
| 3589 |
+
{
|
| 3590 |
+
"epoch": 47.47,
|
| 3591 |
+
"eval_loss": 0.14706237614154816,
|
| 3592 |
+
"eval_runtime": 467.4225,
|
| 3593 |
+
"eval_samples_per_second": 22.414,
|
| 3594 |
+
"eval_steps_per_second": 2.803,
|
| 3595 |
+
"eval_wer": 0.1164754184651016,
|
| 3596 |
+
"step": 46000
|
| 3597 |
+
}
|
| 3598 |
+
],
|
| 3599 |
+
"max_steps": 48450,
|
| 3600 |
+
"num_train_epochs": 50,
|
| 3601 |
+
"total_flos": 5.920193480786585e+20,
|
| 3602 |
+
"trial_name": null,
|
| 3603 |
+
"trial_params": null
|
| 3604 |
+
}
|