Training in progress, step 200, checkpoint
Browse files
last-checkpoint/adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 639691872
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:262adc43b1549747f8133c78c1f3cd45fcd802b97965c2e197dc99a7ceaf6d3b
|
| 3 |
size 639691872
|
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1279529146
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d21352c173c6a408300739dc463c59f3dba7ad1fb915511917540a408a158f27
|
| 3 |
size 1279529146
|
last-checkpoint/scheduler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:50a635eef07ba685f548ffceaa8f54b55dd3ba9de5e267cb2673b0a5d4d9b18e
|
| 3 |
size 1064
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 0.
|
| 5 |
"eval_steps": 500,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
@@ -1507,6 +1507,1506 @@
|
|
| 1507 |
"rewards/margins": 0.3506268858909607,
|
| 1508 |
"rewards/rejected": -0.010964250192046165,
|
| 1509 |
"step": 100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1510 |
}
|
| 1511 |
],
|
| 1512 |
"logging_steps": 1,
|
|
@@ -1521,7 +3021,7 @@
|
|
| 1521 |
"should_evaluate": false,
|
| 1522 |
"should_log": false,
|
| 1523 |
"should_save": true,
|
| 1524 |
-
"should_training_stop":
|
| 1525 |
},
|
| 1526 |
"attributes": {}
|
| 1527 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.05004222312576236,
|
| 5 |
"eval_steps": 500,
|
| 6 |
+
"global_step": 200,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 1507 |
"rewards/margins": 0.3506268858909607,
|
| 1508 |
"rewards/rejected": -0.010964250192046165,
|
| 1509 |
"step": 100
|
| 1510 |
+
},
|
| 1511 |
+
{
|
| 1512 |
+
"epoch": 0.025271322678509994,
|
| 1513 |
+
"grad_norm": 168.01998901367188,
|
| 1514 |
+
"learning_rate": 2.5250000000000004e-06,
|
| 1515 |
+
"logits/chosen": 0.5338684320449829,
|
| 1516 |
+
"logits/rejected": 0.5581062436103821,
|
| 1517 |
+
"logps/chosen": -830.5277099609375,
|
| 1518 |
+
"logps/rejected": -750.9749145507812,
|
| 1519 |
+
"loss": 5.1281,
|
| 1520 |
+
"rewards/accuracies": 0.6875,
|
| 1521 |
+
"rewards/chosen": 0.29842013120651245,
|
| 1522 |
+
"rewards/margins": 0.12503623962402344,
|
| 1523 |
+
"rewards/rejected": 0.1733839064836502,
|
| 1524 |
+
"step": 101
|
| 1525 |
+
},
|
| 1526 |
+
{
|
| 1527 |
+
"epoch": 0.025521533794138806,
|
| 1528 |
+
"grad_norm": 195.6433868408203,
|
| 1529 |
+
"learning_rate": 2.55e-06,
|
| 1530 |
+
"logits/chosen": 0.5473726987838745,
|
| 1531 |
+
"logits/rejected": 0.2889501452445984,
|
| 1532 |
+
"logps/chosen": -904.6567993164062,
|
| 1533 |
+
"logps/rejected": -989.4647216796875,
|
| 1534 |
+
"loss": 4.9998,
|
| 1535 |
+
"rewards/accuracies": 0.75,
|
| 1536 |
+
"rewards/chosen": 0.41574910283088684,
|
| 1537 |
+
"rewards/margins": 0.19607900083065033,
|
| 1538 |
+
"rewards/rejected": 0.21967005729675293,
|
| 1539 |
+
"step": 102
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 0.025771744909767618,
|
| 1543 |
+
"grad_norm": 170.56797790527344,
|
| 1544 |
+
"learning_rate": 2.5750000000000003e-06,
|
| 1545 |
+
"logits/chosen": 0.8648879528045654,
|
| 1546 |
+
"logits/rejected": 0.5083369612693787,
|
| 1547 |
+
"logps/chosen": -743.2974243164062,
|
| 1548 |
+
"logps/rejected": -946.9625244140625,
|
| 1549 |
+
"loss": 5.1817,
|
| 1550 |
+
"rewards/accuracies": 0.6875,
|
| 1551 |
+
"rewards/chosen": 0.336568146944046,
|
| 1552 |
+
"rewards/margins": 0.1252390742301941,
|
| 1553 |
+
"rewards/rejected": 0.21132907271385193,
|
| 1554 |
+
"step": 103
|
| 1555 |
+
},
|
| 1556 |
+
{
|
| 1557 |
+
"epoch": 0.026021956025396427,
|
| 1558 |
+
"grad_norm": 179.4681854248047,
|
| 1559 |
+
"learning_rate": 2.6e-06,
|
| 1560 |
+
"logits/chosen": 0.6028002500534058,
|
| 1561 |
+
"logits/rejected": 0.6547164916992188,
|
| 1562 |
+
"logps/chosen": -834.4608154296875,
|
| 1563 |
+
"logps/rejected": -800.7316284179688,
|
| 1564 |
+
"loss": 5.4172,
|
| 1565 |
+
"rewards/accuracies": 0.5,
|
| 1566 |
+
"rewards/chosen": 0.5305736660957336,
|
| 1567 |
+
"rewards/margins": 0.05496921390295029,
|
| 1568 |
+
"rewards/rejected": 0.47560444474220276,
|
| 1569 |
+
"step": 104
|
| 1570 |
+
},
|
| 1571 |
+
{
|
| 1572 |
+
"epoch": 0.02627216714102524,
|
| 1573 |
+
"grad_norm": 177.14212036132812,
|
| 1574 |
+
"learning_rate": 2.6250000000000003e-06,
|
| 1575 |
+
"logits/chosen": 0.633183479309082,
|
| 1576 |
+
"logits/rejected": 0.49183252453804016,
|
| 1577 |
+
"logps/chosen": -769.5543823242188,
|
| 1578 |
+
"logps/rejected": -841.36181640625,
|
| 1579 |
+
"loss": 5.0401,
|
| 1580 |
+
"rewards/accuracies": 0.6875,
|
| 1581 |
+
"rewards/chosen": 0.3643644452095032,
|
| 1582 |
+
"rewards/margins": 0.16701143980026245,
|
| 1583 |
+
"rewards/rejected": 0.1973530352115631,
|
| 1584 |
+
"step": 105
|
| 1585 |
+
},
|
| 1586 |
+
{
|
| 1587 |
+
"epoch": 0.02652237825665405,
|
| 1588 |
+
"grad_norm": 183.1171112060547,
|
| 1589 |
+
"learning_rate": 2.6500000000000005e-06,
|
| 1590 |
+
"logits/chosen": 0.7262644171714783,
|
| 1591 |
+
"logits/rejected": 0.7168549299240112,
|
| 1592 |
+
"logps/chosen": -836.1957397460938,
|
| 1593 |
+
"logps/rejected": -949.1402587890625,
|
| 1594 |
+
"loss": 5.3246,
|
| 1595 |
+
"rewards/accuracies": 0.5,
|
| 1596 |
+
"rewards/chosen": 0.4721420407295227,
|
| 1597 |
+
"rewards/margins": 0.14782388508319855,
|
| 1598 |
+
"rewards/rejected": 0.32431814074516296,
|
| 1599 |
+
"step": 106
|
| 1600 |
+
},
|
| 1601 |
+
{
|
| 1602 |
+
"epoch": 0.026772589372282864,
|
| 1603 |
+
"grad_norm": 183.04190063476562,
|
| 1604 |
+
"learning_rate": 2.6750000000000002e-06,
|
| 1605 |
+
"logits/chosen": 0.45039886236190796,
|
| 1606 |
+
"logits/rejected": 0.23830902576446533,
|
| 1607 |
+
"logps/chosen": -868.3818359375,
|
| 1608 |
+
"logps/rejected": -880.0626220703125,
|
| 1609 |
+
"loss": 5.1669,
|
| 1610 |
+
"rewards/accuracies": 0.625,
|
| 1611 |
+
"rewards/chosen": 0.38201138377189636,
|
| 1612 |
+
"rewards/margins": 0.16360534727573395,
|
| 1613 |
+
"rewards/rejected": 0.2184060662984848,
|
| 1614 |
+
"step": 107
|
| 1615 |
+
},
|
| 1616 |
+
{
|
| 1617 |
+
"epoch": 0.027022800487911676,
|
| 1618 |
+
"grad_norm": 166.98008728027344,
|
| 1619 |
+
"learning_rate": 2.7000000000000004e-06,
|
| 1620 |
+
"logits/chosen": 0.5699177980422974,
|
| 1621 |
+
"logits/rejected": 0.5048665404319763,
|
| 1622 |
+
"logps/chosen": -841.4978637695312,
|
| 1623 |
+
"logps/rejected": -836.6079711914062,
|
| 1624 |
+
"loss": 4.5338,
|
| 1625 |
+
"rewards/accuracies": 0.8125,
|
| 1626 |
+
"rewards/chosen": 0.591092586517334,
|
| 1627 |
+
"rewards/margins": 0.3148172199726105,
|
| 1628 |
+
"rewards/rejected": 0.2762753963470459,
|
| 1629 |
+
"step": 108
|
| 1630 |
+
},
|
| 1631 |
+
{
|
| 1632 |
+
"epoch": 0.027273011603540488,
|
| 1633 |
+
"grad_norm": 174.3636474609375,
|
| 1634 |
+
"learning_rate": 2.7250000000000006e-06,
|
| 1635 |
+
"logits/chosen": 0.6328776478767395,
|
| 1636 |
+
"logits/rejected": 0.506734311580658,
|
| 1637 |
+
"logps/chosen": -1035.760009765625,
|
| 1638 |
+
"logps/rejected": -1037.7685546875,
|
| 1639 |
+
"loss": 4.5938,
|
| 1640 |
+
"rewards/accuracies": 0.8125,
|
| 1641 |
+
"rewards/chosen": 0.5912416577339172,
|
| 1642 |
+
"rewards/margins": 0.30340731143951416,
|
| 1643 |
+
"rewards/rejected": 0.2878343462944031,
|
| 1644 |
+
"step": 109
|
| 1645 |
+
},
|
| 1646 |
+
{
|
| 1647 |
+
"epoch": 0.0275232227191693,
|
| 1648 |
+
"grad_norm": 172.23876953125,
|
| 1649 |
+
"learning_rate": 2.7500000000000004e-06,
|
| 1650 |
+
"logits/chosen": 0.4434925317764282,
|
| 1651 |
+
"logits/rejected": 0.6842514872550964,
|
| 1652 |
+
"logps/chosen": -983.0321044921875,
|
| 1653 |
+
"logps/rejected": -788.3206787109375,
|
| 1654 |
+
"loss": 4.999,
|
| 1655 |
+
"rewards/accuracies": 0.625,
|
| 1656 |
+
"rewards/chosen": 0.3090551793575287,
|
| 1657 |
+
"rewards/margins": 0.17981156706809998,
|
| 1658 |
+
"rewards/rejected": 0.1292436271905899,
|
| 1659 |
+
"step": 110
|
| 1660 |
+
},
|
| 1661 |
+
{
|
| 1662 |
+
"epoch": 0.02777343383479811,
|
| 1663 |
+
"grad_norm": 172.0596466064453,
|
| 1664 |
+
"learning_rate": 2.7750000000000005e-06,
