Commit
•
aff76c5
1
Parent(s):
5cc90cd
Model save
Browse files- README.md +32 -22
- adapter_model.safetensors +1 -1
- all_results.json +5 -5
- train_results.json +5 -5
- trainer_state.json +2030 -90
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
| 4 |
-
- alignment-handbook
|
| 5 |
- trl
|
| 6 |
- dpo
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
| 9 |
-
datasets:
|
| 10 |
-
- hugodk-sch/aftonposten_title_prefs
|
| 11 |
model-index:
|
| 12 |
- name: aftonposten-6b-align-scan
|
| 13 |
results: []
|
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
# aftonposten-6b-align-scan
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 1.
|
| 24 |
-
- Rewards/chosen:
|
| 25 |
-
- Rewards/rejected:
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,21 +55,33 @@ The following hyperparameters were used during training:
|
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
| 59 |
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
-
- num_epochs:
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
-
| Training Loss | Epoch | Step |
|
| 65 |
-
|
| 66 |
-
| 1.4583 | 0.26 | 100 |
|
| 67 |
-
| 1.279 | 0.52 | 200 |
|
| 68 |
-
| 1.0043 | 0.78 | 300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
| 72 |
|
| 73 |
-
- PEFT 0.
|
| 74 |
-
- Transformers 4.
|
| 75 |
- Pytorch 2.1.2+cu121
|
| 76 |
-
- Datasets 2.
|
| 77 |
- Tokenizers 0.15.1
|
|
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
|
|
|
| 4 |
- trl
|
| 5 |
- dpo
|
| 6 |
+
- alignment-handbook
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
|
|
|
|
|
|
| 9 |
model-index:
|
| 10 |
- name: aftonposten-6b-align-scan
|
| 11 |
results: []
|
|
|
|
| 16 |
|
| 17 |
# aftonposten-6b-align-scan
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 1.6201
|
| 22 |
+
- Rewards/chosen: 0.0327
|
| 23 |
+
- Rewards/rejected: 0.0149
|
| 24 |
+
- Rewards/accuracies: 0.5249
|
| 25 |
+
- Rewards/margins: 0.0178
|
| 26 |
+
- Logps/rejected: -37.4793
|
| 27 |
+
- Logps/chosen: -33.9527
|
| 28 |
+
- Logits/rejected: -2.2332
|
| 29 |
+
- Logits/chosen: -2.2381
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 55 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
- lr_scheduler_type: cosine
|
| 57 |
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 4
|
| 59 |
|
| 60 |
### Training results
|
| 61 |
|
| 62 |
+
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
| 63 |
+
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
| 64 |
+
| 1.4583 | 0.26 | 100 | -2.2357 | -2.2308 | -34.0303 | -37.5236 | 1.6162 | 0.5245 | 0.0017 | 0.0045 | -0.0028 |
|
| 65 |
+
| 1.279 | 0.52 | 200 | -2.2359 | -2.2311 | -34.0825 | -37.5923 | 1.6100 | 0.5257 | -0.0192 | 0.0111 | -0.0303 |
|
| 66 |
+
| 1.0043 | 0.78 | 300 | -2.2312 | -2.2263 | -34.0845 | -37.6004 | 1.5962 | 0.5166 | -0.0200 | 0.0135 | -0.0335 |
|
| 67 |
+
| 0.7239 | 1.04 | 400 | 1.6461 | -0.0219 | -0.0311 | 0.5341 | 0.0092 | -37.5945 | -34.0893 | -2.2276 | -2.2324 |
|
| 68 |
+
| 0.6061 | 1.3 | 500 | 1.6487 | -0.0274 | -0.0429 | 0.5395 | 0.0155 | -37.6239 | -34.1030 | -2.2282 | -2.2330 |
|
| 69 |
+
| 0.9255 | 1.56 | 600 | 1.5912 | 0.0108 | -0.0119 | 0.5544 | 0.0228 | -37.5464 | -34.0074 | -2.2273 | -2.2321 |
|
| 70 |
+
| 0.8252 | 1.82 | 700 | 1.6334 | 0.0226 | 0.0045 | 0.5216 | 0.0180 | -37.5053 | -33.9781 | -2.2298 | -2.2346 |
|
| 71 |
+
| 0.2848 | 2.08 | 800 | 1.6033 | 0.0153 | -0.0031 | 0.5249 | 0.0184 | -37.5244 | -33.9964 | -2.2313 | -2.2361 |
|
| 72 |
+
| 0.3671 | 2.34 | 900 | 1.6569 | 0.0283 | 0.0177 | 0.5162 | 0.0106 | -37.4723 | -33.9637 | -2.2309 | -2.2358 |
|
| 73 |
+
| 0.3936 | 2.6 | 1000 | 1.6203 | 0.0348 | 0.0187 | 0.5428 | 0.0161 | -37.4698 | -33.9475 | -2.2325 | -2.2374 |
|
| 74 |
+
| 0.3156 | 2.86 | 1100 | 1.6012 | 0.0302 | 0.0108 | 0.5606 | 0.0194 | -37.4896 | -33.9592 | -2.2326 | -2.2375 |
|
| 75 |
+
| 0.2893 | 3.12 | 1200 | 1.5705 | 0.0346 | 0.0103 | 0.5365 | 0.0243 | -37.4909 | -33.9480 | -2.2335 | -2.2383 |
|
| 76 |
+
| 0.277 | 3.38 | 1300 | 1.6102 | 0.0314 | 0.0121 | 0.5403 | 0.0194 | -37.4865 | -33.9559 | -2.2333 | -2.2382 |
