checkpoint 9690 contra
Browse files- model.safetensors +1 -1
- optimizer.pt +3 -0
- rng_state.pth +3 -0
- scheduler.pt +3 -0
- trainer_state.json +1010 -0
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1192135096
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:728e0719bfc2b88b533e21bdb4f2454d9fd976656cdd246883e6a01c7411a720
|
| 3 |
size 1192135096
|
optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b45ae3f39d88743b1662eb6134b02ad750d8cfaf6d404377e8efc8e59b6194d0
|
| 3 |
+
size 2384460363
|
rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:91e62442493ae4f2963ab198bb4e1be4db65ab5cdd6b158248d98219357cb2b0
|
| 3 |
+
size 14645
|
scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4af0f0a49e234227ff12101e0294b33302da01a67e316f5465c9475b2201d4ef
|
| 3 |
+
size 1465
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1010 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 500,
|
| 3 |
+
"best_metric": 279.8433532714844,
|
| 4 |
+
"best_model_checkpoint": "output-dpo-contrastive-new/checkpoint-500",
|
| 5 |
+
"epoch": 3.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 9690,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.03096634345545685,
|
| 14 |
+
"grad_norm": 1144.0,
|
| 15 |
+
"learning_rate": 9.997424714962634e-06,
|
| 16 |
+
"loss": 478.5263,
|
| 17 |
+
"step": 100
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.0619326869109137,
|
| 21 |
+
"grad_norm": 249.0,
|
| 22 |
+
"learning_rate": 9.989597261051275e-06,
|
| 23 |
+
"loss": 5.8995,
|
| 24 |
+
"step": 200
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.09289903036637055,
|
| 28 |
+
"grad_norm": 1416.0,
|
| 29 |
+
"learning_rate": 9.976525602356642e-06,
|
| 30 |
+
"loss": 1.974,
|
| 31 |
+
"step": 300
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"epoch": 0.1238653738218274,
|
| 35 |
+
"grad_norm": 124.0,
|
| 36 |
+
"learning_rate": 9.958223477553715e-06,
|
| 37 |
+
"loss": 1.2967,
|
| 38 |
+
"step": 400
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.15483171727728426,
|
| 42 |
+
"grad_norm": 185.0,
|
| 43 |
+
"learning_rate": 9.93471012268208e-06,
|
| 44 |
+
"loss": 0.9707,
|
| 45 |
+
"step": 500
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 0.15483171727728426,
|
| 49 |
+
"eval_logits/chosen": NaN,
|
| 50 |
+
"eval_logits/rejected": NaN,
|
| 51 |
+
"eval_logps/chosen": -7700.236328125,
|
| 52 |
+
"eval_logps/rejected": -5658.39892578125,
|
| 53 |
+
"eval_loss": 279.8433532714844,
|
| 54 |
+
"eval_rewards/accuracies": 0.3200867176055908,
|
| 55 |
+
"eval_rewards/chosen": -724.652587890625,
|
| 56 |
+
"eval_rewards/margins": -192.2952880859375,
|
| 57 |
+
"eval_rewards/rejected": -532.3572998046875,
|
| 58 |
+
"eval_runtime": 1360.3414,
|
| 59 |
+
"eval_samples_per_second": 14.243,
|
| 60 |
+
"eval_steps_per_second": 14.243,
|
| 61 |
+
"step": 500
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"epoch": 0.1857980607327411,
|
| 65 |
+
"grad_norm": 136.0,
|
| 66 |
+
"learning_rate": 9.906010250928317e-06,
|
| 67 |
+
"loss": 0.8311,
|
| 68 |
+
"step": 600
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.21676440418819795,
|
| 72 |
+
"grad_norm": 66.5,
|
| 73 |
+
"learning_rate": 9.872154026651767e-06,
|
| 74 |
+
"loss": 0.6984,
|
| 75 |
+
"step": 700
|
| 76 |
+
},
|
| 77 |
+
{
|
| 78 |
+
"epoch": 0.2477307476436548,
|
| 79 |
+
"grad_norm": 768.0,
|
| 80 |
+
"learning_rate": 9.833177033680945e-06,
|
| 81 |
+
"loss": 0.7479,
|
| 82 |
+
"step": 800
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"epoch": 0.27869709109911167,
|
| 86 |
+
"grad_norm": 73.5,
|
| 87 |
+
"learning_rate": 9.789120237913954e-06,
|
| 88 |
+
"loss": 0.5771,
|
| 89 |
+
"step": 900
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.3096634345545685,
|
| 93 |
+
"grad_norm": 71.5,
|
| 94 |
+
"learning_rate": 9.740029944262193e-06,
|
| 95 |
+
"loss": 0.5536,
|
| 96 |
+
"step": 1000
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"epoch": 0.3096634345545685,
|
| 100 |
+
"eval_logits/chosen": NaN,
|
| 101 |
+
"eval_logits/rejected": NaN,
|
| 102 |
+
"eval_logps/chosen": -8134.36328125,
|
| 103 |
+
"eval_logps/rejected": -5972.11669921875,
|
| 104 |
+
"eval_loss": 297.4208679199219,
|
| 105 |
+
"eval_rewards/accuracies": 0.31869322061538696,
|
| 106 |
+
"eval_rewards/chosen": -768.0653686523438,
|
| 107 |
+
"eval_rewards/margins": -204.33627319335938,
|
| 108 |
+