|
| 1665 |
+
"logits/chosen": 0.4155680537223816,
|
| 1666 |
+
"logits/rejected": 0.46463680267333984,
|
| 1667 |
+
"logps/chosen": -881.0513305664062,
|
| 1668 |
+
"logps/rejected": -925.1255493164062,
|
| 1669 |
+
"loss": 4.7598,
|
| 1670 |
+
"rewards/accuracies": 0.625,
|
| 1671 |
+
"rewards/chosen": 0.6203694939613342,
|
| 1672 |
+
"rewards/margins": 0.26839369535446167,
|
| 1673 |
+
"rewards/rejected": 0.35197576880455017,
|
| 1674 |
+
"step": 111
|
| 1675 |
+
},
|
| 1676 |
+
{
|
| 1677 |
+
"epoch": 0.02802364495042692,
|
| 1678 |
+
"grad_norm": 195.08697509765625,
|
| 1679 |
+
"learning_rate": 2.8000000000000003e-06,
|
| 1680 |
+
"logits/chosen": 0.605146586894989,
|
| 1681 |
+
"logits/rejected": 0.7462798357009888,
|
| 1682 |
+
"logps/chosen": -970.7579345703125,
|
| 1683 |
+
"logps/rejected": -969.8851318359375,
|
| 1684 |
+
"loss": 5.0111,
|
| 1685 |
+
"rewards/accuracies": 0.625,
|
| 1686 |
+
"rewards/chosen": 0.3882797658443451,
|
| 1687 |
+
"rewards/margins": 0.22670859098434448,
|
| 1688 |
+
"rewards/rejected": 0.1615711748600006,
|
| 1689 |
+
"step": 112
|
| 1690 |
+
},
|
| 1691 |
+
{
|
| 1692 |
+
"epoch": 0.028273856066055734,
|
| 1693 |
+
"grad_norm": 157.1236114501953,
|
| 1694 |
+
"learning_rate": 2.825e-06,
|
| 1695 |
+
"logits/chosen": 0.6366948485374451,
|
| 1696 |
+
"logits/rejected": 0.717607855796814,
|
| 1697 |
+
"logps/chosen": -820.1522827148438,
|
| 1698 |
+
"logps/rejected": -789.8685913085938,
|
| 1699 |
+
"loss": 5.0085,
|
| 1700 |
+
"rewards/accuracies": 0.625,
|
| 1701 |
+
"rewards/chosen": 0.4857552647590637,
|
| 1702 |
+
"rewards/margins": 0.18623466789722443,
|
| 1703 |
+
"rewards/rejected": 0.2995206117630005,
|
| 1704 |
+
"step": 113
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"epoch": 0.028524067181684546,
|
| 1708 |
+
"grad_norm": 198.33815002441406,
|
| 1709 |
+
"learning_rate": 2.85e-06,
|
| 1710 |
+
"logits/chosen": 0.7510839700698853,
|
| 1711 |
+
"logits/rejected": 0.6683776378631592,
|
| 1712 |
+
"logps/chosen": -951.9930419921875,
|
| 1713 |
+
"logps/rejected": -911.171142578125,
|
| 1714 |
+
"loss": 5.9293,
|
| 1715 |
+
"rewards/accuracies": 0.4375,
|
| 1716 |
+
"rewards/chosen": 0.3041432499885559,
|
| 1717 |
+
"rewards/margins": -0.05183648690581322,
|
| 1718 |
+
"rewards/rejected": 0.3559797406196594,
|
| 1719 |
+
"step": 114
|
| 1720 |
+
},
|
| 1721 |
+
{
|
| 1722 |
+
"epoch": 0.02877427829731336,
|
| 1723 |
+
"grad_norm": 156.49578857421875,
|
| 1724 |
+
"learning_rate": 2.875e-06,
|
| 1725 |
+
"logits/chosen": 0.3662444055080414,
|
| 1726 |
+
"logits/rejected": 0.35427069664001465,
|
| 1727 |
+
"logps/chosen": -725.6800537109375,
|
| 1728 |
+
"logps/rejected": -791.179931640625,
|
| 1729 |
+
"loss": 4.5455,
|
| 1730 |
+
"rewards/accuracies": 0.75,
|
| 1731 |
+
"rewards/chosen": 0.3427468240261078,
|
| 1732 |
+
"rewards/margins": 0.31882816553115845,
|
| 1733 |
+
"rewards/rejected": 0.023918677121400833,
|
| 1734 |
+
"step": 115
|
| 1735 |
+
},
|
| 1736 |
+
{
|
| 1737 |
+
"epoch": 0.02902448941294217,
|
| 1738 |
+
"grad_norm": 162.18934631347656,
|
| 1739 |
+
"learning_rate": 2.9e-06,
|
| 1740 |
+
"logits/chosen": 0.568724513053894,
|
| 1741 |
+
"logits/rejected": 0.4400956928730011,
|
| 1742 |
+
"logps/chosen": -718.4844360351562,
|
| 1743 |
+
"logps/rejected": -824.3160400390625,
|
| 1744 |
+
"loss": 4.7706,
|
| 1745 |
+
"rewards/accuracies": 0.625,
|
| 1746 |
+
"rewards/chosen": 0.5042533874511719,
|
| 1747 |
+
"rewards/margins": 0.26497164368629456,
|
| 1748 |
+
"rewards/rejected": 0.2392817586660385,
|
| 1749 |
+
"step": 116
|
| 1750 |
+
},
|
| 1751 |
+
{
|
| 1752 |
+
"epoch": 0.029274700528570983,
|
| 1753 |
+
"grad_norm": 179.54246520996094,
|
| 1754 |
+
"learning_rate": 2.925e-06,
|
| 1755 |
+
"logits/chosen": 0.5801786184310913,
|
| 1756 |
+
"logits/rejected": 0.549165666103363,
|
| 1757 |
+
"logps/chosen": -898.4237060546875,
|
| 1758 |
+
"logps/rejected": -969.0765380859375,
|
| 1759 |
+
"loss": 4.3671,
|
| 1760 |
+
"rewards/accuracies": 0.75,
|
| 1761 |
+
"rewards/chosen": 0.66265469789505,
|
| 1762 |
+
"rewards/margins": 0.3924952745437622,
|
| 1763 |
+
"rewards/rejected": 0.27015945315361023,
|
| 1764 |
+
"step": 117
|
| 1765 |
+
},
|
| 1766 |
+
{
|
| 1767 |
+
"epoch": 0.029524911644199795,
|
| 1768 |
+
"grad_norm": 181.2903289794922,
|
| 1769 |
+
"learning_rate": 2.95e-06,
|
| 1770 |
+
"logits/chosen": 0.7072390913963318,
|
| 1771 |
+
"logits/rejected": 0.599428653717041,
|
| 1772 |
+
"logps/chosen": -873.1541748046875,
|
| 1773 |
+
"logps/rejected": -932.2781982421875,
|
| 1774 |
+
"loss": 5.0481,
|
| 1775 |
+
"rewards/accuracies": 0.6875,
|
| 1776 |
+
"rewards/chosen": 0.516876220703125,
|
| 1777 |
+
"rewards/margins": 0.1785745620727539,
|
| 1778 |
+
"rewards/rejected": 0.3383016586303711,
|
| 1779 |
+
"step": 118
|
| 1780 |
+
},
|
| 1781 |
+
{
|
| 1782 |
+
"epoch": 0.029775122759828604,
|
| 1783 |
+
"grad_norm": 170.22119140625,
|
| 1784 |
+
"learning_rate": 2.9750000000000003e-06,
|
| 1785 |
+
"logits/chosen": 0.9537469744682312,
|
| 1786 |
+
"logits/rejected": 0.8464532494544983,
|
| 1787 |
+
"logps/chosen": -755.2413330078125,
|
| 1788 |
+
"logps/rejected": -743.6602172851562,
|
| 1789 |
+
"loss": 4.8546,
|
| 1790 |
+
"rewards/accuracies": 0.625,
|
| 1791 |
+
"rewards/chosen": 0.5026987791061401,
|
| 1792 |
+
"rewards/margins": 0.2353692501783371,
|
| 1793 |
+
"rewards/rejected": 0.26732951402664185,
|
| 1794 |
+
"step": 119
|
| 1795 |
+
},
|
| 1796 |
+
{
|
| 1797 |
+
"epoch": 0.030025333875457416,
|
| 1798 |
+
"grad_norm": 144.19140625,
|
| 1799 |
+
"learning_rate": 3e-06,
|
| 1800 |
+
"logits/chosen": 0.6080533266067505,
|
| 1801 |
+
"logits/rejected": 0.5463143587112427,
|
| 1802 |
+
"logps/chosen": -680.61181640625,
|
| 1803 |
+
"logps/rejected": -740.6124877929688,
|
| 1804 |
+
"loss": 4.1267,
|
| 1805 |
+
"rewards/accuracies": 0.875,
|
| 1806 |
+
"rewards/chosen": 0.5357052683830261,
|
| 1807 |
+
"rewards/margins": 0.4544171392917633,
|
| 1808 |
+
"rewards/rejected": 0.08128809928894043,
|
| 1809 |
+
"step": 120
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 0.03027554499108623,
|
| 1813 |
+
"grad_norm": 152.67770385742188,
|
| 1814 |
+
"learning_rate": 3.0250000000000003e-06,
|
| 1815 |
+
"logits/chosen": 0.433148592710495,
|
| 1816 |
+
"logits/rejected": 0.657333493232727,
|
| 1817 |
+
"logps/chosen": -992.8467407226562,
|
| 1818 |
+
"logps/rejected": -825.07958984375,
|
| 1819 |
+
"loss": 4.248,
|
| 1820 |
+
"rewards/accuracies": 0.75,
|
| 1821 |
+
"rewards/chosen": 0.5468575358390808,
|
| 1822 |
+
"rewards/margins": 0.4552149772644043,
|
| 1823 |
+
"rewards/rejected": 0.09164253622293472,
|
| 1824 |
+
"step": 121
|
| 1825 |
+
},
|
| 1826 |
+
{
|
| 1827 |
+
"epoch": 0.03052575610671504,
|
| 1828 |
+
"grad_norm": 170.1833038330078,
|
| 1829 |
+
"learning_rate": 3.05e-06,
|
| 1830 |
+
"logits/chosen": 1.1314983367919922,
|
| 1831 |
+
"logits/rejected": 1.0099120140075684,
|
| 1832 |
+
"logps/chosen": -732.5499877929688,
|
| 1833 |
+
"logps/rejected": -709.9703979492188,
|
| 1834 |
+
"loss": 5.4108,
|
| 1835 |
+
"rewards/accuracies": 0.5,
|
| 1836 |
+
"rewards/chosen": 0.35310542583465576,
|
| 1837 |
+
"rewards/margins": 0.10829624533653259,
|
| 1838 |
+
"rewards/rejected": 0.24480919539928436,
|
| 1839 |
+
"step": 122
|
| 1840 |
+
},
|
| 1841 |
+
{
|
| 1842 |
+
"epoch": 0.030775967222343853,
|
| 1843 |
+
"grad_norm": 121.0159912109375,
|
| 1844 |
+
"learning_rate": 3.075e-06,
|
| 1845 |
+
"logits/chosen": 0.9415697455406189,
|
| 1846 |
+
"logits/rejected": 0.8197463154792786,
|
| 1847 |
+
"logps/chosen": -769.3919677734375,
|
| 1848 |
+
"logps/rejected": -631.9843139648438,
|
| 1849 |
+
"loss": 3.956,
|
| 1850 |
+
"rewards/accuracies": 0.875,
|
| 1851 |
+
"rewards/chosen": 0.5862818956375122,
|
| 1852 |
+
"rewards/margins": 0.5186235308647156,
|
| 1853 |
+
"rewards/rejected": 0.0676584541797638,
|
| 1854 |
+
"step": 123
|
| 1855 |
+
},
|
| 1856 |
+
{
|
| 1857 |
+
"epoch": 0.031026178337972665,
|
| 1858 |
+
"grad_norm": 176.4051513671875,
|