|
| 77 |
+
| 0.139 | 3.64 | 1400 | 1.6181 | 0.0273 | 0.0092 | 0.5307 | 0.0181 | -37.4937 | -33.9663 | -2.2333 | -2.2381 |
|
| 78 |
+
| 0.24 | 3.9 | 1500 | 1.6201 | 0.0327 | 0.0149 | 0.5249 | 0.0178 | -37.4793 | -33.9527 | -2.2332 | -2.2381 |
|
| 79 |
|
| 80 |
|
| 81 |
### Framework versions
|
| 82 |
|
| 83 |
+
- PEFT 0.10.0
|
| 84 |
+
- Transformers 4.39.0.dev0
|
| 85 |
- Pytorch 2.1.2+cu121
|
| 86 |
+
- Datasets 2.14.6
|
| 87 |
- Tokenizers 0.15.1
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 176183216
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c47bb1a324b5553592841a0c57e589c161272bd4e084e17244ed50fd058977a
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.2312498092651367,
|
| 4 |
"eval_logits/rejected": -2.2264044284820557,
|
| 5 |
"eval_logps/chosen": -34.088890075683594,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.355,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
-
"train_loss":
|
| 17 |
-
"train_runtime":
|
| 18 |
"train_samples": 3079,
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
"eval_logits/chosen": -2.2312498092651367,
|
| 4 |
"eval_logits/rejected": -2.2264044284820557,
|
| 5 |
"eval_logps/chosen": -34.088890075683594,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.355,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.36088944433571457,
|
| 17 |
+
"train_runtime": 10769.3493,
|
| 18 |
"train_samples": 3079,
|
| 19 |
+
"train_samples_per_second": 1.144,
|
| 20 |
+
"train_steps_per_second": 0.143
|
| 21 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 3079,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
+
"train_loss": 0.36088944433571457,
|
| 4 |
+
"train_runtime": 10769.3493,
|
| 5 |
"train_samples": 3079,
|
| 6 |
+
"train_samples_per_second": 1.144,
|
| 7 |
+
"train_steps_per_second": 0.143
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
@@ -492,130 +492,2070 @@
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
-
"
|
| 496 |
-
"
|
| 497 |
-
"logits/
|
| 498 |
-
"
|
| 499 |
-
"logps/
|
| 500 |
-
"
|
| 501 |
-
"
|
| 502 |
-
"rewards/
|
| 503 |
-
"rewards/
|
| 504 |
-
"rewards/
|
|
|
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
-
"
|
| 510 |
-
"
|
| 511 |
-
"logits/
|
| 512 |
-
"
|
| 513 |
-
"logps/
|
| 514 |
-
"
|
| 515 |
-
"
|
| 516 |
-
"rewards/
|
| 517 |
-
"rewards/
|
| 518 |
-
"rewards/
|
|
|
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
-
"
|
| 524 |
-
"
|
| 525 |
-
"logits/
|
| 526 |
-
"
|
| 527 |
-
"logps/
|
| 528 |
-
"
|
| 529 |
-
"
|
| 530 |
-
"rewards/
|
| 531 |
-
"rewards/
|
| 532 |
-
"rewards/
|
|
|
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"
|
| 538 |
-
"
|
| 539 |
-
"logits/
|
| 540 |
-
"
|
| 541 |
-
"logps/
|
| 542 |
-
"
|
| 543 |
-
"
|
| 544 |
-
"rewards/
|
| 545 |
-
"rewards/
|
| 546 |
-
"rewards/
|
|
|
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
-
"
|
| 552 |
-
"
|
| 553 |
-
"logits/
|
| 554 |
-
"
|
| 555 |
-
"logps/
|
| 556 |
-
"
|
| 557 |
-
"
|
| 558 |
-
"rewards/
|
| 559 |
-
"rewards/
|
| 560 |
-
"rewards/
|
|
|
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
-
"
|
| 566 |
-
"
|
| 567 |
-
"logits/
|
| 568 |
-
"
|
| 569 |
-
"logps/
|
| 570 |
-
"
|
| 571 |
-
"
|
| 572 |
-
"rewards/
|
| 573 |
-
"rewards/
|
| 574 |
-
"rewards/
|
|
|
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
-
"
|
| 580 |
-
"
|
| 581 |
-
"logits/
|
| 582 |
-
"
|
| 583 |
-
"logps/
|
| 584 |
-
"
|
| 585 |
-
"
|
| 586 |
-
"rewards/
|
| 587 |
-
"rewards/
|
| 588 |
-
"rewards/
|
|
|
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
-
"
|
| 594 |
-
"
|
| 595 |
-
"logits/
|
| 596 |
-
"
|
| 597 |
-
"logps/
|
| 598 |
-
"
|
| 599 |
-
"
|
| 600 |
-
"rewards/
|
| 601 |
-
"rewards/
|
| 602 |
-
"rewards/
|
|
|
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
-
"epoch": 1.
|
| 607 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 608 |
"total_flos": 0.0,
|
| 609 |
-
"train_loss":
|
| 610 |
-
"train_runtime":
|
| 611 |
-
"train_samples_per_second":
|
| 612 |
-
"train_steps_per_second": 0.