"eval_rewards/rejected": -563.7291259765625,
|
| 109 |
+
"eval_runtime": 1355.184,
|
| 110 |
+
"eval_samples_per_second": 14.298,
|
| 111 |
+
"eval_steps_per_second": 14.298,
|
| 112 |
+
"step": 1000
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"epoch": 0.34062977801002536,
|
| 116 |
+
"grad_norm": 53.5,
|
| 117 |
+
"learning_rate": 9.685957747982618e-06,
|
| 118 |
+
"loss": 0.5381,
|
| 119 |
+
"step": 1100
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"epoch": 0.3715961214654822,
|
| 123 |
+
"grad_norm": 84.5,
|
| 124 |
+
"learning_rate": 9.626960480449692e-06,
|
| 125 |
+
"loss": 0.5969,
|
| 126 |
+
"step": 1200
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"epoch": 0.40256246492093906,
|
| 130 |
+
"grad_norm": 70.5,
|
| 131 |
+
"learning_rate": 9.563100149424045e-06,
|
| 132 |
+
"loss": 0.4972,
|
| 133 |
+
"step": 1300
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"epoch": 0.4335288083763959,
|
| 137 |
+
"grad_norm": 85.5,
|
| 138 |
+
"learning_rate": 9.4944438738806e-06,
|
| 139 |
+
"loss": 0.4799,
|
| 140 |
+
"step": 1400
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 0.46449515183185275,
|
| 144 |
+
"grad_norm": 79.0,
|
| 145 |
+
"learning_rate": 9.421063813464661e-06,
|
| 146 |
+
"loss": 0.4689,
|
| 147 |
+
"step": 1500
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"epoch": 0.46449515183185275,
|
| 151 |
+
"eval_logits/chosen": NaN,
|
| 152 |
+
"eval_logits/rejected": NaN,
|
| 153 |
+
"eval_logps/chosen": -8173.689453125,
|
| 154 |
+
"eval_logps/rejected": -5999.046875,
|
| 155 |
+
"eval_loss": 298.9976806640625,
|
| 156 |
+
"eval_rewards/accuracies": 0.3185900151729584,
|
| 157 |
+
"eval_rewards/chosen": -771.9979248046875,
|
| 158 |
+
"eval_rewards/margins": -205.575927734375,
|
| 159 |
+
"eval_rewards/rejected": -566.4219360351562,
|
| 160 |
+
"eval_runtime": 1358.7276,
|
| 161 |
+
"eval_samples_per_second": 14.26,
|
| 162 |
+
"eval_steps_per_second": 14.26,
|
| 163 |
+
"step": 1500
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.4954614952873096,
|
| 167 |
+
"grad_norm": 64.5,
|
| 168 |
+
"learning_rate": 9.343037092650156e-06,
|
| 169 |
+
"loss": 0.4951,
|
| 170 |
+
"step": 1600
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.5264278387427664,
|
| 174 |
+
"grad_norm": 55.25,
|
| 175 |
+
"learning_rate": 9.260445719679651e-06,
|
| 176 |
+
"loss": 0.4569,
|
| 177 |
+
"step": 1700
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.5573941821982233,
|
| 181 |
+
"grad_norm": 71.0,
|
| 182 |
+
"learning_rate": 9.173376500371441e-06,
|
| 183 |
+
"loss": 0.4647,
|
| 184 |
+
"step": 1800
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 0.5883605256536801,
|
| 188 |
+
"grad_norm": 76.0,
|
| 189 |
+
"learning_rate": 9.081920946884217e-06,
|
| 190 |
+
"loss": 0.4954,
|
| 191 |
+
"step": 1900
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.619326869109137,
|
| 195 |
+
"grad_norm": 48.75,
|
| 196 |
+
"learning_rate": 8.986175181535266e-06,
|
| 197 |
+
"loss": 0.4217,
|
| 198 |
+
"step": 2000
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.619326869109137,
|
| 202 |
+
"eval_logits/chosen": NaN,
|
| 203 |
+
"eval_logits/rejected": NaN,
|
| 204 |
+
"eval_logps/chosen": -8315.658203125,
|
| 205 |
+
"eval_logps/rejected": -6106.26806640625,
|
| 206 |
+
"eval_loss": 304.197021484375,
|
| 207 |
+
"eval_rewards/accuracies": 0.3195706009864807,
|
| 208 |
+
"eval_rewards/chosen": -786.1946411132812,
|
| 209 |
+
"eval_rewards/margins": -209.05052185058594,
|
| 210 |
+
"eval_rewards/rejected": -577.1441650390625,
|
| 211 |
+
"eval_runtime": 1355.5933,
|
| 212 |
+
"eval_samples_per_second": 14.293,
|
| 213 |
+
"eval_steps_per_second": 14.293,
|
| 214 |
+
"step": 2000
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 0.6502932125645938,
|
| 218 |
+
"grad_norm": 43.75,
|
| 219 |
+
"learning_rate": 8.886239835773252e-06,
|
| 220 |
+
"loss": 0.4427,
|
| 221 |
+
"step": 2100
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.6812595560200507,
|
| 225 |
+
"grad_norm": 47.25,
|
| 226 |
+
"learning_rate": 8.782219944411774e-06,
|
| 227 |
+
"loss": 0.4756,
|
| 228 |
+
"step": 2200
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 0.7122258994755075,
|
| 232 |
+
"grad_norm": 64.5,
|
| 233 |
+
"learning_rate": 8.674224835234879e-06,
|
| 234 |
+
"loss": 0.4246,
|
| 235 |
+
"step": 2300
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.7431922429309644,