| 1859 |
+
"learning_rate": 3.1000000000000004e-06,
|
| 1860 |
+
"logits/chosen": 0.5172820687294006,
|
| 1861 |
+
"logits/rejected": 0.3080348074436188,
|
| 1862 |
+
"logps/chosen": -974.28173828125,
|
| 1863 |
+
"logps/rejected": -859.1981201171875,
|
| 1864 |
+
"loss": 4.7509,
|
| 1865 |
+
"rewards/accuracies": 0.75,
|
| 1866 |
+
"rewards/chosen": 0.5816446542739868,
|
| 1867 |
+
"rewards/margins": 0.2865919768810272,
|
| 1868 |
+
"rewards/rejected": 0.2950526475906372,
|
| 1869 |
+
"step": 124
|
| 1870 |
+
},
|
| 1871 |
+
{
|
| 1872 |
+
"epoch": 0.03127638945360148,
|
| 1873 |
+
"grad_norm": 155.2854461669922,
|
| 1874 |
+
"learning_rate": 3.125e-06,
|
| 1875 |
+
"logits/chosen": 1.1160801649093628,
|
| 1876 |
+
"logits/rejected": 0.9693074226379395,
|
| 1877 |
+
"logps/chosen": -568.8187255859375,
|
| 1878 |
+
"logps/rejected": -709.5980834960938,
|
| 1879 |
+
"loss": 4.9778,
|
| 1880 |
+
"rewards/accuracies": 0.75,
|
| 1881 |
+
"rewards/chosen": 0.40664032101631165,
|
| 1882 |
+
"rewards/margins": 0.20195113122463226,
|
| 1883 |
+
"rewards/rejected": 0.2046891748905182,
|
| 1884 |
+
"step": 125
|
| 1885 |
+
},
|
| 1886 |
+
{
|
| 1887 |
+
"epoch": 0.03152660056923029,
|
| 1888 |
+
"grad_norm": 174.4468536376953,
|
| 1889 |
+
"learning_rate": 3.1500000000000003e-06,
|
| 1890 |
+
"logits/chosen": 0.5777021646499634,
|
| 1891 |
+
"logits/rejected": 0.6705082058906555,
|
| 1892 |
+
"logps/chosen": -1009.0233154296875,
|
| 1893 |
+
"logps/rejected": -1065.4202880859375,
|
| 1894 |
+
"loss": 4.7682,
|
| 1895 |
+
"rewards/accuracies": 0.625,
|
| 1896 |
+
"rewards/chosen": 0.6260993480682373,
|
| 1897 |
+
"rewards/margins": 0.3270193338394165,
|
| 1898 |
+
"rewards/rejected": 0.29908010363578796,
|
| 1899 |
+
"step": 126
|
| 1900 |
+
},
|
| 1901 |
+
{
|
| 1902 |
+
"epoch": 0.0317768116848591,
|
| 1903 |
+
"grad_norm": 135.96408081054688,
|
| 1904 |
+
"learning_rate": 3.175e-06,
|
| 1905 |
+
"logits/chosen": 0.9410853385925293,
|
| 1906 |
+
"logits/rejected": 0.9606730937957764,
|
| 1907 |
+
"logps/chosen": -729.5543212890625,
|
| 1908 |
+
"logps/rejected": -651.3915405273438,
|
| 1909 |
+
"loss": 4.3649,
|
| 1910 |
+
"rewards/accuracies": 0.625,
|
| 1911 |
+
"rewards/chosen": 0.4558297395706177,
|
| 1912 |
+
"rewards/margins": 0.4536476135253906,
|
| 1913 |
+
"rewards/rejected": 0.002182198688387871,
|
| 1914 |
+
"step": 127
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"epoch": 0.032027022800487914,
|
| 1918 |
+
"grad_norm": 170.97409057617188,
|
| 1919 |
+
"learning_rate": 3.2000000000000003e-06,
|
| 1920 |
+
"logits/chosen": 0.5134232640266418,
|
| 1921 |
+
"logits/rejected": 0.6284916400909424,
|
| 1922 |
+
"logps/chosen": -871.0223999023438,
|
| 1923 |
+
"logps/rejected": -807.8709106445312,
|
| 1924 |
+
"loss": 4.8398,
|
| 1925 |
+
"rewards/accuracies": 0.625,
|
| 1926 |
+
"rewards/chosen": 0.44705837965011597,
|
| 1927 |
+
"rewards/margins": 0.23864485323429108,
|
| 1928 |
+
"rewards/rejected": 0.20841355621814728,
|
| 1929 |
+
"step": 128
|
| 1930 |
+
},
|
| 1931 |
+
{
|
| 1932 |
+
"epoch": 0.03227723391611673,
|
| 1933 |
+
"grad_norm": 134.9060821533203,
|
| 1934 |
+
"learning_rate": 3.2250000000000005e-06,
|
| 1935 |
+
"logits/chosen": 0.295400470495224,
|
| 1936 |
+
"logits/rejected": 0.3619856834411621,
|
| 1937 |
+
"logps/chosen": -827.4248046875,
|
| 1938 |
+
"logps/rejected": -775.391357421875,
|
| 1939 |
+
"loss": 4.6104,
|
| 1940 |
+
"rewards/accuracies": 0.75,
|
| 1941 |
+
"rewards/chosen": 0.45551204681396484,
|
| 1942 |
+
"rewards/margins": 0.318456768989563,
|
| 1943 |
+
"rewards/rejected": 0.13705529272556305,
|
| 1944 |
+
"step": 129
|
| 1945 |
+
},
|
| 1946 |
+
{
|
| 1947 |
+
"epoch": 0.03252744503174553,
|
| 1948 |
+
"grad_norm": 139.79693603515625,
|
| 1949 |
+
"learning_rate": 3.2500000000000002e-06,
|
| 1950 |
+
"logits/chosen": 0.7282382845878601,
|
| 1951 |
+
"logits/rejected": 0.6856253743171692,
|
| 1952 |
+
"logps/chosen": -897.2752685546875,
|
| 1953 |
+
"logps/rejected": -857.0909423828125,
|
| 1954 |
+
"loss": 3.9534,
|
| 1955 |
+
"rewards/accuracies": 0.875,
|
| 1956 |
+
"rewards/chosen": 0.6775761842727661,
|
| 1957 |
+
"rewards/margins": 0.49155694246292114,
|
| 1958 |
+
"rewards/rejected": 0.18601924180984497,
|
| 1959 |
+
"step": 130
|
| 1960 |
+
},
|
| 1961 |
+
{
|
| 1962 |
+
"epoch": 0.032777656147374344,
|
| 1963 |
+
"grad_norm": 165.3737030029297,
|
| 1964 |
+
"learning_rate": 3.2750000000000004e-06,
|
| 1965 |
+
"logits/chosen": 0.6580913066864014,
|
| 1966 |
+
"logits/rejected": 0.40577080845832825,
|
| 1967 |
+
"logps/chosen": -777.4276733398438,
|
| 1968 |
+
"logps/rejected": -795.449951171875,
|
| 1969 |
+
"loss": 4.3611,
|
| 1970 |
+
"rewards/accuracies": 0.75,
|
| 1971 |
+
"rewards/chosen": 0.5797842144966125,
|
| 1972 |
+
"rewards/margins": 0.39686450362205505,
|
| 1973 |
+
"rewards/rejected": 0.1829197257757187,
|
| 1974 |
+
"step": 131
|
| 1975 |
+
},
|
| 1976 |
+
{
|
| 1977 |
+
"epoch": 0.033027867263003156,
|
| 1978 |
+
"grad_norm": 148.64382934570312,
|
| 1979 |
+
"learning_rate": 3.3000000000000006e-06,
|
| 1980 |
+
"logits/chosen": 0.7884936928749084,
|
| 1981 |
+
"logits/rejected": 0.8930270075798035,
|
| 1982 |
+
"logps/chosen": -911.4806518554688,
|
| 1983 |
+
"logps/rejected": -826.3285522460938,
|
| 1984 |
+
"loss": 3.9331,
|
| 1985 |
+
"rewards/accuracies": 0.9375,
|
| 1986 |
+
"rewards/chosen": 0.6084189414978027,
|
| 1987 |
+
"rewards/margins": 0.5333962440490723,
|
| 1988 |
+
"rewards/rejected": 0.07502269744873047,
|
| 1989 |
+
"step": 132
|
| 1990 |
+
},
|
| 1991 |
+
{
|
| 1992 |
+
"epoch": 0.03327807837863197,
|
| 1993 |
+
"grad_norm": 149.3182373046875,
|
| 1994 |
+
"learning_rate": 3.3250000000000004e-06,
|
| 1995 |
+
"logits/chosen": 0.9997194409370422,
|
| 1996 |
+
"logits/rejected": 0.8741428256034851,
|
| 1997 |
+
"logps/chosen": -750.9700927734375,
|
| 1998 |
+
"logps/rejected": -828.2926635742188,
|
| 1999 |
+
"loss": 4.5261,
|
| 2000 |
+
"rewards/accuracies": 0.75,
|
| 2001 |
+
"rewards/chosen": 0.5206612944602966,
|
| 2002 |
+
"rewards/margins": 0.34372854232788086,
|
| 2003 |
+
"rewards/rejected": 0.17693273723125458,
|
| 2004 |
+
"step": 133
|
| 2005 |
+
},
|
| 2006 |
+
{
|
| 2007 |
+
"epoch": 0.03352828949426078,
|
| 2008 |
+
"grad_norm": 205.8184051513672,
|
| 2009 |
+
"learning_rate": 3.3500000000000005e-06,
|
| 2010 |
+
"logits/chosen": 0.5166525840759277,
|
| 2011 |
+
"logits/rejected": 0.3342657685279846,
|
| 2012 |
+
"logps/chosen": -820.9586181640625,
|
| 2013 |
+
"logps/rejected": -1032.162841796875,
|
| 2014 |
+
"loss": 4.9892,
|
| 2015 |
+
"rewards/accuracies": 0.75,
|
| 2016 |
+
"rewards/chosen": 0.6503936648368835,
|
| 2017 |
+
"rewards/margins": 0.24925732612609863,
|
| 2018 |
+
"rewards/rejected": 0.4011363089084625,
|
| 2019 |
+
"step": 134
|
| 2020 |
+
},
|
| 2021 |
+
{
|
| 2022 |
+
"epoch": 0.03377850060988959,
|
| 2023 |
+
"grad_norm": 148.48712158203125,
|
| 2024 |
+
"learning_rate": 3.3750000000000003e-06,
|
| 2025 |
+
"logits/chosen": 0.3050788342952728,
|
| 2026 |
+
"logits/rejected": 0.27589350938796997,
|
| 2027 |
+
"logps/chosen": -950.6005249023438,
|
| 2028 |
+
"logps/rejected": -985.9448852539062,
|
| 2029 |
+
"loss": 4.3223,
|
| 2030 |
+
"rewards/accuracies": 0.75,
|
| 2031 |
+
"rewards/chosen": 0.6733807325363159,
|
| 2032 |
+
"rewards/margins": 0.41189804673194885,
|
| 2033 |
+
"rewards/rejected": 0.26148271560668945,
|
| 2034 |
+
"step": 135
|
| 2035 |
+
},
|
| 2036 |
+
{
|
| 2037 |
+
"epoch": 0.034028711725518405,
|
| 2038 |
+
"grad_norm": 162.81271362304688,
|
| 2039 |
+
"learning_rate": 3.4000000000000005e-06,
|
| 2040 |
+
"logits/chosen": 0.8448665142059326,
|
| 2041 |
+
"logits/rejected": 0.6966872811317444,
|
| 2042 |
+
"logps/chosen": -743.9957275390625,
|
| 2043 |
+
"logps/rejected": -750.2032470703125,
|
| 2044 |
+
"loss": 5.0515,
|
| 2045 |
+
"rewards/accuracies": 0.5625,
|
| 2046 |
+
"rewards/chosen": 0.3949689567089081,
|
| 2047 |
+
"rewards/margins": 0.22736862301826477,
|
| 2048 |
+
"rewards/rejected": 0.1676003336906433,