|
| 613 |
}
|
| 614 |
],
|
| 615 |
"logging_steps": 10,
|
| 616 |
-
"max_steps":
|
| 617 |
"num_input_tokens_seen": 0,
|
| 618 |
-
"num_train_epochs":
|
| 619 |
"save_steps": 100,
|
| 620 |
"total_flos": 0.0,
|
| 621 |
"train_batch_size": 4,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 4.0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 1540,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
+
"grad_norm": 104.0,
|
| 496 |
+
"learning_rate": 4.84533120650964e-06,
|
| 497 |
+
"logits/chosen": -2.0658023357391357,
|
| 498 |
+
"logits/rejected": -2.0529816150665283,
|
| 499 |
+
"logps/chosen": -32.19905090332031,
|
| 500 |
+
"logps/rejected": -32.87395477294922,
|
| 501 |
+
"loss": 0.7465,
|
| 502 |
+
"rewards/accuracies": 0.887499988079071,
|
| 503 |
+
"rewards/chosen": 0.13173677027225494,
|
| 504 |
+
"rewards/margins": 0.22489500045776367,
|
| 505 |
+
"rewards/rejected": -0.09315825998783112,
|
| 506 |
"step": 310
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.83,
|
| 510 |
+
"grad_norm": 42.5,
|
| 511 |
+
"learning_rate": 4.825108134172131e-06,
|
| 512 |
+
"logits/chosen": -1.9775402545928955,
|
| 513 |
+
"logits/rejected": -1.9688619375228882,
|
| 514 |
+
"logps/chosen": -31.920490264892578,
|
| 515 |
+
"logps/rejected": -30.43929672241211,
|
| 516 |
+
"loss": 0.712,
|
| 517 |
+
"rewards/accuracies": 0.875,
|
| 518 |
+
"rewards/chosen": 0.17379958927631378,
|
| 519 |
+
"rewards/margins": 0.2529620826244354,
|
| 520 |
+
"rewards/rejected": -0.07916250824928284,
|
| 521 |
"step": 320
|
| 522 |
},
|
| 523 |
{
|
| 524 |
"epoch": 0.86,
|
| 525 |
+
"grad_norm": 88.0,
|
| 526 |
+
"learning_rate": 4.80369052967602e-06,
|
| 527 |
+
"logits/chosen": -1.916343331336975,
|
| 528 |
+
"logits/rejected": -1.928449273109436,
|
| 529 |
+
"logps/chosen": -29.984766006469727,
|
| 530 |
+
"logps/rejected": -33.69157791137695,
|
| 531 |
+
"loss": 0.603,
|
| 532 |
+
"rewards/accuracies": 0.875,
|
| 533 |
+
"rewards/chosen": 0.19780051708221436,
|
| 534 |
+
"rewards/margins": 0.3135302662849426,
|
| 535 |
+
"rewards/rejected": -0.11572977155447006,
|
| 536 |
"step": 330
|
| 537 |
},
|
| 538 |
{
|
| 539 |
"epoch": 0.88,
|
| 540 |
+
"grad_norm": 93.0,
|
| 541 |
+
"learning_rate": 4.781089396387968e-06,
|
| 542 |
+
"logits/chosen": -1.8849292993545532,
|
| 543 |
+
"logits/rejected": -1.8756763935089111,
|
| 544 |
+
"logps/chosen": -34.16048812866211,
|
| 545 |
+
"logps/rejected": -36.131134033203125,
|
| 546 |
+
"loss": 0.854,
|
| 547 |
+
"rewards/accuracies": 0.887499988079071,
|
| 548 |
+
"rewards/chosen": 0.21537461876869202,
|
| 549 |
+
"rewards/margins": 0.32513895630836487,
|
| 550 |
+
"rewards/rejected": -0.10976432263851166,
|
| 551 |
"step": 340
|
| 552 |
},
|
| 553 |
{
|
| 554 |
"epoch": 0.91,
|
| 555 |
+
"grad_norm": 64.0,
|
| 556 |
+
"learning_rate": 4.757316345716554e-06,
|
| 557 |
+
"logits/chosen": -1.94269597530365,
|
| 558 |
+
"logits/rejected": -1.9433438777923584,
|
| 559 |
+
"logps/chosen": -33.91167068481445,
|
| 560 |
+
"logps/rejected": -34.086830139160156,
|
| 561 |
+
"loss": 0.6834,
|
| 562 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 563 |
+
"rewards/chosen": 0.1974446177482605,
|
| 564 |
+
"rewards/margins": 0.29258665442466736,
|
| 565 |
+
"rewards/rejected": -0.09514202922582626,
|
| 566 |
"step": 350
|
| 567 |
},
|
| 568 |
{
|
| 569 |
"epoch": 0.94,
|
| 570 |
+
"grad_norm": 81.0,
|
| 571 |
+
"learning_rate": 4.73238359114687e-06,
|
| 572 |
+
"logits/chosen": -2.0718166828155518,
|
| 573 |
+
"logits/rejected": -2.0780444145202637,
|
| 574 |
+
"logps/chosen": -31.210739135742188,
|
| 575 |
+
"logps/rejected": -32.90790557861328,
|
| 576 |
+
"loss": 0.7717,
|
| 577 |
+
"rewards/accuracies": 0.824999988079071,
|
| 578 |
+
"rewards/chosen": 0.14500285685062408,
|
| 579 |
+
"rewards/margins": 0.24085240066051483,
|
| 580 |
+
"rewards/rejected": -0.09584951400756836,
|
| 581 |
"step": 360
|
| 582 |
},
|
| 583 |
{
|
| 584 |
"epoch": 0.96,
|
| 585 |
+
"grad_norm": 200.0,
|
| 586 |
+
"learning_rate": 4.706303941965804e-06,
|
| 587 |
+
"logits/chosen": -2.0039541721343994,
|
| 588 |
+
"logits/rejected": -2.0034594535827637,
|
| 589 |
+
"logps/chosen": -32.95136260986328,
|
| 590 |
+
"logps/rejected": -36.23757553100586,
|
| 591 |
+
"loss": 0.7959,
|