|
| 239 |
+
"grad_norm": 71.5,
|
| 240 |
+
"learning_rate": 8.562368014090532e-06,
|
| 241 |
+
"loss": 0.4434,
|
| 242 |
+
"step": 2400
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 0.7741585863864212,
|
| 246 |
+
"grad_norm": 77.0,
|
| 247 |
+
"learning_rate": 8.446767045592829e-06,
|
| 248 |
+
"loss": 0.5978,
|
| 249 |
+
"step": 2500
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.7741585863864212,
|
| 253 |
+
"eval_logits/chosen": NaN,
|
| 254 |
+
"eval_logits/rejected": NaN,
|
| 255 |
+
"eval_logps/chosen": -8511.75,
|
| 256 |
+
"eval_logps/rejected": -6250.36181640625,
|
| 257 |
+
"eval_loss": 311.8685607910156,
|
| 258 |
+
"eval_rewards/accuracies": 0.3185900151729584,
|
| 259 |
+
"eval_rewards/chosen": -805.803955078125,
|
| 260 |
+
"eval_rewards/margins": -214.25033569335938,
|
| 261 |
+
"eval_rewards/rejected": -591.5535888671875,
|
| 262 |
+
"eval_runtime": 1359.1867,
|
| 263 |
+
"eval_samples_per_second": 14.256,
|
| 264 |
+
"eval_steps_per_second": 14.256,
|
| 265 |
+
"step": 2500
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.8051249298418781,
|
| 269 |
+
"grad_norm": 49.75,
|
| 270 |
+
"learning_rate": 8.327543429558335e-06,
|
| 271 |
+
"loss": 0.4466,
|
| 272 |
+
"step": 2600
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 0.836091273297335,
|
| 276 |
+
"grad_norm": 52.25,
|
| 277 |
+
"learning_rate": 8.20482247330641e-06,
|
| 278 |
+
"loss": 0.4372,
|
| 279 |
+
"step": 2700
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 0.8670576167527918,
|
| 283 |
+
"grad_norm": 62.0,
|
| 284 |
+
"learning_rate": 8.07873315995776e-06,
|
| 285 |
+
"loss": 0.4154,
|
| 286 |
+
"step": 2800
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 0.8980239602082487,
|
| 290 |
+
"grad_norm": 69.0,
|
| 291 |
+
"learning_rate": 7.9494080128696e-06,
|
| 292 |
+
"loss": 0.4073,
|
| 293 |
+
"step": 2900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 0.9289903036637055,
|
| 297 |
+
"grad_norm": 46.75,
|
| 298 |
+
"learning_rate": 7.816982956349941e-06,
|
| 299 |
+
"loss": 0.399,
|
| 300 |
+
"step": 3000
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 0.9289903036637055,
|
| 304 |
+
"eval_logits/chosen": NaN,
|
| 305 |
+
"eval_logits/rejected": NaN,
|
| 306 |
+
"eval_logps/chosen": -8489.6416015625,
|
| 307 |
+
"eval_logps/rejected": -6233.3359375,
|
| 308 |
+
"eval_loss": 311.04168701171875,
|
| 309 |
+
"eval_rewards/accuracies": 0.31926095485687256,
|
| 310 |
+
"eval_rewards/chosen": -803.5930786132812,
|
| 311 |
+
"eval_rewards/margins": -213.7421875,
|
| 312 |
+
"eval_rewards/rejected": -589.8509521484375,
|
| 313 |
+
"eval_runtime": 1346.8135,
|
| 314 |
+
"eval_samples_per_second": 14.387,
|
| 315 |
+
"eval_steps_per_second": 14.387,
|
| 316 |
+
"step": 3000
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"epoch": 0.9599566471191624,
|
| 320 |
+
"grad_norm": 40.75,
|
| 321 |
+
"learning_rate": 7.681597172797377e-06,
|
| 322 |
+
"loss": 0.4016,
|
| 323 |
+
"step": 3100
|
| 324 |
+
},
|
| 325 |
+
{
|
| 326 |
+
"epoch": 0.9909229905746192,
|
| 327 |
+
"grad_norm": 61.5,
|
| 328 |
+
"learning_rate": 7.543392956416542e-06,
|
| 329 |
+
"loss": 0.4017,
|
| 330 |
+
"step": 3200
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 1.0216764404188199,
|
| 334 |
+
"grad_norm": 49.75,
|
| 335 |
+
"learning_rate": 7.4025155636629546e-06,
|
| 336 |
+
"loss": 0.3778,
|
| 337 |
+
"step": 3300
|
| 338 |
+
},
|
| 339 |
+
{
|
| 340 |
+
"epoch": 1.0526427838742767,
|
| 341 |
+
"grad_norm": 45.0,
|
| 342 |
+
"learning_rate": 7.25911306057447e-06,
|
| 343 |
+
"loss": 0.3591,
|
| 344 |
+
"step": 3400
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 1.0836091273297335,
|
| 348 |
+
"grad_norm": 61.5,
|
| 349 |
+
"learning_rate": 7.113336167149775e-06,
|
| 350 |
+
"loss": 0.3661,
|
| 351 |
+
"step": 3500
|
| 352 |
+
},
|
| 353 |
+
{
|
| 354 |
+
"epoch": 1.0836091273297335,
|
| 355 |
+
"eval_logits/chosen": NaN,
|
| 356 |
+
"eval_logits/rejected": NaN,
|
| 357 |
+
"eval_logps/chosen": -8532.7568359375,
|
| 358 |
+
"eval_logps/rejected": -6264.232421875,
|
| 359 |
+
"eval_loss": 312.83892822265625,
|
| 360 |
+
"eval_rewards/accuracies": 0.31828033924102783,
|
| 361 |
+
"eval_rewards/chosen": -807.90478515625,
|
| 362 |
+
"eval_rewards/margins": -214.964111328125,
|
| 363 |
+
"eval_rewards/rejected": -592.940673828125,
|
| 364 |
+
"eval_runtime": 1355.1706,
|