|
| 2049 |
+
"step": 136
|
| 2050 |
+
},
|
| 2051 |
+
{
|
| 2052 |
+
"epoch": 0.03427892284114722,
|
| 2053 |
+
"grad_norm": 139.54498291015625,
|
| 2054 |
+
"learning_rate": 3.4250000000000007e-06,
|
| 2055 |
+
"logits/chosen": 0.553779661655426,
|
| 2056 |
+
"logits/rejected": 0.5152968168258667,
|
| 2057 |
+
"logps/chosen": -796.459228515625,
|
| 2058 |
+
"logps/rejected": -836.67822265625,
|
| 2059 |
+
"loss": 4.3161,
|
| 2060 |
+
"rewards/accuracies": 0.75,
|
| 2061 |
+
"rewards/chosen": 0.7378758192062378,
|
| 2062 |
+
"rewards/margins": 0.4128691554069519,
|
| 2063 |
+
"rewards/rejected": 0.3250066041946411,
|
| 2064 |
+
"step": 137
|
| 2065 |
+
},
|
| 2066 |
+
{
|
| 2067 |
+
"epoch": 0.03452913395677603,
|
| 2068 |
+
"grad_norm": 151.59085083007812,
|
| 2069 |
+
"learning_rate": 3.45e-06,
|
| 2070 |
+
"logits/chosen": 0.5669974684715271,
|
| 2071 |
+
"logits/rejected": 0.6564877033233643,
|
| 2072 |
+
"logps/chosen": -815.4319458007812,
|
| 2073 |
+
"logps/rejected": -738.3330078125,
|
| 2074 |
+
"loss": 4.5803,
|
| 2075 |
+
"rewards/accuracies": 0.8125,
|
| 2076 |
+
"rewards/chosen": 0.4881832003593445,
|
| 2077 |
+
"rewards/margins": 0.350455105304718,
|
| 2078 |
+
"rewards/rejected": 0.13772808015346527,
|
| 2079 |
+
"step": 138
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 0.03477934507240484,
|
| 2083 |
+
"grad_norm": 148.89613342285156,
|
| 2084 |
+
"learning_rate": 3.475e-06,
|
| 2085 |
+
"logits/chosen": 0.5828185081481934,
|
| 2086 |
+
"logits/rejected": 0.3963644206523895,
|
| 2087 |
+
"logps/chosen": -909.1538696289062,
|
| 2088 |
+
"logps/rejected": -997.1546630859375,
|
| 2089 |
+
"loss": 3.4615,
|
| 2090 |
+
"rewards/accuracies": 0.875,
|
| 2091 |
+
"rewards/chosen": 0.8202193975448608,
|
| 2092 |
+
"rewards/margins": 0.8132323622703552,
|
| 2093 |
+
"rewards/rejected": 0.006987094879150391,
|
| 2094 |
+
"step": 139
|
| 2095 |
+
},
|
| 2096 |
+
{
|
| 2097 |
+
"epoch": 0.035029556188033655,
|
| 2098 |
+
"grad_norm": 153.6221160888672,
|
| 2099 |
+
"learning_rate": 3.5e-06,
|
| 2100 |
+
"logits/chosen": 1.0873308181762695,
|
| 2101 |
+
"logits/rejected": 1.020272970199585,
|
| 2102 |
+
"logps/chosen": -749.4591674804688,
|
| 2103 |
+
"logps/rejected": -737.8289794921875,
|
| 2104 |
+
"loss": 4.6594,
|
| 2105 |
+
"rewards/accuracies": 0.6875,
|
| 2106 |
+
"rewards/chosen": 0.841954231262207,
|
| 2107 |
+
"rewards/margins": 0.32822534441947937,
|
| 2108 |
+
"rewards/rejected": 0.51372891664505,
|
| 2109 |
+
"step": 140
|
| 2110 |
+
},
|
| 2111 |
+
{
|
| 2112 |
+
"epoch": 0.03527976730366247,
|
| 2113 |
+
"grad_norm": 163.4862518310547,
|
| 2114 |
+
"learning_rate": 3.525e-06,
|
| 2115 |
+
"logits/chosen": 0.6650976538658142,
|
| 2116 |
+
"logits/rejected": 0.5198307037353516,
|
| 2117 |
+
"logps/chosen": -900.0272216796875,
|
| 2118 |
+
"logps/rejected": -877.53466796875,
|
| 2119 |
+
"loss": 5.235,
|
| 2120 |
+
"rewards/accuracies": 0.625,
|
| 2121 |
+
"rewards/chosen": 0.7146763801574707,
|
| 2122 |
+
"rewards/margins": 0.21902096271514893,
|
| 2123 |
+
"rewards/rejected": 0.4956553280353546,
|
| 2124 |
+
"step": 141
|
| 2125 |
+
},
|
| 2126 |
+
{
|
| 2127 |
+
"epoch": 0.03552997841929128,
|
| 2128 |
+
"grad_norm": 175.59779357910156,
|
| 2129 |
+
"learning_rate": 3.5500000000000003e-06,
|
| 2130 |
+
"logits/chosen": 0.48263072967529297,
|
| 2131 |
+
"logits/rejected": 0.3903447985649109,
|
| 2132 |
+
"logps/chosen": -1006.64111328125,
|
| 2133 |
+
"logps/rejected": -1018.9843139648438,
|
| 2134 |
+
"loss": 4.2582,
|
| 2135 |
+
"rewards/accuracies": 0.875,
|
| 2136 |
+
"rewards/chosen": 0.7219852805137634,
|
| 2137 |
+
"rewards/margins": 0.40434205532073975,
|
| 2138 |
+
"rewards/rejected": 0.3176431953907013,
|
| 2139 |
+
"step": 142
|
| 2140 |
+
},
|
| 2141 |
+
{
|
| 2142 |
+
"epoch": 0.03578018953492009,
|
| 2143 |
+
"grad_norm": 129.54666137695312,
|
| 2144 |
+
"learning_rate": 3.575e-06,
|
| 2145 |
+
"logits/chosen": 0.9929988384246826,
|
| 2146 |
+
"logits/rejected": 0.9958979487419128,
|
| 2147 |
+
"logps/chosen": -597.6022338867188,
|
| 2148 |
+
"logps/rejected": -560.0516967773438,
|
| 2149 |
+
"loss": 4.6597,
|
| 2150 |
+
"rewards/accuracies": 0.625,
|
| 2151 |
+
"rewards/chosen": 0.4857937693595886,
|
| 2152 |
+
"rewards/margins": 0.47004491090774536,
|
| 2153 |
+
"rewards/rejected": 0.01574888825416565,
|
| 2154 |
+
"step": 143
|
| 2155 |
+
},
|
| 2156 |
+
{
|
| 2157 |
+
"epoch": 0.036030400650548904,
|
| 2158 |
+
"grad_norm": 144.1933135986328,
|
| 2159 |
+
"learning_rate": 3.6000000000000003e-06,
|
| 2160 |
+
"logits/chosen": 0.5605639219284058,
|
| 2161 |
+
"logits/rejected": 0.44549262523651123,
|
| 2162 |
+
"logps/chosen": -890.560302734375,
|
| 2163 |
+
"logps/rejected": -883.78369140625,
|
| 2164 |
+
"loss": 3.4495,
|
| 2165 |
+
"rewards/accuracies": 0.8125,
|
| 2166 |
+
"rewards/chosen": 1.008372187614441,
|
| 2167 |
+
"rewards/margins": 0.813175618648529,
|
| 2168 |
+
"rewards/rejected": 0.1951965093612671,
|
| 2169 |
+
"step": 144
|
| 2170 |
+
},
|
| 2171 |
+
{
|
| 2172 |
+
"epoch": 0.036280611766177716,
|
| 2173 |
+
"grad_norm": 164.4710235595703,
|
| 2174 |
+
"learning_rate": 3.625e-06,
|
| 2175 |
+
"logits/chosen": 0.3835439085960388,
|
| 2176 |
+
"logits/rejected": 0.2644839286804199,
|
| 2177 |
+
"logps/chosen": -756.0498046875,
|
| 2178 |
+
"logps/rejected": -946.8538818359375,
|
| 2179 |
+
"loss": 4.4913,
|
| 2180 |
+
"rewards/accuracies": 0.6875,
|
| 2181 |
+
"rewards/chosen": 0.6012188196182251,
|
| 2182 |
+
"rewards/margins": 0.5210594534873962,
|
| 2183 |
+
"rewards/rejected": 0.08015939593315125,
|
| 2184 |
+
"step": 145
|
| 2185 |
+
},
|
| 2186 |
+
{
|
| 2187 |
+
"epoch": 0.03653082288180652,
|
| 2188 |
+
"grad_norm": 117.87698364257812,
|
| 2189 |
+
"learning_rate": 3.65e-06,
|
| 2190 |
+
"logits/chosen": 0.598304033279419,
|
| 2191 |
+
"logits/rejected": 0.47974368929862976,
|
| 2192 |
+
"logps/chosen": -849.1965942382812,
|
| 2193 |
+
"logps/rejected": -900.4439697265625,
|
| 2194 |
+
"loss": 3.3203,
|
| 2195 |
+
"rewards/accuracies": 0.875,
|
| 2196 |
+
"rewards/chosen": 0.8496798276901245,
|
| 2197 |
+
"rewards/margins": 0.8870599269866943,
|
| 2198 |
+
"rewards/rejected": -0.037380047142505646,
|
| 2199 |
+
"step": 146
|
| 2200 |
+
},
|
| 2201 |
+
{
|
| 2202 |
+
"epoch": 0.03678103399743533,
|
| 2203 |
+
"grad_norm": 164.40049743652344,
|
| 2204 |
+
"learning_rate": 3.6750000000000004e-06,
|
| 2205 |
+
"logits/chosen": 0.6101506948471069,
|
| 2206 |
+
"logits/rejected": 0.6310951113700867,
|
| 2207 |
+
"logps/chosen": -801.3853149414062,
|
| 2208 |
+
"logps/rejected": -883.5050048828125,
|
| 2209 |
+
"loss": 4.7674,
|
| 2210 |
+
"rewards/accuracies": 0.75,
|
| 2211 |
+
"rewards/chosen": 0.7482701539993286,
|
| 2212 |
+
"rewards/margins": 0.41238662600517273,
|
| 2213 |
+
"rewards/rejected": 0.33588361740112305,
|
| 2214 |
+
"step": 147
|
| 2215 |
+
},
|
| 2216 |
+
{
|
| 2217 |
+
"epoch": 0.037031245113064146,
|
| 2218 |
+
"grad_norm": 163.932861328125,
|
| 2219 |
+
"learning_rate": 3.7e-06,
|
| 2220 |
+
"logits/chosen": 1.0259884595870972,
|
| 2221 |
+
"logits/rejected": 1.1341297626495361,
|
| 2222 |
+
"logps/chosen": -858.99169921875,
|
| 2223 |
+
"logps/rejected": -760.257568359375,
|
| 2224 |
+
"loss": 4.0584,
|
| 2225 |
+
"rewards/accuracies": 0.625,
|
| 2226 |
+
"rewards/chosen": 0.6627655029296875,
|
| 2227 |
+
"rewards/margins": 0.5813624858856201,
|
| 2228 |
+
"rewards/rejected": 0.08140295743942261,
|
| 2229 |
+
"step": 148
|
| 2230 |
+
},
|
| 2231 |
+
{
|
| 2232 |
+
"epoch": 0.03728145622869296,
|
| 2233 |
+
"grad_norm": 178.82398986816406,
|
| 2234 |
+
"learning_rate": 3.7250000000000003e-06,
|
| 2235 |
+
"logits/chosen": 0.8155965805053711,
|
| 2236 |
+
"logits/rejected": 0.6099752187728882,
|
| 2237 |
+
"logps/chosen": -817.9868774414062,
|
| 2238 |
+
"logps/rejected": -1045.9879150390625,
|
| 2239 |
+
"loss": 4.7943,
|
| 2240 |
+
"rewards/accuracies": 0.6875,
|
| 2241 |
+
"rewards/chosen": 0.6769387125968933,
|
| 2242 |
+
"rewards/margins": 0.3597002923488617,