| 592 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 593 |
+
"rewards/chosen": 0.19742996990680695,
|
| 594 |
+
"rewards/margins": 0.2725493013858795,
|
| 595 |
+
"rewards/rejected": -0.07511930912733078,
|
| 596 |
"step": 370
|
| 597 |
},
|
| 598 |
{
|
| 599 |
"epoch": 0.99,
|
| 600 |
+
"grad_norm": 67.0,
|
| 601 |
+
"learning_rate": 4.679090796681225e-06,
|
| 602 |
+
"logits/chosen": -2.038403272628784,
|
| 603 |
+
"logits/rejected": -2.0337538719177246,
|
| 604 |
+
"logps/chosen": -30.219844818115234,
|
| 605 |
+
"logps/rejected": -29.44011878967285,
|
| 606 |
+
"loss": 0.6897,
|
| 607 |
+
"rewards/accuracies": 0.887499988079071,
|
| 608 |
+
"rewards/chosen": 0.1963024139404297,
|
| 609 |
+
"rewards/margins": 0.286792516708374,
|
| 610 |
+
"rewards/rejected": -0.09049010276794434,
|
| 611 |
"step": 380
|
| 612 |
},
|
| 613 |
{
|
| 614 |
+
"epoch": 1.01,
|
| 615 |
+
"grad_norm": 68.0,
|
| 616 |
+
"learning_rate": 4.650758136138454e-06,
|
| 617 |
+
"logits/chosen": -1.8102906942367554,
|
| 618 |
+
"logits/rejected": -1.8167164325714111,
|
| 619 |
+
"logps/chosen": -31.897790908813477,
|
| 620 |
+
"logps/rejected": -36.53219985961914,
|
| 621 |
+
"loss": 0.8214,
|
| 622 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 623 |
+
"rewards/chosen": 0.1951691210269928,
|
| 624 |
+
"rewards/margins": 0.36073189973831177,
|
| 625 |
+
"rewards/rejected": -0.16556277871131897,
|
| 626 |
+
"step": 390
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 1.04,
|
| 630 |
+
"grad_norm": 54.0,
|
| 631 |
+
"learning_rate": 4.621320516337559e-06,
|
| 632 |
+
"logits/chosen": -1.9675865173339844,
|
| 633 |
+
"logits/rejected": -1.961277723312378,
|
| 634 |
+
"logps/chosen": -33.26747131347656,
|
| 635 |
+
"logps/rejected": -32.52979278564453,
|
| 636 |
+
"loss": 0.7239,
|
| 637 |
+
"rewards/accuracies": 0.824999988079071,
|
| 638 |
+
"rewards/chosen": 0.226090669631958,
|
| 639 |
+
"rewards/margins": 0.3144562244415283,
|
| 640 |
+
"rewards/rejected": -0.08836553245782852,
|
| 641 |
+
"step": 400
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"epoch": 1.04,
|
| 645 |
+
"eval_logits/chosen": -2.2324159145355225,
|
| 646 |
+
"eval_logits/rejected": -2.227566957473755,
|
| 647 |
+
"eval_logps/chosen": -34.08934783935547,
|
| 648 |
+
"eval_logps/rejected": -37.594459533691406,
|
| 649 |
+
"eval_loss": 1.646134853363037,
|
| 650 |
+
"eval_rewards/accuracies": 0.5340532064437866,
|
| 651 |
+
"eval_rewards/chosen": -0.02191990241408348,
|
| 652 |
+
"eval_rewards/margins": 0.009215297177433968,
|
| 653 |
+
"eval_rewards/rejected": -0.03113519586622715,
|
| 654 |
+
"eval_runtime": 145.6073,
|
| 655 |
+
"eval_samples_per_second": 2.356,
|
| 656 |
+
"eval_steps_per_second": 0.295,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 1.06,
|
| 661 |
+
"grad_norm": 67.5,
|
| 662 |
+
"learning_rate": 4.590793060955158e-06,
|
| 663 |
+
"logits/chosen": -1.9676673412322998,
|
| 664 |
+
"logits/rejected": -1.9748865365982056,
|
| 665 |
+
"logps/chosen": -28.631023406982422,
|
| 666 |
+
"logps/rejected": -29.46906089782715,
|
| 667 |
+
"loss": 0.6247,
|
| 668 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 669 |
+
"rewards/chosen": 0.1957395225763321,
|
| 670 |
+
"rewards/margins": 0.34407514333724976,
|
| 671 |
+
"rewards/rejected": -0.14833566546440125,
|
| 672 |
+
"step": 410
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 1.09,
|
| 676 |
+
"grad_norm": 75.5,
|
| 677 |
+
"learning_rate": 4.559191453574582e-06,
|
| 678 |
+
"logits/chosen": -1.9874868392944336,
|
| 679 |
+
"logits/rejected": -1.986395239830017,
|
| 680 |
+
"logps/chosen": -33.74576187133789,
|
| 681 |
+
"logps/rejected": -31.112499237060547,
|
| 682 |
+
"loss": 0.8861,
|
| 683 |
+
"rewards/accuracies": 0.800000011920929,
|
| 684 |
+
"rewards/chosen": 0.20197901129722595,
|
| 685 |
+
"rewards/margins": 0.2697620093822479,
|
| 686 |
+
"rewards/rejected": -0.06778299063444138,
|
| 687 |
+
"step": 420
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 1.12,
|
| 691 |
+
"grad_norm": 117.0,
|
| 692 |
+
"learning_rate": 4.52653192962838e-06,
|
| 693 |
+
"logits/chosen": -1.9822708368301392,
|
| 694 |
+
"logits/rejected": -1.964942216873169,
|
| 695 |
+
"logps/chosen": -30.64545249938965,
|
| 696 |
+
"logps/rejected": -33.29753112792969,
|
| 697 |
+