| 365 |
+
"eval_samples_per_second": 14.298,
|
| 366 |
+
"eval_steps_per_second": 14.298,
|
| 367 |
+
"step": 3500
|
| 368 |
+
},
|
| 369 |
+
{
|
| 370 |
+
"epoch": 1.1145754707851903,
|
| 371 |
+
"grad_norm": 41.25,
|
| 372 |
+
"learning_rate": 6.965338098937496e-06,
|
| 373 |
+
"loss": 0.3844,
|
| 374 |
+
"step": 3600
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 1.1455418142406473,
|
| 378 |
+
"grad_norm": 62.5,
|
| 379 |
+
"learning_rate": 6.815274406002428e-06,
|
| 380 |
+
"loss": 0.3653,
|
| 381 |
+
"step": 3700
|
| 382 |
+
},
|
| 383 |
+
{
|
| 384 |
+
"epoch": 1.176508157696104,
|
| 385 |
+
"grad_norm": 95.0,
|
| 386 |
+
"learning_rate": 6.663302809438097e-06,
|
| 387 |
+
"loss": 0.3915,
|
| 388 |
+
"step": 3800
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 1.2074745011515609,
|
| 392 |
+
"grad_norm": 50.25,
|
| 393 |
+
"learning_rate": 6.509583035597538e-06,
|
| 394 |
+
"loss": 0.3743,
|
| 395 |
+
"step": 3900
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"epoch": 1.2384408446070179,
|
| 399 |
+
"grad_norm": 57.75,
|
| 400 |
+
"learning_rate": 6.35427664821648e-06,
|
| 401 |
+
"loss": 0.3638,
|
| 402 |
+
"step": 4000
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 1.2384408446070179,
|
| 406 |
+
"eval_logits/chosen": NaN,
|
| 407 |
+
"eval_logits/rejected": NaN,
|
| 408 |
+
"eval_logps/chosen": -8631.439453125,
|
| 409 |
+
"eval_logps/rejected": -6336.66796875,
|
| 410 |
+
"eval_loss": 316.6842346191406,
|
| 411 |
+
"eval_rewards/accuracies": 0.3185900151729584,
|
| 412 |
+
"eval_rewards/chosen": -817.7728271484375,
|
| 413 |
+
"eval_rewards/margins": -217.5886993408203,
|
| 414 |
+
"eval_rewards/rejected": -600.1841430664062,
|
| 415 |
+
"eval_runtime": 1356.2341,
|
| 416 |
+
"eval_samples_per_second": 14.287,
|
| 417 |
+
"eval_steps_per_second": 14.287,
|
| 418 |
+
"step": 4000
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 1.2694071880624747,
|
| 422 |
+
"grad_norm": 68.0,
|
| 423 |
+
"learning_rate": 6.197546878605376e-06,
|
| 424 |
+
"loss": 0.4077,
|
| 425 |
+
"step": 4100
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 1.3003735315179314,
|
| 429 |
+
"grad_norm": 44.0,
|
| 430 |
+
"learning_rate": 6.039558454088796e-06,
|
| 431 |
+
"loss": 0.3837,
|
| 432 |
+
"step": 4200
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"epoch": 1.3313398749733882,
|
| 436 |
+
"grad_norm": 47.75,
|
| 437 |
+
"learning_rate": 5.88047742487244e-06,
|
| 438 |
+
"loss": 0.3916,
|
| 439 |
+
"step": 4300
|
| 440 |
+
},
|
| 441 |
+
{
|
| 442 |
+
"epoch": 1.362306218428845,
|
| 443 |
+
"grad_norm": 70.5,
|
| 444 |
+
"learning_rate": 5.720470989519773e-06,
|
| 445 |
+
"loss": 0.3604,
|
| 446 |
+
"step": 4400
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 1.393272561884302,
|
| 450 |
+
"grad_norm": 42.5,
|
| 451 |
+
"learning_rate": 5.559707319221725e-06,
|
| 452 |
+
"loss": 0.3803,
|
| 453 |
+
"step": 4500
|
| 454 |
+
},
|
| 455 |
+
{
|
| 456 |
+
"epoch": 1.393272561884302,
|
| 457 |
+
"eval_logits/chosen": NaN,
|
| 458 |
+
"eval_logits/rejected": NaN,
|
| 459 |
+
"eval_logps/chosen": -8557.1044921875,
|
| 460 |
+
"eval_logps/rejected": -6282.44775390625,
|
| 461 |
+
"eval_loss": 313.7301940917969,
|
| 462 |
+
"eval_rewards/accuracies": 0.3190028965473175,
|
| 463 |
+
"eval_rewards/chosen": -810.3394165039062,
|
| 464 |
+
"eval_rewards/margins": -215.5771942138672,
|
| 465 |
+
"eval_rewards/rejected": -594.7621459960938,
|
| 466 |
+
"eval_runtime": 1359.1293,
|
| 467 |
+
"eval_samples_per_second": 14.256,
|
| 468 |
+
"eval_steps_per_second": 14.256,
|
| 469 |
+
"step": 4500
|
| 470 |
+
},
|
| 471 |
+
{
|
| 472 |
+
"epoch": 1.4242389053397588,
|
| 473 |
+
"grad_norm": 37.25,
|
| 474 |
+
"learning_rate": 5.3983553810441004e-06,
|
| 475 |
+
"loss": 0.3853,
|
| 476 |
+
"step": 4600
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 1.4552052487952156,
|
| 480 |
+
"grad_norm": 49.0,
|
| 481 |
+
"learning_rate": 5.236584760338523e-06,
|
| 482 |
+
"loss": 0.3786,
|
| 483 |
+
"step": 4700
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"epoch": 1.4861715922506726,
|
| 487 |
+
"grad_norm": 38.75,
|
| 488 |
+
"learning_rate": 5.074565482503543e-06,
|
| 489 |
+
"loss": 0.3855,
|
| 490 |
+
"step": 4800
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"epoch": 1.5171379357061294,
|
| 494 |
+
"grad_norm": 36.5,
|
| 495 |
+
"learning_rate": 4.912467834283229e-06,