|
| 2243 |
+
"rewards/rejected": 0.31723839044570923,
|
| 2244 |
+
"step": 149
|
| 2245 |
+
},
|
| 2246 |
+
{
|
| 2247 |
+
"epoch": 0.03753166734432177,
|
| 2248 |
+
"grad_norm": 169.52249145507812,
|
| 2249 |
+
"learning_rate": 3.7500000000000005e-06,
|
| 2250 |
+
"logits/chosen": 0.5471953749656677,
|
| 2251 |
+
"logits/rejected": 0.5662561058998108,
|
| 2252 |
+
"logps/chosen": -986.181884765625,
|
| 2253 |
+
"logps/rejected": -1016.1942138671875,
|
| 2254 |
+
"loss": 3.7821,
|
| 2255 |
+
"rewards/accuracies": 0.6875,
|
| 2256 |
+
"rewards/chosen": 0.6381077766418457,
|
| 2257 |
+
"rewards/margins": 0.8503923416137695,
|
| 2258 |
+
"rewards/rejected": -0.21228459477424622,
|
| 2259 |
+
"step": 150
|
| 2260 |
+
},
|
| 2261 |
+
{
|
| 2262 |
+
"epoch": 0.03778187845995058,
|
| 2263 |
+
"grad_norm": 182.09280395507812,
|
| 2264 |
+
"learning_rate": 3.7750000000000003e-06,
|
| 2265 |
+
"logits/chosen": 0.6060510873794556,
|
| 2266 |
+
"logits/rejected": 0.4075266420841217,
|
| 2267 |
+
"logps/chosen": -821.666015625,
|
| 2268 |
+
"logps/rejected": -908.2159423828125,
|
| 2269 |
+
"loss": 4.6201,
|
| 2270 |
+
"rewards/accuracies": 0.8125,
|
| 2271 |
+
"rewards/chosen": 0.5977217555046082,
|
| 2272 |
+
"rewards/margins": 0.3014836609363556,
|
| 2273 |
+
"rewards/rejected": 0.29623815417289734,
|
| 2274 |
+
"step": 151
|
| 2275 |
+
},
|
| 2276 |
+
{
|
| 2277 |
+
"epoch": 0.038032089575579395,
|
| 2278 |
+
"grad_norm": 149.020263671875,
|
| 2279 |
+
"learning_rate": 3.8000000000000005e-06,
|
| 2280 |
+
"logits/chosen": 0.9842261075973511,
|
| 2281 |
+
"logits/rejected": 0.8099168539047241,
|
| 2282 |
+
"logps/chosen": -836.44287109375,
|
| 2283 |
+
"logps/rejected": -982.6625366210938,
|
| 2284 |
+
"loss": 3.7042,
|
| 2285 |
+
"rewards/accuracies": 0.9375,
|
| 2286 |
+
"rewards/chosen": 1.0862705707550049,
|
| 2287 |
+
"rewards/margins": 0.7122077941894531,
|
| 2288 |
+
"rewards/rejected": 0.37406274676322937,
|
| 2289 |
+
"step": 152
|
| 2290 |
+
},
|
| 2291 |
+
{
|
| 2292 |
+
"epoch": 0.03828230069120821,
|
| 2293 |
+
"grad_norm": 159.7766571044922,
|
| 2294 |
+
"learning_rate": 3.825000000000001e-06,
|
| 2295 |
+
"logits/chosen": 0.7192105650901794,
|
| 2296 |
+
"logits/rejected": 0.5668576955795288,
|
| 2297 |
+
"logps/chosen": -801.5166015625,
|
| 2298 |
+
"logps/rejected": -979.3016357421875,
|
| 2299 |
+
"loss": 4.1193,
|
| 2300 |
+
"rewards/accuracies": 0.8125,
|
| 2301 |
+
"rewards/chosen": 0.6885128021240234,
|
| 2302 |
+
"rewards/margins": 0.4991152584552765,
|
| 2303 |
+
"rewards/rejected": 0.18939754366874695,
|
| 2304 |
+
"step": 153
|
| 2305 |
+
},
|
| 2306 |
+
{
|
| 2307 |
+
"epoch": 0.03853251180683702,
|
| 2308 |
+
"grad_norm": 166.79933166503906,
|
| 2309 |
+
"learning_rate": 3.85e-06,
|
| 2310 |
+
"logits/chosen": 0.7860902547836304,
|
| 2311 |
+
"logits/rejected": 0.8038786053657532,
|
| 2312 |
+
"logps/chosen": -810.453857421875,
|
| 2313 |
+
"logps/rejected": -723.3671875,
|
| 2314 |
+
"loss": 4.1486,
|
| 2315 |
+
"rewards/accuracies": 0.875,
|
| 2316 |
+
"rewards/chosen": 0.6507831811904907,
|
| 2317 |
+
"rewards/margins": 0.6382051110267639,
|
| 2318 |
+
"rewards/rejected": 0.012578055262565613,
|
| 2319 |
+
"step": 154
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 0.03878272292246583,
|
| 2323 |
+
"grad_norm": 173.12013244628906,
|
| 2324 |
+
"learning_rate": 3.875e-06,
|
| 2325 |
+
"logits/chosen": 0.9088850021362305,
|
| 2326 |
+
"logits/rejected": 0.7919357419013977,
|
| 2327 |
+
"logps/chosen": -913.0542602539062,
|
| 2328 |
+
"logps/rejected": -926.7158203125,
|
| 2329 |
+
"loss": 4.406,
|
| 2330 |
+
"rewards/accuracies": 0.6875,
|
| 2331 |
+
"rewards/chosen": 0.8225826621055603,
|
| 2332 |
+
"rewards/margins": 0.5563416481018066,
|
| 2333 |
+
"rewards/rejected": 0.2662409842014313,
|
| 2334 |
+
"step": 155
|
| 2335 |
+
},
|
| 2336 |
+
{
|
| 2337 |
+
"epoch": 0.039032934038094644,
|
| 2338 |
+
"grad_norm": 171.97964477539062,
|
| 2339 |
+
"learning_rate": 3.900000000000001e-06,
|
| 2340 |
+
"logits/chosen": 0.6967225074768066,
|
| 2341 |
+
"logits/rejected": 0.540969967842102,
|
| 2342 |
+
"logps/chosen": -849.8792114257812,
|
| 2343 |
+
"logps/rejected": -913.679443359375,
|
| 2344 |
+
"loss": 4.5975,
|
| 2345 |
+
"rewards/accuracies": 0.6875,
|
| 2346 |
+
"rewards/chosen": 0.7197656035423279,
|
| 2347 |
+
"rewards/margins": 0.35648685693740845,
|
| 2348 |
+
"rewards/rejected": 0.3632788062095642,
|
| 2349 |
+
"step": 156
|
| 2350 |
+
},
|
| 2351 |
+
{
|
| 2352 |
+
"epoch": 0.039283145153723456,
|
| 2353 |
+
"grad_norm": 144.37672424316406,
|
| 2354 |
+
"learning_rate": 3.9250000000000005e-06,
|
| 2355 |
+
"logits/chosen": 0.6681056022644043,
|
| 2356 |
+
"logits/rejected": 0.7472302317619324,
|
| 2357 |
+
"logps/chosen": -798.53759765625,
|
| 2358 |
+
"logps/rejected": -854.90087890625,
|
| 2359 |
+
"loss": 3.7286,
|
| 2360 |
+
"rewards/accuracies": 0.9375,
|
| 2361 |
+
"rewards/chosen": 0.7991371154785156,
|
| 2362 |
+
"rewards/margins": 0.7337710857391357,
|
| 2363 |
+
"rewards/rejected": 0.06536597013473511,
|
| 2364 |
+
"step": 157
|
| 2365 |
+
},
|
| 2366 |
+
{
|
| 2367 |
+
"epoch": 0.03953335626935227,
|
| 2368 |
+
"grad_norm": 141.87095642089844,
|
| 2369 |
+
"learning_rate": 3.95e-06,
|
| 2370 |
+
"logits/chosen": 0.6761986613273621,
|
| 2371 |
+
"logits/rejected": 0.579059362411499,
|
| 2372 |
+
"logps/chosen": -909.6781616210938,
|
| 2373 |
+
"logps/rejected": -862.220458984375,
|
| 2374 |
+
"loss": 3.7281,
|
| 2375 |
+
"rewards/accuracies": 0.9375,
|
| 2376 |
+
"rewards/chosen": 0.5470597743988037,
|
| 2377 |
+
"rewards/margins": 0.6505722999572754,
|
| 2378 |
+
"rewards/rejected": -0.10351258516311646,
|
| 2379 |
+
"step": 158
|
| 2380 |
+
},
|
| 2381 |
+
{
|
| 2382 |
+
"epoch": 0.03978356738498108,
|
| 2383 |
+
"grad_norm": 112.09781646728516,
|
| 2384 |
+
"learning_rate": 3.975000000000001e-06,
|
| 2385 |
+
"logits/chosen": 0.5438903570175171,
|
| 2386 |
+
"logits/rejected": 0.7648389339447021,
|
| 2387 |
+
"logps/chosen": -990.7408447265625,
|
| 2388 |
+
"logps/rejected": -835.4803466796875,
|
| 2389 |
+
"loss": 2.564,
|
| 2390 |
+
"rewards/accuracies": 1.0,
|
| 2391 |
+
"rewards/chosen": 1.2782984972000122,
|
| 2392 |
+
"rewards/margins": 1.3758175373077393,
|
| 2393 |
+
"rewards/rejected": -0.09751906245946884,
|
| 2394 |
+
"step": 159
|
| 2395 |
+
},
|
| 2396 |
+
{
|
| 2397 |
+
"epoch": 0.04003377850060989,
|
| 2398 |
+
"grad_norm": 133.63601684570312,
|
| 2399 |
+
"learning_rate": 4.000000000000001e-06,
|
| 2400 |
+
"logits/chosen": 0.560940146446228,
|
| 2401 |
+
"logits/rejected": 0.672624945640564,
|
| 2402 |
+
"logps/chosen": -807.1983642578125,
|
| 2403 |
+
"logps/rejected": -861.266845703125,
|
| 2404 |
+
"loss": 3.7409,
|
| 2405 |
+
"rewards/accuracies": 0.9375,
|
| 2406 |
+
"rewards/chosen": 0.8021640777587891,
|
| 2407 |
+
"rewards/margins": 0.6544591784477234,
|
| 2408 |
+
"rewards/rejected": 0.14770489931106567,
|
| 2409 |
+
"step": 160
|
| 2410 |
+
},
|
| 2411 |
+
{
|
| 2412 |
+
"epoch": 0.0402839896162387,
|
| 2413 |
+
"grad_norm": 168.26402282714844,
|
| 2414 |
+
"learning_rate": 4.0250000000000004e-06,
|
| 2415 |
+
"logits/chosen": 0.27081358432769775,
|
| 2416 |
+
"logits/rejected": 0.36320167779922485,
|
| 2417 |
+
"logps/chosen": -950.2415161132812,
|
| 2418 |
+
"logps/rejected": -933.5248413085938,
|
| 2419 |
+
"loss": 3.8679,
|
| 2420 |
+
"rewards/accuracies": 0.8125,
|
| 2421 |
+
"rewards/chosen": 0.9346640706062317,
|
| 2422 |
+
"rewards/margins": 0.7012830972671509,
|
| 2423 |
+
"rewards/rejected": 0.23338094353675842,
|
| 2424 |
+
"step": 161
|
| 2425 |
+
},
|
| 2426 |
+
{
|
| 2427 |
+
"epoch": 0.04053420073186751,
|
| 2428 |
+
"grad_norm": 189.97183227539062,
|
| 2429 |
+
"learning_rate": 4.05e-06,
|
| 2430 |
+
"logits/chosen": 0.7229450941085815,
|
| 2431 |
+
"logits/rejected": 0.6292408108711243,
|
| 2432 |
+
"logps/chosen": -815.431396484375,
|
| 2433 |
+
"logps/rejected": -854.240478515625,
|
| 2434 |
+
"loss": 5.1314,
|
| 2435 |
+
"rewards/accuracies": 0.75,