"loss": 0.9065,
|
| 698 |
+
"rewards/accuracies": 0.887499988079071,
|
| 699 |
+
"rewards/chosen": 0.13270007073879242,
|
| 700 |
+
"rewards/margins": 0.2811988890171051,
|
| 701 |
+
"rewards/rejected": -0.1484988033771515,
|
| 702 |
+
"step": 430
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 1.14,
|
| 706 |
+
"grad_norm": 49.0,
|
| 707 |
+
"learning_rate": 4.492831268057307e-06,
|
| 708 |
+
"logits/chosen": -2.014035224914551,
|
| 709 |
+
"logits/rejected": -2.016022205352783,
|
| 710 |
+
"logps/chosen": -36.08784866333008,
|
| 711 |
+
"logps/rejected": -35.217376708984375,
|
| 712 |
+
"loss": 0.5593,
|
| 713 |
+
"rewards/accuracies": 0.887499988079071,
|
| 714 |
+
"rewards/chosen": 0.20980596542358398,
|
| 715 |
+
"rewards/margins": 0.35277143120765686,
|
| 716 |
+
"rewards/rejected": -0.14296546578407288,
|
| 717 |
+
"step": 440
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.17,
|
| 721 |
+
"grad_norm": 55.75,
|
| 722 |
+
"learning_rate": 4.458106782690094e-06,
|
| 723 |
+
"logits/chosen": -2.093276262283325,
|
| 724 |
+
"logits/rejected": -2.093129873275757,
|
| 725 |
+
"logps/chosen": -32.13650894165039,
|
| 726 |
+
"logps/rejected": -33.764305114746094,
|
| 727 |
+
"loss": 0.8816,
|
| 728 |
+
"rewards/accuracies": 0.875,
|
| 729 |
+
"rewards/chosen": 0.198737233877182,
|
| 730 |
+
"rewards/margins": 0.2804142236709595,
|
| 731 |
+
"rewards/rejected": -0.08167698234319687,
|
| 732 |
+
"step": 450
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"epoch": 1.19,
|
| 736 |
+
"grad_norm": 100.5,
|
| 737 |
+
"learning_rate": 4.422376313348405e-06,
|
| 738 |
+
"logits/chosen": -2.038951873779297,
|
| 739 |
+
"logits/rejected": -2.0315186977386475,
|
| 740 |
+
"logps/chosen": -31.86245346069336,
|
| 741 |
+
"logps/rejected": -36.433692932128906,
|
| 742 |
+
"loss": 0.7285,
|
| 743 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 744 |
+
"rewards/chosen": 0.21740083396434784,
|
| 745 |
+
"rewards/margins": 0.36926203966140747,
|
| 746 |
+
"rewards/rejected": -0.15186122059822083,
|
| 747 |
+
"step": 460
|
| 748 |
+
},
|
| 749 |
+
{
|
| 750 |
+
"epoch": 1.22,
|
| 751 |
+
"grad_norm": 126.5,
|
| 752 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 753 |
+
"logits/chosen": -1.940791130065918,
|
| 754 |
+
"logits/rejected": -1.9372183084487915,
|
| 755 |
+
"logps/chosen": -33.40504837036133,
|
| 756 |
+
"logps/rejected": -33.189998626708984,
|
| 757 |
+
"loss": 0.609,
|
| 758 |
+
"rewards/accuracies": 0.875,
|
| 759 |
+
"rewards/chosen": 0.21089871227741241,
|
| 760 |
+
"rewards/margins": 0.3173474669456482,
|
| 761 |
+
"rewards/rejected": -0.10644875466823578,
|
| 762 |
+
"step": 470
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 1.25,
|
| 766 |
+
"grad_norm": 104.5,
|
| 767 |
+
"learning_rate": 4.347971356735789e-06,
|
| 768 |
+
"logits/chosen": -2.065157651901245,
|
| 769 |
+
"logits/rejected": -2.0581443309783936,
|
| 770 |
+
"logps/chosen": -30.389184951782227,
|
| 771 |
+
"logps/rejected": -32.29039764404297,
|
| 772 |
+
"loss": 0.7324,
|
| 773 |
+
"rewards/accuracies": 0.875,
|
| 774 |
+
"rewards/chosen": 0.17574059963226318,
|
| 775 |
+
"rewards/margins": 0.29450365900993347,
|
| 776 |
+
"rewards/rejected": -0.11876306682825089,
|
| 777 |
+
"step": 480
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 1.27,
|
| 781 |
+
"grad_norm": 68.0,
|
| 782 |
+
"learning_rate": 4.309335095262675e-06,
|
| 783 |
+
"logits/chosen": -2.0097293853759766,
|
| 784 |
+
"logits/rejected": -2.0112791061401367,
|
| 785 |
+
"logps/chosen": -34.72505569458008,
|
| 786 |
+
"logps/rejected": -33.918052673339844,
|
| 787 |
+
"loss": 0.5935,
|
| 788 |
+
"rewards/accuracies": 0.887499988079071,
|
| 789 |
+
"rewards/chosen": 0.2512527406215668,
|
| 790 |
+
"rewards/margins": 0.3618691861629486,
|
| 791 |
+
"rewards/rejected": -0.11061643064022064,
|
| 792 |
+
"step": 490
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 1.3,
|
| 796 |
+
"grad_norm": 107.0,
|
| 797 |
+
"learning_rate": 4.269769281772082e-06,
|
| 798 |
+
"logits/chosen": -1.8976560831069946,
|
| 799 |
+
"logits/rejected": -1.8952856063842773,
|
| 800 |
+
"logps/chosen": -32.58599090576172,
|
| 801 |
+
"logps/rejected": -36.84335708618164,
|
| 802 |
+
"loss": 0.6061,
|
| 803 |
+
"rewards/accuracies": 0.925000011920929,
|
| 804 |