|
| 496 |
+
"loss": 0.3645,
|
| 497 |
+
"step": 4900
|
| 498 |
+
},
|
| 499 |
+
{
|
| 500 |
+
"epoch": 1.5481042791615862,
|
| 501 |
+
"grad_norm": 49.25,
|
| 502 |
+
"learning_rate": 4.750462184791093e-06,
|
| 503 |
+
"loss": 0.3757,
|
| 504 |
+
"step": 5000
|
| 505 |
+
},
|
| 506 |
+
{
|
| 507 |
+
"epoch": 1.5481042791615862,
|
| 508 |
+
"eval_logits/chosen": NaN,
|
| 509 |
+
"eval_logits/rejected": NaN,
|
| 510 |
+
"eval_logps/chosen": -8567.33203125,
|
| 511 |
+
"eval_logps/rejected": -6289.53173828125,
|
| 512 |
+
"eval_loss": 314.1842346191406,
|
| 513 |
+
"eval_rewards/accuracies": 0.3201383054256439,
|
| 514 |
+
"eval_rewards/chosen": -811.3621215820312,
|
| 515 |
+
"eval_rewards/margins": -215.89157104492188,
|
| 516 |
+
"eval_rewards/rejected": -595.470458984375,
|
| 517 |
+
"eval_runtime": 1365.2519,
|
| 518 |
+
"eval_samples_per_second": 14.192,
|
| 519 |
+
"eval_steps_per_second": 14.192,
|
| 520 |
+
"step": 5000
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 1.579070622617043,
|
| 524 |
+
"grad_norm": 82.5,
|
| 525 |
+
"learning_rate": 4.588718806447441e-06,
|
| 526 |
+
"loss": 0.3887,
|
| 527 |
+
"step": 5100
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 1.6100369660724998,
|
| 531 |
+
"grad_norm": 46.0,
|
| 532 |
+
"learning_rate": 4.427407696018343e-06,
|
| 533 |
+
"loss": 0.3759,
|
| 534 |
+
"step": 5200
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"epoch": 1.6410033095279568,
|
| 538 |
+
"grad_norm": 39.25,
|
| 539 |
+
"learning_rate": 4.266698395944332e-06,
|
| 540 |
+
"loss": 0.3836,
|
| 541 |
+
"step": 5300
|
| 542 |
+
},
|
| 543 |
+
{
|
| 544 |
+
"epoch": 1.6719696529834136,
|
| 545 |
+
"grad_norm": 42.5,
|
| 546 |
+
"learning_rate": 4.1067598161466e-06,
|
| 547 |
+
"loss": 0.3727,
|
| 548 |
+
"step": 5400
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 1.7029359964388706,
|
| 552 |
+
"grad_norm": 64.5,
|
| 553 |
+
"learning_rate": 3.947760056498012e-06,
|
| 554 |
+
"loss": 0.3917,
|
| 555 |
+
"step": 5500
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 1.7029359964388706,
|
| 559 |
+
"eval_logits/chosen": NaN,
|
| 560 |
+
"eval_logits/rejected": NaN,
|
| 561 |
+
"eval_logps/chosen": -8585.994140625,
|
| 562 |
+
"eval_logps/rejected": -6303.09765625,
|
| 563 |
+
"eval_loss": 314.93072509765625,
|
| 564 |
+
"eval_rewards/accuracies": 0.31920933723449707,
|
| 565 |
+
"eval_rewards/chosen": -813.2284545898438,
|
| 566 |
+
"eval_rewards/margins": -216.4012451171875,
|
| 567 |
+
"eval_rewards/rejected": -596.8271484375,
|
| 568 |
+
"eval_runtime": 1366.4753,
|
| 569 |
+
"eval_samples_per_second": 14.18,
|
| 570 |
+
"eval_steps_per_second": 14.18,
|
| 571 |
+
"step": 5500
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"epoch": 1.7339023398943274,
|
| 575 |
+
"grad_norm": 52.5,
|
| 576 |
+
"learning_rate": 3.7898662301454724e-06,
|
| 577 |
+
"loss": 0.3941,
|
| 578 |
+
"step": 5600
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 1.7648686833497842,
|
| 582 |
+
"grad_norm": 42.75,
|
| 583 |
+
"learning_rate": 3.6332442878693896e-06,
|
| 584 |
+
"loss": 0.3701,
|
| 585 |
+
"step": 5700
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 1.795835026805241,
|
| 589 |
+
"grad_norm": 42.75,
|
| 590 |
+
"learning_rate": 3.4780588436648223e-06,
|
| 591 |
+
"loss": 0.3707,
|
| 592 |
+
"step": 5800
|
| 593 |
+
},
|
| 594 |
+
{
|
| 595 |
+
"epoch": 1.8268013702606978,
|
| 596 |
+
"grad_norm": 52.75,
|
| 597 |
+
"learning_rate": 3.3244730017275974e-06,
|
| 598 |
+
"loss": 0.3729,
|
| 599 |
+
"step": 5900
|
| 600 |
+
},
|
| 601 |
+
{
|
| 602 |
+
"epoch": 1.8577677137161548,
|
| 603 |
+
"grad_norm": 49.75,
|
| 604 |
+
"learning_rate": 3.172648185027306e-06,
|
| 605 |
+
"loss": 0.3716,
|
| 606 |
+
"step": 6000
|
| 607 |
+
},
|
| 608 |
+
{
|
| 609 |
+
"epoch": 1.8577677137161548,
|
| 610 |
+
"eval_logits/chosen": NaN,
|
| 611 |
+
"eval_logits/rejected": NaN,
|
| 612 |
+
"eval_logps/chosen": -8575.375,
|
| 613 |
+
"eval_logps/rejected": -6295.24853515625,
|
| 614 |
+
"eval_loss": 314.5245361328125,
|
| 615 |
+
"eval_rewards/accuracies": 0.3183319568634033,
|
| 616 |
+
"eval_rewards/chosen": -812.16650390625,
|
| 617 |
+
"eval_rewards/margins": -216.12428283691406,
|
| 618 |
+
"eval_rewards/rejected": -596.042236328125,
|
| 619 |
+
"eval_runtime": 1368.3845,
|
| 620 |
+
"eval_samples_per_second": 14.16,
|
| 621 |
+