|
| 2436 |
+
"rewards/chosen": 0.556159257888794,
|
| 2437 |
+
"rewards/margins": 0.33854156732559204,
|
| 2438 |
+
"rewards/rejected": 0.2176176756620407,
|
| 2439 |
+
"step": 162
|
| 2440 |
+
},
|
| 2441 |
+
{
|
| 2442 |
+
"epoch": 0.04078441184749632,
|
| 2443 |
+
"grad_norm": 194.52952575683594,
|
| 2444 |
+
"learning_rate": 4.075e-06,
|
| 2445 |
+
"logits/chosen": 0.7778934240341187,
|
| 2446 |
+
"logits/rejected": 0.737458348274231,
|
| 2447 |
+
"logps/chosen": -692.4363403320312,
|
| 2448 |
+
"logps/rejected": -810.5929565429688,
|
| 2449 |
+
"loss": 4.9497,
|
| 2450 |
+
"rewards/accuracies": 0.625,
|
| 2451 |
+
"rewards/chosen": 0.5305599570274353,
|
| 2452 |
+
"rewards/margins": 0.48551517724990845,
|
| 2453 |
+
"rewards/rejected": 0.045044802129268646,
|
| 2454 |
+
"step": 163
|
| 2455 |
+
},
|
| 2456 |
+
{
|
| 2457 |
+
"epoch": 0.041034622963125135,
|
| 2458 |
+
"grad_norm": 143.23312377929688,
|
| 2459 |
+
"learning_rate": 4.1e-06,
|
| 2460 |
+
"logits/chosen": 0.7502599358558655,
|
| 2461 |
+
"logits/rejected": 0.7829670906066895,
|
| 2462 |
+
"logps/chosen": -908.2646484375,
|
| 2463 |
+
"logps/rejected": -864.3381958007812,
|
| 2464 |
+
"loss": 3.4274,
|
| 2465 |
+
"rewards/accuracies": 0.9375,
|
| 2466 |
+
"rewards/chosen": 0.8055950403213501,
|
| 2467 |
+
"rewards/margins": 0.9003482460975647,
|
| 2468 |
+
"rewards/rejected": -0.09475325047969818,
|
| 2469 |
+
"step": 164
|
| 2470 |
+
},
|
| 2471 |
+
{
|
| 2472 |
+
"epoch": 0.04128483407875395,
|
| 2473 |
+
"grad_norm": 157.2545623779297,
|
| 2474 |
+
"learning_rate": 4.125e-06,
|
| 2475 |
+
"logits/chosen": 0.42867469787597656,
|
| 2476 |
+
"logits/rejected": 0.26693516969680786,
|
| 2477 |
+
"logps/chosen": -951.4996948242188,
|
| 2478 |
+
"logps/rejected": -904.564697265625,
|
| 2479 |
+
"loss": 3.9217,
|
| 2480 |
+
"rewards/accuracies": 0.8125,
|
| 2481 |
+
"rewards/chosen": 0.9759159088134766,
|
| 2482 |
+
"rewards/margins": 0.6854816675186157,
|
| 2483 |
+
"rewards/rejected": 0.29043421149253845,
|
| 2484 |
+
"step": 165
|
| 2485 |
+
},
|
| 2486 |
+
{
|
| 2487 |
+
"epoch": 0.04153504519438276,
|
| 2488 |
+
"grad_norm": 137.5034637451172,
|
| 2489 |
+
"learning_rate": 4.15e-06,
|
| 2490 |
+
"logits/chosen": 0.5330204367637634,
|
| 2491 |
+
"logits/rejected": 0.5395705699920654,
|
| 2492 |
+
"logps/chosen": -767.4264526367188,
|
| 2493 |
+
"logps/rejected": -729.0516357421875,
|
| 2494 |
+
"loss": 3.6721,
|
| 2495 |
+
"rewards/accuracies": 0.75,
|
| 2496 |
+
"rewards/chosen": 0.8080553412437439,
|
| 2497 |
+
"rewards/margins": 0.718656063079834,
|
| 2498 |
+
"rewards/rejected": 0.08939927071332932,
|
| 2499 |
+
"step": 166
|
| 2500 |
+
},
|
| 2501 |
+
{
|
| 2502 |
+
"epoch": 0.04178525631001157,
|
| 2503 |
+
"grad_norm": 167.63526916503906,
|
| 2504 |
+
"learning_rate": 4.175e-06,
|
| 2505 |
+
"logits/chosen": 0.5876973867416382,
|
| 2506 |
+
"logits/rejected": 0.5106618404388428,
|
| 2507 |
+
"logps/chosen": -769.3453369140625,
|
| 2508 |
+
"logps/rejected": -762.6663208007812,
|
| 2509 |
+
"loss": 4.5432,
|
| 2510 |
+
"rewards/accuracies": 0.5625,
|
| 2511 |
+
"rewards/chosen": 0.6535158157348633,
|
| 2512 |
+
"rewards/margins": 0.46440136432647705,
|
| 2513 |
+
"rewards/rejected": 0.1891145408153534,
|
| 2514 |
+
"step": 167
|
| 2515 |
+
},
|
| 2516 |
+
{
|
| 2517 |
+
"epoch": 0.042035467425640384,
|
| 2518 |
+
"grad_norm": 153.6588592529297,
|
| 2519 |
+
"learning_rate": 4.2000000000000004e-06,
|
| 2520 |
+
"logits/chosen": 1.0562717914581299,
|
| 2521 |
+
"logits/rejected": 0.8941137790679932,
|
| 2522 |
+
"logps/chosen": -774.98779296875,
|
| 2523 |
+
"logps/rejected": -924.650146484375,
|
| 2524 |
+
"loss": 4.5292,
|
| 2525 |
+
"rewards/accuracies": 0.625,
|
| 2526 |
+
"rewards/chosen": 0.6449779272079468,
|
| 2527 |
+
"rewards/margins": 0.7065171599388123,
|
| 2528 |
+
"rewards/rejected": -0.061539266258478165,
|
| 2529 |
+
"step": 168
|
| 2530 |
+
},
|
| 2531 |
+
{
|
| 2532 |
+
"epoch": 0.042285678541269196,
|
| 2533 |
+
"grad_norm": 135.03091430664062,
|
| 2534 |
+
"learning_rate": 4.225e-06,
|
| 2535 |
+
"logits/chosen": 0.9676786065101624,
|
| 2536 |
+
"logits/rejected": 0.8078603744506836,
|
| 2537 |
+
"logps/chosen": -831.5836791992188,
|
| 2538 |
+
"logps/rejected": -804.6873168945312,
|
| 2539 |
+
"loss": 3.2118,
|
| 2540 |
+
"rewards/accuracies": 0.875,
|
| 2541 |
+
"rewards/chosen": 0.6791290044784546,
|
| 2542 |
+
"rewards/margins": 0.9013684391975403,
|
| 2543 |
+
"rewards/rejected": -0.2222394049167633,
|
| 2544 |
+
"step": 169
|
| 2545 |
+
},
|
| 2546 |
+
{
|
| 2547 |
+
"epoch": 0.04253588965689801,
|
| 2548 |
+
"grad_norm": 130.77609252929688,
|
| 2549 |
+
"learning_rate": 4.25e-06,
|
| 2550 |
+
"logits/chosen": 0.6716828346252441,
|
| 2551 |
+
"logits/rejected": 0.4985966086387634,
|
| 2552 |
+
"logps/chosen": -858.6065063476562,
|
| 2553 |
+
"logps/rejected": -982.9849853515625,
|
| 2554 |
+
"loss": 3.028,
|
| 2555 |
+
"rewards/accuracies": 0.875,
|
| 2556 |
+
"rewards/chosen": 0.8909362554550171,
|
| 2557 |
+
"rewards/margins": 1.2361092567443848,
|
| 2558 |
+
"rewards/rejected": -0.3451731204986572,
|
| 2559 |
+
"step": 170
|
| 2560 |
+
},
|
| 2561 |
+
{
|
| 2562 |
+
"epoch": 0.04278610077252682,
|
| 2563 |
+
"grad_norm": 141.84657287597656,
|
| 2564 |
+
"learning_rate": 4.2750000000000006e-06,
|
| 2565 |
+
"logits/chosen": 0.6962553858757019,
|
| 2566 |
+
"logits/rejected": 0.9387693405151367,
|
| 2567 |
+
"logps/chosen": -731.3770141601562,
|
| 2568 |
+
"logps/rejected": -636.809326171875,
|
| 2569 |
+
"loss": 4.3014,
|
| 2570 |
+
"rewards/accuracies": 0.75,
|
| 2571 |
+
"rewards/chosen": 0.762712836265564,
|
| 2572 |
+
"rewards/margins": 0.5948913097381592,
|
| 2573 |
+
"rewards/rejected": 0.16782140731811523,
|
| 2574 |
+
"step": 171
|
| 2575 |
+
},
|
| 2576 |
+
{
|
| 2577 |
+
"epoch": 0.04303631188815563,
|
| 2578 |
+
"grad_norm": 138.4955596923828,
|
| 2579 |
+
"learning_rate": 4.3e-06,
|
| 2580 |
+
"logits/chosen": 0.44833868741989136,
|
| 2581 |
+
"logits/rejected": 0.3760189712047577,
|
| 2582 |
+
"logps/chosen": -954.2752685546875,
|
| 2583 |
+
"logps/rejected": -976.9806518554688,
|
| 2584 |
+
"loss": 3.0478,
|
| 2585 |
+
"rewards/accuracies": 0.9375,
|
| 2586 |
+
"rewards/chosen": 1.0003539323806763,
|
| 2587 |
+
"rewards/margins": 1.2028504610061646,
|
| 2588 |
+
"rewards/rejected": -0.20249652862548828,
|
| 2589 |
+
"step": 172
|
| 2590 |
+
},
|
| 2591 |
+
{
|
| 2592 |
+
"epoch": 0.043286523003784445,
|
| 2593 |
+
"grad_norm": 125.87564849853516,
|
| 2594 |
+
"learning_rate": 4.325e-06,
|
| 2595 |
+
"logits/chosen": 0.14536720514297485,
|
| 2596 |
+
"logits/rejected": 0.34332674741744995,
|
| 2597 |
+
"logps/chosen": -845.4380493164062,
|
| 2598 |
+
"logps/rejected": -711.8458862304688,
|
| 2599 |
+
"loss": 3.6443,
|
| 2600 |
+
"rewards/accuracies": 0.75,
|
| 2601 |
+
"rewards/chosen": 0.7356263399124146,
|
| 2602 |
+
"rewards/margins": 0.8952075242996216,
|
| 2603 |
+
"rewards/rejected": -0.1595812439918518,
|
| 2604 |
+
"step": 173
|
| 2605 |
+
},
|
| 2606 |
+
{
|
| 2607 |
+
"epoch": 0.04353673411941326,
|
| 2608 |
+
"grad_norm": 153.5838623046875,
|
| 2609 |
+
"learning_rate": 4.350000000000001e-06,
|
| 2610 |
+
"logits/chosen": 0.7080106139183044,
|
| 2611 |
+
"logits/rejected": 0.696812629699707,
|
| 2612 |
+
"logps/chosen": -778.3582153320312,
|
| 2613 |
+
"logps/rejected": -852.6756591796875,
|
| 2614 |
+
"loss": 3.8358,
|
| 2615 |
+
"rewards/accuracies": 0.75,
|
| 2616 |
+
"rewards/chosen": 0.716163694858551,
|
| 2617 |
+
"rewards/margins": 0.6860113143920898,
|
| 2618 |
+
"rewards/rejected": 0.030152369290590286,
|
| 2619 |
+
"step": 174
|
| 2620 |
+
},
|
| 2621 |
+
{
|
| 2622 |
+
"epoch": 0.04378694523504207,
|
| 2623 |
+
"grad_norm": 148.90675354003906,
|
| 2624 |
+
"learning_rate": 4.3750000000000005e-06,
|
| 2625 |
+
"logits/chosen": 0.7510695457458496,
|
| 2626 |
+
"logits/rejected": 0.7831482887268066,
|
| 2627 |
+
"logps/chosen": -791.2515258789062,
|
| 2628 |
+
"logps/rejected": -890.5167846679688,
|
| 2629 |