+
"rewards/chosen": 0.23451784253120422,
|
| 805 |
+
"rewards/margins": 0.38613152503967285,
|
| 806 |
+
"rewards/rejected": -0.15161368250846863,
|
| 807 |
+
"step": 500
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 1.3,
|
| 811 |
+
"eval_logits/chosen": -2.2330281734466553,
|
| 812 |
+
"eval_logits/rejected": -2.228160858154297,
|
| 813 |
+
"eval_logps/chosen": -34.10296630859375,
|
| 814 |
+
"eval_logps/rejected": -37.62388229370117,
|
| 815 |
+
"eval_loss": 1.6487250328063965,
|
| 816 |
+
"eval_rewards/accuracies": 0.5394518375396729,
|
| 817 |
+
"eval_rewards/chosen": -0.02736487239599228,
|
| 818 |
+
"eval_rewards/margins": 0.015538448467850685,
|
| 819 |
+
"eval_rewards/rejected": -0.04290332272648811,
|
| 820 |
+
"eval_runtime": 145.2589,
|
| 821 |
+
"eval_samples_per_second": 2.361,
|
| 822 |
+
"eval_steps_per_second": 0.296,
|
| 823 |
+
"step": 500
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 1.32,
|
| 827 |
+
"grad_norm": 63.5,
|
| 828 |
+
"learning_rate": 4.22929424333435e-06,
|
| 829 |
+
"logits/chosen": -2.002866744995117,
|
| 830 |
+
"logits/rejected": -2.007615566253662,
|
| 831 |
+
"logps/chosen": -32.78615188598633,
|
| 832 |
+
"logps/rejected": -31.923254013061523,
|
| 833 |
+
"loss": 0.7109,
|
| 834 |
+
"rewards/accuracies": 0.887499988079071,
|
| 835 |
+
"rewards/chosen": 0.2684818208217621,
|
| 836 |
+
"rewards/margins": 0.37934139370918274,
|
| 837 |
+
"rewards/rejected": -0.11085958778858185,
|
| 838 |
+
"step": 510
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 1.35,
|
| 842 |
+
"grad_norm": 90.0,
|
| 843 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 844 |
+
"logits/chosen": -2.034381151199341,
|
| 845 |
+
"logits/rejected": -2.0453076362609863,
|
| 846 |
+
"logps/chosen": -30.794418334960938,
|
| 847 |
+
"logps/rejected": -31.83915138244629,
|
| 848 |
+
"loss": 0.6671,
|
| 849 |
+
"rewards/accuracies": 0.9375,
|
| 850 |
+
"rewards/chosen": 0.2383977472782135,
|
| 851 |
+
"rewards/margins": 0.37098294496536255,
|
| 852 |
+
"rewards/rejected": -0.13258522748947144,
|
| 853 |
+
"step": 520
|
| 854 |
+
},
|
| 855 |
+
{
|
| 856 |
+
"epoch": 1.38,
|
| 857 |
+
"grad_norm": 62.25,
|
| 858 |
+
"learning_rate": 4.145700124802693e-06,
|
| 859 |
+
"logits/chosen": -1.9620320796966553,
|
| 860 |
+
"logits/rejected": -1.9585500955581665,
|
| 861 |
+
"logps/chosen": -31.819278717041016,
|
| 862 |
+
"logps/rejected": -32.71375274658203,
|
| 863 |
+
"loss": 0.6704,
|
| 864 |
+
"rewards/accuracies": 0.925000011920929,
|
| 865 |
+
"rewards/chosen": 0.24049386382102966,
|
| 866 |
+
"rewards/margins": 0.38088440895080566,
|
| 867 |
+
"rewards/rejected": -0.14039050042629242,
|
| 868 |
+
"step": 530
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"epoch": 1.4,
|
| 872 |
+
"grad_norm": 50.5,
|
| 873 |
+
"learning_rate": 4.102623991469562e-06,
|
| 874 |
+
"logits/chosen": -1.8313617706298828,
|
| 875 |
+
"logits/rejected": -1.84051513671875,
|
| 876 |
+
"logps/chosen": -32.123390197753906,
|
| 877 |
+
"logps/rejected": -32.177490234375,
|
| 878 |
+
"loss": 0.6783,
|
| 879 |
+
"rewards/accuracies": 0.887499988079071,
|
| 880 |
+
"rewards/chosen": 0.17839959263801575,
|
| 881 |
+
"rewards/margins": 0.30700013041496277,
|
| 882 |
+
"rewards/rejected": -0.12860055267810822,
|
| 883 |
+
"step": 540
|
| 884 |
+
},
|
| 885 |
+
{
|
| 886 |
+
"epoch": 1.43,
|
| 887 |
+
"grad_norm": 72.5,
|
| 888 |
+
"learning_rate": 4.058724504646834e-06,
|
| 889 |
+
"logits/chosen": -1.9312222003936768,
|
| 890 |
+
"logits/rejected": -1.9248243570327759,
|
| 891 |
+
"logps/chosen": -33.194034576416016,
|
| 892 |
+
"logps/rejected": -31.20892906188965,
|
| 893 |
+
"loss": 0.6703,
|
| 894 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 895 |
+
"rewards/chosen": 0.2057083547115326,
|
| 896 |
+
"rewards/margins": 0.29274117946624756,
|
| 897 |
+
"rewards/rejected": -0.08703283220529556,
|
| 898 |
+
"step": 550
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 1.45,
|
| 902 |
+
"grad_norm": 141.0,
|
| 903 |
+
"learning_rate": 4.014024217844167e-06,
|
| 904 |
+
"logits/chosen": -2.0215349197387695,
|
| 905 |
+
"logits/rejected": -2.019446849822998,
|
| 906 |
+
"logps/chosen": -33.937469482421875,
|
| 907 |
+
"logps/rejected": -31.723003387451172,
|
| 908 |
+
"loss": 0.9186,