"eval_steps_per_second": 14.16,
|
| 622 |
+
"step": 6000
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"epoch": 1.8887340571716116,
|
| 626 |
+
"grad_norm": 54.25,
|
| 627 |
+
"learning_rate": 3.0227439656472878e-06,
|
| 628 |
+
"loss": 0.3626,
|
| 629 |
+
"step": 6100
|
| 630 |
+
},
|
| 631 |
+
{
|
| 632 |
+
"epoch": 1.9197004006270686,
|
| 633 |
+
"grad_norm": 38.5,
|
| 634 |
+
"learning_rate": 2.87491789706995e-06,
|
| 635 |
+
"loss": 0.393,
|
| 636 |
+
"step": 6200
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 1.9506667440825254,
|
| 640 |
+
"grad_norm": 40.5,
|
| 641 |
+
"learning_rate": 2.729325348583711e-06,
|
| 642 |
+
"loss": 0.3619,
|
| 643 |
+
"step": 6300
|
| 644 |
+
},
|
| 645 |
+
{
|
| 646 |
+
"epoch": 1.9816330875379822,
|
| 647 |
+
"grad_norm": 80.0,
|
| 648 |
+
"learning_rate": 2.5861193419855634e-06,
|
| 649 |
+
"loss": 0.4156,
|
| 650 |
+
"step": 6400
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 2.0123865373821825,
|
| 654 |
+
"grad_norm": 82.5,
|
| 655 |
+
"learning_rate": 2.4454503907509493e-06,
|
| 656 |
+
"loss": 0.3607,
|
| 657 |
+
"step": 6500
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 2.0123865373821825,
|
| 661 |
+
"eval_logits/chosen": NaN,
|
| 662 |
+
"eval_logits/rejected": NaN,
|
| 663 |
+
"eval_logps/chosen": -8613.890625,
|
| 664 |
+
"eval_logps/rejected": -6323.0517578125,
|
| 665 |
+
"eval_loss": 316.09857177734375,
|
| 666 |
+
"eval_rewards/accuracies": 0.3194673955440521,
|
| 667 |
+
"eval_rewards/chosen": -816.01806640625,
|
| 668 |
+
"eval_rewards/margins": -217.19549560546875,
|
| 669 |
+
"eval_rewards/rejected": -598.8226318359375,
|
| 670 |
+
"eval_runtime": 1371.8671,
|
| 671 |
+
"eval_samples_per_second": 14.124,
|
| 672 |
+
"eval_steps_per_second": 14.124,
|
| 673 |
+
"step": 6500
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"epoch": 2.0433528808376398,
|
| 677 |
+
"grad_norm": 43.25,
|
| 678 |
+
"learning_rate": 2.307466341839918e-06,
|
| 679 |
+
"loss": 0.3691,
|
| 680 |
+
"step": 6600
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 2.0743192242930966,
|
| 684 |
+
"grad_norm": 51.0,
|
| 685 |
+
"learning_rate": 2.1723122203058867e-06,
|
| 686 |
+
"loss": 0.3404,
|
| 687 |
+
"step": 6700
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 2.1052855677485534,
|
| 691 |
+
"grad_norm": 28.5,
|
| 692 |
+
"learning_rate": 2.040130076870296e-06,
|
| 693 |
+
"loss": 0.3692,
|
| 694 |
+
"step": 6800
|
| 695 |
+
},
|
| 696 |
+
{
|
| 697 |
+
"epoch": 2.13625191120401,
|
| 698 |
+
"grad_norm": 66.5,
|
| 699 |
+
"learning_rate": 1.9110588386233686e-06,
|
| 700 |
+
"loss": 0.3788,
|
| 701 |
+
"step": 6900
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 2.167218254659467,
|
| 705 |
+
"grad_norm": 83.0,
|
| 706 |
+
"learning_rate": 1.785234163007899e-06,
|
| 707 |
+
"loss": 0.3582,
|
| 708 |
+
"step": 7000
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 2.167218254659467,
|
| 712 |
+
"eval_logits/chosen": NaN,
|
| 713 |
+
"eval_logits/rejected": NaN,
|
| 714 |
+
"eval_logps/chosen": -8605.3466796875,
|
| 715 |
+
"eval_logps/rejected": -6316.7080078125,
|
| 716 |
+
"eval_loss": 315.78204345703125,
|
| 717 |
+
"eval_rewards/accuracies": 0.3188480734825134,
|
| 718 |
+
"eval_rewards/chosen": -815.16357421875,
|
| 719 |
+
"eval_rewards/margins": -216.9755096435547,
|
| 720 |
+
"eval_rewards/rejected": -598.1881103515625,
|
| 721 |
+
"eval_runtime": 1363.6925,
|
| 722 |
+
"eval_samples_per_second": 14.208,
|
| 723 |
+
"eval_steps_per_second": 14.208,
|
| 724 |
+
"step": 7000
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"epoch": 2.1981845981149237,
|
| 728 |
+
"grad_norm": 49.5,
|
| 729 |
+
"learning_rate": 1.6627882952395197e-06,
|
| 730 |
+
"loss": 0.3762,
|
| 731 |
+
"step": 7100
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 2.2291509415703805,
|
| 735 |
+
"grad_norm": 43.25,
|
| 736 |
+
"learning_rate": 1.543849929313328e-06,
|
| 737 |
+
"loss": 0.368,
|
| 738 |
+
"step": 7200
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"epoch": 2.2601172850258378,
|
| 742 |
+
"grad_norm": 43.0,
|
| 743 |
+
"learning_rate": 1.4285440727429296e-06,
|
| 744 |
+
"loss": 0.3496,
|
| 745 |
+
"step": 7300
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 2.2910836284812945,
|
| 749 |
+
"grad_norm": 53.0,
|
| 750 |
+
"learning_rate": 1.3169919151740884e-06,
|
| 751 |
+
"loss": 0.3826,
|
| 752 |
+