+
"loss": 3.5614,
|
| 2630 |
+
"rewards/accuracies": 0.75,
|
| 2631 |
+
"rewards/chosen": 0.9652303457260132,
|
| 2632 |
+
"rewards/margins": 0.8300256133079529,
|
| 2633 |
+
"rewards/rejected": 0.13520470261573792,
|
| 2634 |
+
"step": 175
|
| 2635 |
+
},
|
| 2636 |
+
{
|
| 2637 |
+
"epoch": 0.044037156350670875,
|
| 2638 |
+
"grad_norm": 113.08384704589844,
|
| 2639 |
+
"learning_rate": 4.4e-06,
|
| 2640 |
+
"logits/chosen": 0.5231994986534119,
|
| 2641 |
+
"logits/rejected": 0.4796106517314911,
|
| 2642 |
+
"logps/chosen": -742.1634521484375,
|
| 2643 |
+
"logps/rejected": -823.2677612304688,
|
| 2644 |
+
"loss": 3.6957,
|
| 2645 |
+
"rewards/accuracies": 0.8125,
|
| 2646 |
+
"rewards/chosen": 0.830481231212616,
|
| 2647 |
+
"rewards/margins": 0.7274767160415649,
|
| 2648 |
+
"rewards/rejected": 0.10300463438034058,
|
| 2649 |
+
"step": 176
|
| 2650 |
+
},
|
| 2651 |
+
{
|
| 2652 |
+
"epoch": 0.04428736746629969,
|
| 2653 |
+
"grad_norm": 99.84241485595703,
|
| 2654 |
+
"learning_rate": 4.425e-06,
|
| 2655 |
+
"logits/chosen": 0.47837042808532715,
|
| 2656 |
+
"logits/rejected": 0.3609747290611267,
|
| 2657 |
+
"logps/chosen": -739.2203979492188,
|
| 2658 |
+
"logps/rejected": -757.5416259765625,
|
| 2659 |
+
"loss": 3.096,
|
| 2660 |
+
"rewards/accuracies": 0.6875,
|
| 2661 |
+
"rewards/chosen": 1.5671029090881348,
|
| 2662 |
+
"rewards/margins": 1.6435580253601074,
|
| 2663 |
+
"rewards/rejected": -0.07645538449287415,
|
| 2664 |
+
"step": 177
|
| 2665 |
+
},
|
| 2666 |
+
{
|
| 2667 |
+
"epoch": 0.0445375785819285,
|
| 2668 |
+
"grad_norm": 141.23953247070312,
|
| 2669 |
+
"learning_rate": 4.450000000000001e-06,
|
| 2670 |
+
"logits/chosen": 0.6331477761268616,
|
| 2671 |
+
"logits/rejected": 0.8105359077453613,
|
| 2672 |
+
"logps/chosen": -774.68017578125,
|
| 2673 |
+
"logps/rejected": -789.3485107421875,
|
| 2674 |
+
"loss": 4.5236,
|
| 2675 |
+
"rewards/accuracies": 0.6875,
|
| 2676 |
+
"rewards/chosen": 0.7711273431777954,
|
| 2677 |
+
"rewards/margins": 0.5388712286949158,
|
| 2678 |
+
"rewards/rejected": 0.23225615918636322,
|
| 2679 |
+
"step": 178
|
| 2680 |
+
},
|
| 2681 |
+
{
|
| 2682 |
+
"epoch": 0.04478778969755731,
|
| 2683 |
+
"grad_norm": 151.33343505859375,
|
| 2684 |
+
"learning_rate": 4.475e-06,
|
| 2685 |
+
"logits/chosen": 0.8690697550773621,
|
| 2686 |
+
"logits/rejected": 0.7319327592849731,
|
| 2687 |
+
"logps/chosen": -755.5984497070312,
|
| 2688 |
+
"logps/rejected": -873.7236938476562,
|
| 2689 |
+
"loss": 3.0347,
|
| 2690 |
+
"rewards/accuracies": 0.875,
|
| 2691 |
+
"rewards/chosen": 0.950842022895813,
|
| 2692 |
+
"rewards/margins": 1.3323326110839844,
|
| 2693 |
+
"rewards/rejected": -0.3814905285835266,
|
| 2694 |
+
"step": 179
|
| 2695 |
+
},
|
| 2696 |
+
{
|
| 2697 |
+
"epoch": 0.045038000813186124,
|
| 2698 |
+
"grad_norm": 122.43180084228516,
|
| 2699 |
+
"learning_rate": 4.5e-06,
|
| 2700 |
+
"logits/chosen": 0.41904574632644653,
|
| 2701 |
+
"logits/rejected": 0.33404892683029175,
|
| 2702 |
+
"logps/chosen": -989.1569213867188,
|
| 2703 |
+
"logps/rejected": -961.5740356445312,
|
| 2704 |
+
"loss": 2.3291,
|
| 2705 |
+
"rewards/accuracies": 0.9375,
|
| 2706 |
+
"rewards/chosen": 1.683314561843872,
|
| 2707 |
+
"rewards/margins": 1.5795354843139648,
|
| 2708 |
+
"rewards/rejected": 0.10377904027700424,
|
| 2709 |
+
"step": 180
|
| 2710 |
+
},
|
| 2711 |
+
{
|
| 2712 |
+
"epoch": 0.045288211928814937,
|
| 2713 |
+
"grad_norm": 120.82964324951172,
|
| 2714 |
+
"learning_rate": 4.525000000000001e-06,
|
| 2715 |
+
"logits/chosen": 0.17690542340278625,
|
| 2716 |
+
"logits/rejected": 0.21482543647289276,
|
| 2717 |
+
"logps/chosen": -854.7735595703125,
|
| 2718 |
+
"logps/rejected": -930.084228515625,
|
| 2719 |
+
"loss": 2.745,
|
| 2720 |
+
"rewards/accuracies": 0.875,
|
| 2721 |
+
"rewards/chosen": 1.029003620147705,
|
| 2722 |
+
"rewards/margins": 1.4290024042129517,
|
| 2723 |
+
"rewards/rejected": -0.39999890327453613,
|
| 2724 |
+
"step": 181
|
| 2725 |
+
},
|
| 2726 |
+
{
|
| 2727 |
+
"epoch": 0.04553842304444375,
|
| 2728 |
+
"grad_norm": 98.66107177734375,
|
| 2729 |
+
"learning_rate": 4.5500000000000005e-06,
|
| 2730 |
+
"logits/chosen": 0.4460292458534241,
|
| 2731 |
+
"logits/rejected": 0.509110689163208,
|
| 2732 |
+
"logps/chosen": -777.5671997070312,
|
| 2733 |
+
"logps/rejected": -822.9481811523438,
|
| 2734 |
+
"loss": 2.5145,
|
| 2735 |
+
"rewards/accuracies": 0.875,
|
| 2736 |
+
"rewards/chosen": 1.2376422882080078,
|
| 2737 |
+
"rewards/margins": 1.845048427581787,
|
| 2738 |
+
"rewards/rejected": -0.6074063181877136,
|
| 2739 |
+
"step": 182
|
| 2740 |
+
},
|
| 2741 |
+
{
|
| 2742 |
+
"epoch": 0.04578863416007256,
|
| 2743 |
+
"grad_norm": 200.034912109375,
|
| 2744 |
+
"learning_rate": 4.575e-06,
|
| 2745 |
+
"logits/chosen": 0.20393875241279602,
|
| 2746 |
+
"logits/rejected": 0.21906021237373352,
|
| 2747 |
+
"logps/chosen": -783.5596313476562,
|
| 2748 |
+
"logps/rejected": -711.1970825195312,
|
| 2749 |
+
"loss": 4.7052,
|
| 2750 |
+
"rewards/accuracies": 0.75,
|
| 2751 |
+
"rewards/chosen": 0.6918686032295227,
|
| 2752 |
+
"rewards/margins": 0.5280711054801941,
|
| 2753 |
+
"rewards/rejected": 0.1637975126504898,
|
| 2754 |
+
"step": 183
|
| 2755 |
+
},
|
| 2756 |
+
{
|
| 2757 |
+
"epoch": 0.04603884527570137,
|
| 2758 |
+
"grad_norm": 141.71707153320312,
|
| 2759 |
+
"learning_rate": 4.600000000000001e-06,
|
| 2760 |
+
"logits/chosen": 0.7268654704093933,
|
| 2761 |
+
"logits/rejected": 0.5893440246582031,
|
| 2762 |
+
"logps/chosen": -908.1588134765625,
|
| 2763 |
+
"logps/rejected": -839.3931884765625,
|
| 2764 |
+
"loss": 4.0667,
|
| 2765 |
+
"rewards/accuracies": 0.6875,
|
| 2766 |
+
"rewards/chosen": 1.4157216548919678,
|
| 2767 |
+
"rewards/margins": 0.8883066773414612,
|
| 2768 |
+
"rewards/rejected": 0.5274151563644409,
|
| 2769 |
+
"step": 184
|
| 2770 |
+
},
|
| 2771 |
+
{
|
| 2772 |
+
"epoch": 0.046289056391330186,
|
| 2773 |
+
"grad_norm": 97.34073638916016,
|
| 2774 |
+
"learning_rate": 4.625000000000001e-06,
|
| 2775 |
+
"logits/chosen": 0.6829476952552795,
|
| 2776 |
+
"logits/rejected": 0.8145304322242737,
|
| 2777 |
+
"logps/chosen": -716.8839111328125,
|
| 2778 |
+
"logps/rejected": -681.6070556640625,
|
| 2779 |
+
"loss": 2.9681,
|
| 2780 |
+
"rewards/accuracies": 0.9375,
|
| 2781 |
+
"rewards/chosen": 0.7450221180915833,
|
| 2782 |
+
"rewards/margins": 1.0474132299423218,
|
| 2783 |
+
"rewards/rejected": -0.30239105224609375,
|
| 2784 |
+
"step": 185
|
| 2785 |
+
},
|
| 2786 |
+
{
|
| 2787 |
+
"epoch": 0.046539267506959,
|
| 2788 |
+
"grad_norm": 193.9061279296875,
|
| 2789 |
+
"learning_rate": 4.65e-06,
|
| 2790 |
+
"logits/chosen": 0.6067227721214294,
|
| 2791 |
+
"logits/rejected": 0.533854603767395,
|
| 2792 |
+
"logps/chosen": -710.9017944335938,
|
| 2793 |
+
"logps/rejected": -903.33251953125,
|
| 2794 |
+
"loss": 5.8841,
|
| 2795 |
+
"rewards/accuracies": 0.5625,
|
| 2796 |
+
"rewards/chosen": 0.7944581508636475,
|
| 2797 |
+
"rewards/margins": 0.6155319213867188,
|
| 2798 |
+
"rewards/rejected": 0.17892615497112274,
|
| 2799 |
+
"step": 186
|
| 2800 |
+
},
|
| 2801 |
+
{
|
| 2802 |
+
"epoch": 0.04678947862258781,
|
| 2803 |
+
"grad_norm": 139.4214630126953,
|
| 2804 |
+
"learning_rate": 4.675000000000001e-06,
|
| 2805 |
+
"logits/chosen": 0.47161051630973816,
|
| 2806 |
+
"logits/rejected": 0.5310673713684082,
|
| 2807 |
+
"logps/chosen": -758.6500244140625,
|
| 2808 |
+
"logps/rejected": -694.9664306640625,
|
| 2809 |
+
"loss": 3.156,
|
| 2810 |
+
"rewards/accuracies": 0.75,
|
| 2811 |
+
"rewards/chosen": 1.4704225063323975,
|
| 2812 |
+
"rewards/margins": 2.380772113800049,
|
| 2813 |
+
"rewards/rejected": -0.9103498458862305,
|
| 2814 |
+
"step": 187
|
| 2815 |
+
},
|
| 2816 |
+
{
|
| 2817 |
+
"epoch": 0.04703968973821662,
|
| 2818 |
+
"grad_norm": 148.84498596191406,
|
| 2819 |
+
"learning_rate": 4.7e-06,
|
| 2820 |
+
"logits/chosen": 0.7463999390602112,
|
| 2821 |
+
"logits/rejected": 0.6421334743499756,
|
| 2822 |
+
"logps/chosen": -696.7813110351562,