|
| 909 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 910 |
+
"rewards/chosen": 0.25349074602127075,
|
| 911 |
+
"rewards/margins": 0.34373739361763,
|
| 912 |
+
"rewards/rejected": -0.09024664759635925,
|
| 913 |
+
"step": 560
|
| 914 |
+
},
|
| 915 |
+
{
|
| 916 |
+
"epoch": 1.48,
|
| 917 |
+
"grad_norm": 62.25,
|
| 918 |
+
"learning_rate": 3.968546095984911e-06,
|
| 919 |
+
"logits/chosen": -1.8566452264785767,
|
| 920 |
+
"logits/rejected": -1.8544600009918213,
|
| 921 |
+
"logps/chosen": -32.25889587402344,
|
| 922 |
+
"logps/rejected": -31.205657958984375,
|
| 923 |
+
"loss": 0.8459,
|
| 924 |
+
"rewards/accuracies": 0.8125,
|
| 925 |
+
"rewards/chosen": 0.21500520408153534,
|
| 926 |
+
"rewards/margins": 0.2963467836380005,
|
| 927 |
+
"rewards/rejected": -0.08134157210588455,
|
| 928 |
+
"step": 570
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"epoch": 1.51,
|
| 932 |
+
"grad_norm": 46.75,
|
| 933 |
+
"learning_rate": 3.922313503607806e-06,
|
| 934 |
+
"logits/chosen": -1.9905545711517334,
|
| 935 |
+
"logits/rejected": -1.987287163734436,
|
| 936 |
+
"logps/chosen": -30.5700740814209,
|
| 937 |
+
"logps/rejected": -34.977298736572266,
|
| 938 |
+
"loss": 0.721,
|
| 939 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 940 |
+
"rewards/chosen": 0.21382951736450195,
|
| 941 |
+
"rewards/margins": 0.3542833626270294,
|
| 942 |
+
"rewards/rejected": -0.14045384526252747,
|
| 943 |
+
"step": 580
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.53,
|
| 947 |
+
"grad_norm": 38.0,
|
| 948 |
+
"learning_rate": 3.875350192863368e-06,
|
| 949 |
+
"logits/chosen": -1.9324638843536377,
|
| 950 |
+
"logits/rejected": -1.9361374378204346,
|
| 951 |
+
"logps/chosen": -29.091140747070312,
|
| 952 |
+
"logps/rejected": -30.924423217773438,
|
| 953 |
+
"loss": 0.6955,
|
| 954 |
+
"rewards/accuracies": 0.862500011920929,
|
| 955 |
+
"rewards/chosen": 0.20955555140972137,
|
| 956 |
+
"rewards/margins": 0.31577685475349426,
|
| 957 |
+
"rewards/rejected": -0.1062212809920311,
|
| 958 |
+
"step": 590
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 1.56,
|
| 962 |
+
"grad_norm": 75.5,
|
| 963 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 964 |
+
"logits/chosen": -1.9715423583984375,
|
| 965 |
+
"logits/rejected": -1.9713079929351807,
|
| 966 |
+
"logps/chosen": -31.609033584594727,
|
| 967 |
+
"logps/rejected": -31.35397720336914,
|
| 968 |
+
"loss": 0.9255,
|
| 969 |
+
"rewards/accuracies": 0.875,
|
| 970 |
+
"rewards/chosen": 0.24499893188476562,
|
| 971 |
+
"rewards/margins": 0.3093833923339844,
|
| 972 |
+
"rewards/rejected": -0.06438445299863815,
|
| 973 |
+
"step": 600
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.56,
|
| 977 |
+
"eval_logits/chosen": -2.232149600982666,
|
| 978 |
+
"eval_logits/rejected": -2.227276086807251,
|
| 979 |
+
"eval_logps/chosen": -34.00742721557617,
|
| 980 |
+
"eval_logps/rejected": -37.54637908935547,
|
| 981 |
+
"eval_loss": 1.591224193572998,
|
| 982 |
+
"eval_rewards/accuracies": 0.5544019937515259,
|
| 983 |
+
"eval_rewards/chosen": 0.010849811136722565,
|
| 984 |
+
"eval_rewards/margins": 0.022752860561013222,
|
| 985 |
+
"eval_rewards/rejected": -0.011903048492968082,
|
| 986 |
+
"eval_runtime": 145.3935,
|
| 987 |
+
"eval_samples_per_second": 2.359,
|
| 988 |
+
"eval_steps_per_second": 0.296,
|
| 989 |
+
"step": 600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 1.58,
|
| 993 |
+
"grad_norm": 159.0,
|
| 994 |
+
"learning_rate": 3.7793282895240927e-06,
|
| 995 |
+
"logits/chosen": -2.0326523780822754,
|
| 996 |
+
"logits/rejected": -2.0333547592163086,
|
| 997 |
+
"logps/chosen": -34.38473129272461,
|
| 998 |
+
"logps/rejected": -33.29423141479492,
|
| 999 |
+
"loss": 0.5891,
|
| 1000 |
+
"rewards/accuracies": 0.875,
|
| 1001 |
+
"rewards/chosen": 0.25816404819488525,
|
| 1002 |
+
"rewards/margins": 0.3421815037727356,
|
| 1003 |
+
"rewards/rejected": -0.08401744067668915,
|
| 1004 |
+
"step": 610
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 1.61,
|
| 1008 |
+
"grad_norm": 91.0,
|
| 1009 |
+
"learning_rate": 3.730319028506478e-06,
|
| 1010 |
+
"logits/chosen": -1.9972617626190186,
|
| 1011 |
+
"logits/rejected": -1.994637131690979,
|
| 1012 |
+
"logps/chosen": -32.45311737060547,
|
| 1013 |
+