"step": 7400
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 2.3220499719367513,
|
| 756 |
+
"grad_norm": 42.5,
|
| 757 |
+
"learning_rate": 1.2093107010110516e-06,
|
| 758 |
+
"loss": 0.3704,
|
| 759 |
+
"step": 7500
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 2.3220499719367513,
|
| 763 |
+
"eval_logits/chosen": NaN,
|
| 764 |
+
"eval_logits/rejected": NaN,
|
| 765 |
+
"eval_logps/chosen": -8608.7568359375,
|
| 766 |
+
"eval_logps/rejected": -6319.107421875,
|
| 767 |
+
"eval_loss": 315.918701171875,
|
| 768 |
+
"eval_rewards/accuracies": 0.31920933723449707,
|
| 769 |
+
"eval_rewards/chosen": -815.5045166015625,
|
| 770 |
+
"eval_rewards/margins": -217.0764617919922,
|
| 771 |
+
"eval_rewards/rejected": -598.4281616210938,
|
| 772 |
+
"eval_runtime": 1364.6449,
|
| 773 |
+
"eval_samples_per_second": 14.199,
|
| 774 |
+
"eval_steps_per_second": 14.199,
|
| 775 |
+
"step": 7500
|
| 776 |
+
},
|
| 777 |
+
{
|
| 778 |
+
"epoch": 2.353016315392208,
|
| 779 |
+
"grad_norm": 52.75,
|
| 780 |
+
"learning_rate": 1.1056136061894386e-06,
|
| 781 |
+
"loss": 0.3592,
|
| 782 |
+
"step": 7600
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 2.383982658847665,
|
| 786 |
+
"grad_norm": 61.5,
|
| 787 |
+
"learning_rate": 1.006009619225199e-06,
|
| 788 |
+
"loss": 0.3523,
|
| 789 |
+
"step": 7700
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"epoch": 2.4149490023031217,
|
| 793 |
+
"grad_norm": 40.75,
|
| 794 |
+
"learning_rate": 9.106034266646735e-07,
|
| 795 |
+
"loss": 0.3726,
|
| 796 |
+
"step": 7800
|
| 797 |
+
},
|
| 798 |
+
{
|
| 799 |
+
"epoch": 2.4459153457585785,
|
| 800 |
+
"grad_norm": 46.75,
|
| 801 |
+
"learning_rate": 8.194953030561226e-07,
|
| 802 |
+
"loss": 0.3816,
|
| 803 |
+
"step": 7900
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 2.4768816892140357,
|
| 807 |
+
"grad_norm": 78.0,
|
| 808 |
+
"learning_rate": 7.327810055584211e-07,
|
| 809 |
+
"loss": 0.3597,
|
| 810 |
+
"step": 8000
|
| 811 |
+
},
|
| 812 |
+
{
|
| 813 |
+
"epoch": 2.4768816892140357,
|
| 814 |
+
"eval_logits/chosen": NaN,
|
| 815 |
+
"eval_logits/rejected": NaN,
|
| 816 |
+
"eval_logps/chosen": -8607.6982421875,
|
| 817 |
+
"eval_logps/rejected": -6318.22314453125,
|
| 818 |
+
"eval_loss": 315.8908386230469,
|
| 819 |
+
"eval_rewards/accuracies": 0.31874483823776245,
|
| 820 |
+
"eval_rewards/chosen": -815.398681640625,
|
| 821 |
+
"eval_rewards/margins": -217.05902099609375,
|
| 822 |
+
"eval_rewards/rejected": -598.3396606445312,
|
| 823 |
+
"eval_runtime": 1361.5153,
|
| 824 |
+
"eval_samples_per_second": 14.231,
|
| 825 |
+
"eval_steps_per_second": 14.231,
|
| 826 |
+
"step": 8000
|
| 827 |
+
},
|
| 828 |
+
{
|
| 829 |
+
"epoch": 2.507848032669492,
|
| 830 |
+
"grad_norm": 46.75,
|
| 831 |
+
"learning_rate": 6.505516732976153e-07,
|
| 832 |
+
"loss": 0.3639,
|
| 833 |
+
"step": 8100
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 2.5388143761249493,
|
| 837 |
+
"grad_norm": 46.25,
|
| 838 |
+
"learning_rate": 5.728937315771954e-07,
|
| 839 |
+
"loss": 0.3778,
|
| 840 |
+
"step": 8200
|
| 841 |
+
},
|
| 842 |
+
{
|
| 843 |
+
"epoch": 2.569780719580406,
|
| 844 |
+
"grad_norm": 100.5,
|
| 845 |
+
"learning_rate": 4.99888801042701e-07,
|
| 846 |
+
"loss": 0.3645,
|
| 847 |
+
"step": 8300
|
| 848 |
+
},
|
| 849 |
+
{
|
| 850 |
+
"epoch": 2.600747063035863,
|
| 851 |
+
"grad_norm": 51.5,
|
| 852 |
+
"learning_rate": 4.316136118961656e-07,
|
| 853 |
+
"loss": 0.3746,
|
| 854 |
+
"step": 8400
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 2.6317134064913197,
|
| 858 |
+
"grad_norm": 67.5,
|
| 859 |
+
"learning_rate": 3.6813992325055504e-07,
|
| 860 |
+
"loss": 0.366,
|
| 861 |
+
"step": 8500
|
| 862 |
+
},
|
| 863 |
+
{
|
| 864 |
+
"epoch": 2.6317134064913197,
|
| 865 |
+
"eval_logits/chosen": NaN,
|
| 866 |
+
"eval_logits/rejected": NaN,
|
| 867 |
+
"eval_logps/chosen": -8610.2001953125,
|
| 868 |
+
"eval_logps/rejected": -6320.1904296875,
|
| 869 |
+
"eval_loss": 315.9734191894531,
|
| 870 |
+
"eval_rewards/accuracies": 0.3197254240512848,
|
| 871 |
+
"eval_rewards/chosen": -815.64892578125,
|
| 872 |
+
"eval_rewards/margins": -217.1125030517578,
|
| 873 |
+
"eval_rewards/rejected": -598.5364379882812,
|
| 874 |
+
"eval_runtime": 1480.2365,
|
| 875 |
+
"eval_samples_per_second": 13.09,
|
| 876 |
+
"eval_steps_per_second": 13.09,
|
| 877 |
+
"step": 8500
|
| 878 |
+
},
|