|
| 2823 |
+
"logps/rejected": -788.3897094726562,
|
| 2824 |
+
"loss": 4.4071,
|
| 2825 |
+
"rewards/accuracies": 0.6875,
|
| 2826 |
+
"rewards/chosen": 0.5864818692207336,
|
| 2827 |
+
"rewards/margins": 0.9901420474052429,
|
| 2828 |
+
"rewards/rejected": -0.40365999937057495,
|
| 2829 |
+
"step": 188
|
| 2830 |
+
},
|
| 2831 |
+
{
|
| 2832 |
+
"epoch": 0.047289900853845435,
|
| 2833 |
+
"grad_norm": 96.11073303222656,
|
| 2834 |
+
"learning_rate": 4.7250000000000005e-06,
|
| 2835 |
+
"logits/chosen": 1.0857046842575073,
|
| 2836 |
+
"logits/rejected": 0.8691186308860779,
|
| 2837 |
+
"logps/chosen": -646.8231201171875,
|
| 2838 |
+
"logps/rejected": -615.858642578125,
|
| 2839 |
+
"loss": 2.8434,
|
| 2840 |
+
"rewards/accuracies": 0.875,
|
| 2841 |
+
"rewards/chosen": 0.9308801293373108,
|
| 2842 |
+
"rewards/margins": 1.6092588901519775,
|
| 2843 |
+
"rewards/rejected": -0.6783787608146667,
|
| 2844 |
+
"step": 189
|
| 2845 |
+
},
|
| 2846 |
+
{
|
| 2847 |
+
"epoch": 0.04754011196947425,
|
| 2848 |
+
"grad_norm": 171.08673095703125,
|
| 2849 |
+
"learning_rate": 4.75e-06,
|
| 2850 |
+
"logits/chosen": 0.7702385783195496,
|
| 2851 |
+
"logits/rejected": 0.643980860710144,
|
| 2852 |
+
"logps/chosen": -950.1005249023438,
|
| 2853 |
+
"logps/rejected": -999.3270874023438,
|
| 2854 |
+
"loss": 3.6033,
|
| 2855 |
+
"rewards/accuracies": 0.8125,
|
| 2856 |
+
"rewards/chosen": 1.6129151582717896,
|
| 2857 |
+
"rewards/margins": 1.880535364151001,
|
| 2858 |
+
"rewards/rejected": -0.26762041449546814,
|
| 2859 |
+
"step": 190
|
| 2860 |
+
},
|
| 2861 |
+
{
|
| 2862 |
+
"epoch": 0.04779032308510306,
|
| 2863 |
+
"grad_norm": 243.4348602294922,
|
| 2864 |
+
"learning_rate": 4.775e-06,
|
| 2865 |
+
"logits/chosen": 0.26067548990249634,
|
| 2866 |
+
"logits/rejected": 0.34642136096954346,
|
| 2867 |
+
"logps/chosen": -981.1875610351562,
|
| 2868 |
+
"logps/rejected": -782.4407958984375,
|
| 2869 |
+
"loss": 5.5213,
|
| 2870 |
+
"rewards/accuracies": 0.75,
|
| 2871 |
+
"rewards/chosen": 0.8981813788414001,
|
| 2872 |
+
"rewards/margins": 0.879020094871521,
|
| 2873 |
+
"rewards/rejected": 0.019161269068717957,
|
| 2874 |
+
"step": 191
|
| 2875 |
+
},
|
| 2876 |
+
{
|
| 2877 |
+
"epoch": 0.048040534200731864,
|
| 2878 |
+
"grad_norm": 173.99644470214844,
|
| 2879 |
+
"learning_rate": 4.800000000000001e-06,
|
| 2880 |
+
"logits/chosen": 0.672372579574585,
|
| 2881 |
+
"logits/rejected": 0.7054111957550049,
|
| 2882 |
+
"logps/chosen": -784.5167846679688,
|
| 2883 |
+
"logps/rejected": -712.2625732421875,
|
| 2884 |
+
"loss": 4.8373,
|
| 2885 |
+
"rewards/accuracies": 0.625,
|
| 2886 |
+
"rewards/chosen": 0.7209684252738953,
|
| 2887 |
+
"rewards/margins": 0.6813942193984985,
|
| 2888 |
+
"rewards/rejected": 0.03957414627075195,
|
| 2889 |
+
"step": 192
|
| 2890 |
+
},
|
| 2891 |
+
{
|
| 2892 |
+
"epoch": 0.04829074531636068,
|
| 2893 |
+
"grad_norm": 142.26370239257812,
|
| 2894 |
+
"learning_rate": 4.825e-06,
|
| 2895 |
+
"logits/chosen": 0.4435230791568756,
|
| 2896 |
+
"logits/rejected": 0.3302009105682373,
|
| 2897 |
+
"logps/chosen": -968.1591186523438,
|
| 2898 |
+
"logps/rejected": -1088.6114501953125,
|
| 2899 |
+
"loss": 3.1433,
|
| 2900 |
+
"rewards/accuracies": 0.875,
|
| 2901 |
+
"rewards/chosen": 1.4998180866241455,
|
| 2902 |
+
"rewards/margins": 1.830888032913208,
|
| 2903 |
+
"rewards/rejected": -0.3310699760913849,
|
| 2904 |
+
"step": 193
|
| 2905 |
+
},
|
| 2906 |
+
{
|
| 2907 |
+
"epoch": 0.04854095643198949,
|
| 2908 |
+
"grad_norm": 175.26205444335938,
|
| 2909 |
+
"learning_rate": 4.85e-06,
|
| 2910 |
+
"logits/chosen": 0.4360215365886688,
|
| 2911 |
+
"logits/rejected": 0.4668835401535034,
|
| 2912 |
+
"logps/chosen": -1311.623291015625,
|
| 2913 |
+
"logps/rejected": -1099.365478515625,
|
| 2914 |
+
"loss": 3.7134,
|
| 2915 |
+
"rewards/accuracies": 0.8125,
|
| 2916 |
+
"rewards/chosen": 1.3552098274230957,
|
| 2917 |
+
"rewards/margins": 1.2448309659957886,
|
| 2918 |
+
"rewards/rejected": 0.11037885397672653,
|
| 2919 |
+
"step": 194
|
| 2920 |
+
},
|
| 2921 |
+
{
|
| 2922 |
+
"epoch": 0.0487911675476183,
|
| 2923 |
+
"grad_norm": 162.4991455078125,
|
| 2924 |
+
"learning_rate": 4.875e-06,
|
| 2925 |
+
"logits/chosen": 0.43502160906791687,
|
| 2926 |
+
"logits/rejected": 0.3366948366165161,
|
| 2927 |
+
"logps/chosen": -917.325439453125,
|
| 2928 |
+
"logps/rejected": -1000.3974609375,
|
| 2929 |
+
"loss": 3.5885,
|
| 2930 |
+
"rewards/accuracies": 0.6875,
|
| 2931 |
+
"rewards/chosen": 1.0496270656585693,
|
| 2932 |
+
"rewards/margins": 1.293393611907959,
|
| 2933 |
+
"rewards/rejected": -0.24376648664474487,
|
| 2934 |
+
"step": 195
|
| 2935 |
+
},
|
| 2936 |
+
{
|
| 2937 |
+
"epoch": 0.049041378663247114,
|
| 2938 |
+
"grad_norm": 140.2455291748047,
|
| 2939 |
+
"learning_rate": 4.9000000000000005e-06,
|
| 2940 |
+
"logits/chosen": 0.7639466524124146,
|
| 2941 |
+
"logits/rejected": 0.8074101805686951,
|
| 2942 |
+
"logps/chosen": -978.490966796875,
|
| 2943 |
+
"logps/rejected": -858.6104736328125,
|
| 2944 |
+
"loss": 3.4015,
|
| 2945 |
+
"rewards/accuracies": 0.75,
|
| 2946 |
+
"rewards/chosen": 1.7361470460891724,
|
| 2947 |
+
"rewards/margins": 2.0762417316436768,
|
| 2948 |
+
"rewards/rejected": -0.3400946259498596,
|
| 2949 |
+
"step": 196
|
| 2950 |
+
},
|
| 2951 |
+
{
|
| 2952 |
+
"epoch": 0.049291589778875926,
|
| 2953 |
+
"grad_norm": 133.20802307128906,
|
| 2954 |
+
"learning_rate": 4.925e-06,
|
| 2955 |
+
"logits/chosen": 0.5749949812889099,
|
| 2956 |
+
"logits/rejected": 0.297177255153656,
|
| 2957 |
+
"logps/chosen": -712.5463256835938,
|
| 2958 |
+
"logps/rejected": -823.659912109375,
|
| 2959 |
+
"loss": 4.2319,
|
| 2960 |
+
"rewards/accuracies": 0.75,
|
| 2961 |
+
"rewards/chosen": 0.6333484649658203,
|
| 2962 |
+
"rewards/margins": 1.0953937768936157,
|
| 2963 |
+
"rewards/rejected": -0.462045282125473,
|
| 2964 |
+
"step": 197
|
| 2965 |
+
},
|
| 2966 |
+
{
|
| 2967 |
+
"epoch": 0.04954180089450474,
|
| 2968 |
+
"grad_norm": 85.10523986816406,
|
| 2969 |
+
"learning_rate": 4.95e-06,
|
| 2970 |
+
"logits/chosen": 0.45457780361175537,
|
| 2971 |
+
"logits/rejected": 0.7557879686355591,
|
| 2972 |
+
"logps/chosen": -828.2936401367188,
|
| 2973 |
+
"logps/rejected": -675.2957763671875,
|
| 2974 |
+
"loss": 2.1404,
|
| 2975 |
+
"rewards/accuracies": 0.9375,
|
| 2976 |
+
"rewards/chosen": 1.1196131706237793,
|
| 2977 |
+
"rewards/margins": 1.659280776977539,
|
| 2978 |
+
"rewards/rejected": -0.5396679043769836,
|
| 2979 |
+
"step": 198
|
| 2980 |
+
},
|
| 2981 |
+
{
|
| 2982 |
+
"epoch": 0.04979201201013355,
|
| 2983 |
+
"grad_norm": 96.11859130859375,
|
| 2984 |
+
"learning_rate": 4.975000000000001e-06,
|
| 2985 |
+
"logits/chosen": 0.8775396943092346,
|
| 2986 |
+
"logits/rejected": 0.8114781379699707,
|
| 2987 |
+
"logps/chosen": -720.8642578125,
|
| 2988 |
+
"logps/rejected": -787.377197265625,
|
| 2989 |
+
"loss": 1.9458,
|
| 2990 |
+
"rewards/accuracies": 0.875,
|
| 2991 |
+
"rewards/chosen": 1.2211010456085205,
|
| 2992 |
+
"rewards/margins": 1.7165383100509644,
|
| 2993 |
+
"rewards/rejected": -0.4954374134540558,
|
| 2994 |
+
"step": 199
|
| 2995 |
+
},
|
| 2996 |
+
{
|
| 2997 |
+
"epoch": 0.05004222312576236,
|
| 2998 |
+
"grad_norm": 161.368408203125,
|
| 2999 |
+
"learning_rate": 5e-06,
|
| 3000 |
+
"logits/chosen": 0.3539942502975464,
|
| 3001 |
+
"logits/rejected": 0.3524795472621918,
|
| 3002 |
+
"logps/chosen": -840.6842651367188,
|
| 3003 |
+
"logps/rejected": -821.8270874023438,
|
| 3004 |
+
"loss": 3.4471,
|
| 3005 |
+
"rewards/accuracies": 0.875,
|
| 3006 |
+
"rewards/chosen": 0.9537111520767212,
|
| 3007 |
+
"rewards/margins": 1.3869084119796753,
|
| 3008 |
+
"rewards/rejected": -0.4331972002983093,
|
| 3009 |
+
"step": 200
|
| 3010 |
}
|
| 3011 |
],
|
| 3012 |
"logging_steps": 1,
|
|
|
|
| 3021 |
"should_evaluate": false,
|
| 3022 |
"should_log": false,
|
| 3023 |
"should_save": true,
|
| 3024 |
+
"should_training_stop": true
|
| 3025 |
},
|
| 3026 |
"attributes": {}
|
| 3027 |
}
|