| 879 |
+
{
|
| 880 |
+
"epoch": 2.6626797499467765,
|
| 881 |
+
"grad_norm": 47.75,
|
| 882 |
+
"learning_rate": 3.095344477089462e-07,
|
| 883 |
+
"loss": 0.3551,
|
| 884 |
+
"step": 8600
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 2.6936460934022337,
|
| 888 |
+
"grad_norm": 54.75,
|
| 889 |
+
"learning_rate": 2.5585878124774754e-07,
|
| 890 |
+
"loss": 0.3719,
|
| 891 |
+
"step": 8700
|
| 892 |
+
},
|
| 893 |
+
{
|
| 894 |
+
"epoch": 2.72461243685769,
|
| 895 |
+
"grad_norm": 51.0,
|
| 896 |
+
"learning_rate": 2.0716933847761134e-07,
|
| 897 |
+
"loss": 0.3659,
|
| 898 |
+
"step": 8800
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 2.7555787803131473,
|
| 902 |
+
"grad_norm": 47.5,
|
| 903 |
+
"learning_rate": 1.6351729335012334e-07,
|
| 904 |
+
"loss": 0.3829,
|
| 905 |
+
"step": 8900
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 2.786545123768604,
|
| 909 |
+
"grad_norm": 63.5,
|
| 910 |
+
"learning_rate": 1.2494852537256296e-07,
|
| 911 |
+
"loss": 0.3503,
|
| 912 |
+
"step": 9000
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"epoch": 2.786545123768604,
|
| 916 |
+
"eval_logits/chosen": NaN,
|
| 917 |
+
"eval_logits/rejected": NaN,
|
| 918 |
+
"eval_logps/chosen": -8609.8154296875,
|
| 919 |
+
"eval_logps/rejected": -6319.88720703125,
|
| 920 |
+
"eval_loss": 315.95770263671875,
|
| 921 |
+
"eval_rewards/accuracies": 0.3191577196121216,
|
| 922 |
+
"eval_rewards/chosen": -815.6104736328125,
|
| 923 |
+
"eval_rewards/margins": -217.10435485839844,
|
| 924 |
+
"eval_rewards/rejected": -598.5061645507812,
|
| 925 |
+
"eval_runtime": 1438.727,
|
| 926 |
+
"eval_samples_per_second": 13.467,
|
| 927 |
+
"eval_steps_per_second": 13.467,
|
| 928 |
+
"step": 9000
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"epoch": 2.817511467224061,
|
| 932 |
+
"grad_norm": 46.5,
|
| 933 |
+
"learning_rate": 9.150357138727028e-08,
|
| 934 |
+
"loss": 0.3738,
|
| 935 |
+
"step": 9100
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 2.8484778106795177,
|
| 939 |
+
"grad_norm": 52.5,
|
| 940 |
+
"learning_rate": 6.321758296630398e-08,
|
| 941 |
+
"loss": 0.3754,
|
| 942 |
+
"step": 9200
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 2.8794441541349745,
|
| 946 |
+
"grad_norm": 68.5,
|
| 947 |
+
"learning_rate": 4.0120289466166754e-08,
|
| 948 |
+
"loss": 0.3747,
|
| 949 |
+
"step": 9300
|
| 950 |
+
},
|
| 951 |
+
{
|
| 952 |
+
"epoch": 2.9104104975904312,
|
| 953 |
+
"grad_norm": 46.75,
|
| 954 |
+
"learning_rate": 2.2235966781427586e-08,
|
| 955 |
+
"loss": 0.3565,
|
| 956 |
+
"step": 9400
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 2.941376841045888,
|
| 960 |
+
"grad_norm": 58.0,
|
| 961 |
+
"learning_rate": 9.583411830087485e-09,
|
| 962 |
+
"loss": 0.4153,
|
| 963 |
+
"step": 9500
|
| 964 |
+
},
|
| 965 |
+
{
|
| 966 |
+
"epoch": 2.941376841045888,
|
| 967 |
+
"eval_logits/chosen": NaN,
|
| 968 |
+
"eval_logits/rejected": NaN,
|
| 969 |
+
"eval_logps/chosen": -8609.7314453125,
|
| 970 |
+
"eval_logps/rejected": -6319.83056640625,
|
| 971 |
+
"eval_loss": 315.9565734863281,
|
| 972 |
+
"eval_rewards/accuracies": 0.3196222186088562,
|
| 973 |
+
"eval_rewards/chosen": -815.6021728515625,
|
| 974 |
+
"eval_rewards/margins": -217.1017303466797,
|
| 975 |
+
"eval_rewards/rejected": -598.50048828125,
|
| 976 |
+
"eval_runtime": 1365.7529,
|
| 977 |
+
"eval_samples_per_second": 14.187,
|
| 978 |
+
"eval_steps_per_second": 14.187,
|
| 979 |
+
"step": 9500
|
| 980 |
+
},
|
| 981 |
+
{
|
| 982 |
+
"epoch": 2.9723431845013453,
|
| 983 |
+
"grad_norm": 59.25,
|
| 984 |
+
"learning_rate": 2.1759227974949006e-09,
|
| 985 |
+
"loss": 0.3534,
|
| 986 |
+
"step": 9600
|
| 987 |
+
}
|
| 988 |
+
],
|
| 989 |
+
"logging_steps": 100,
|
| 990 |
+
"max_steps": 9690,
|
| 991 |
+
"num_input_tokens_seen": 0,
|
| 992 |
+
"num_train_epochs": 3,
|
| 993 |
+
"save_steps": 500,
|
| 994 |
+
"stateful_callbacks": {
|
| 995 |
+
"TrainerControl": {
|
| 996 |
+
"args": {
|
| 997 |
+
"should_epoch_stop": false,
|
| 998 |
+
"should_evaluate": false,
|
| 999 |
+
"should_log": false,
|
| 1000 |
+
"should_save": true,
|
| 1001 |
+
"should_training_stop": true
|
| 1002 |
+
},
|
| 1003 |
+
"attributes": {}
|
| 1004 |
+
}
|
| 1005 |
+
},
|
| 1006 |
+
"total_flos": 0.0,
|
| 1007 |
+
"train_batch_size": 1,
|
| 1008 |
+
"trial_name": null,
|
| 1009 |
+
"trial_params": null
|
| 1010 |
+
}
|