"logps/rejected": -32.34403991699219,
|
| 1014 |
+
"loss": 0.5185,
|
| 1015 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1016 |
+
"rewards/chosen": 0.2689127027988434,
|
| 1017 |
+
"rewards/margins": 0.378562867641449,
|
| 1018 |
+
"rewards/rejected": -0.10965017974376678,
|
| 1019 |
+
"step": 620
|
| 1020 |
+
},
|
| 1021 |
+
{
|
| 1022 |
+
"epoch": 1.64,
|
| 1023 |
+
"grad_norm": 113.5,
|
| 1024 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 1025 |
+
"logits/chosen": -2.012416124343872,
|
| 1026 |
+
"logits/rejected": -2.003248929977417,
|
| 1027 |
+
"logps/chosen": -32.12730026245117,
|
| 1028 |
+
"logps/rejected": -31.12014389038086,
|
| 1029 |
+
"loss": 0.7937,
|
| 1030 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1031 |
+
"rewards/chosen": 0.2595270574092865,
|
| 1032 |
+
"rewards/margins": 0.33343806862831116,
|
| 1033 |
+
"rewards/rejected": -0.07391100376844406,
|
| 1034 |
+
"step": 630
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 1.66,
|
| 1038 |
+
"grad_norm": 75.0,
|
| 1039 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1040 |
+
"logits/chosen": -2.016336441040039,
|
| 1041 |
+
"logits/rejected": -2.013164758682251,
|
| 1042 |
+
"logps/chosen": -31.525562286376953,
|
| 1043 |
+
"logps/rejected": -32.66303634643555,
|
| 1044 |
+
"loss": 0.4959,
|
| 1045 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1046 |
+
"rewards/chosen": 0.26914453506469727,
|
| 1047 |
+
"rewards/margins": 0.36659201979637146,
|
| 1048 |
+
"rewards/rejected": -0.097447469830513,
|
| 1049 |
+
"step": 640
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 1.69,
|
| 1053 |
+
"grad_norm": 48.0,
|
| 1054 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1055 |
+
"logits/chosen": -2.0207347869873047,
|
| 1056 |
+
"logits/rejected": -2.023106575012207,
|
| 1057 |
+
"logps/chosen": -32.97084426879883,
|
| 1058 |
+
"logps/rejected": -34.033267974853516,
|
| 1059 |
+
"loss": 0.4312,
|
| 1060 |
+
"rewards/accuracies": 0.9375,
|
| 1061 |
+
"rewards/chosen": 0.2731453478336334,
|
| 1062 |
+
"rewards/margins": 0.413092702627182,
|
| 1063 |
+
"rewards/rejected": -0.1399472951889038,
|
| 1064 |
+
"step": 650
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 1.71,
|
| 1068 |
+
"grad_norm": 48.0,
|
| 1069 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1070 |
+
"logits/chosen": -1.934343695640564,
|
| 1071 |
+
"logits/rejected": -1.9348455667495728,
|
| 1072 |
+
"logps/chosen": -33.32109832763672,
|
| 1073 |
+
"logps/rejected": -31.806873321533203,
|
| 1074 |
+
"loss": 0.6159,
|
| 1075 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1076 |
+
"rewards/chosen": 0.25029245018959045,
|
| 1077 |
+
"rewards/margins": 0.33013930916786194,
|
| 1078 |
+
"rewards/rejected": -0.0798468366265297,
|
| 1079 |
+
"step": 660
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 1.74,
|
| 1083 |
+
"grad_norm": 51.0,
|
| 1084 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1085 |
+
"logits/chosen": -1.9735260009765625,
|
| 1086 |
+
"logits/rejected": -1.9637184143066406,
|
| 1087 |
+
"logps/chosen": -33.36211013793945,
|
| 1088 |
+
"logps/rejected": -32.63333511352539,
|
| 1089 |
+
"loss": 0.4666,
|
| 1090 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1091 |
+
"rewards/chosen": 0.25566938519477844,
|
| 1092 |
+
"rewards/margins": 0.38588738441467285,
|
| 1093 |
+
"rewards/rejected": -0.13021798431873322,
|
| 1094 |
+
"step": 670
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 1.77,
|
| 1098 |
+
"grad_norm": 56.0,
|
| 1099 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1100 |
+
"logits/chosen": -1.8723722696304321,
|
| 1101 |
+
"logits/rejected": -1.8691799640655518,
|
| 1102 |
+
"logps/chosen": -30.749767303466797,
|
| 1103 |
+
"logps/rejected": -34.51530838012695,
|
| 1104 |
+
"loss": 0.5406,
|
| 1105 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1106 |
+
"rewards/chosen": 0.2844772934913635,
|
| 1107 |
+
"rewards/margins": 0.3631672263145447,
|
| 1108 |
+
"rewards/rejected": -0.07868987321853638,
|
| 1109 |
+
"step": 680
|
| 1110 |
+
},
|
| 1111 |
+
{
|
| 1112 |
+
"epoch": 1.79,
|
| 1113 |
+
"grad_norm": 96.0,
|
| 1114 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1115 |
+
"logits/chosen": -1.931316614151001,
|
| 1116 |
+
"logits/rejected": -1.9304965734481812,
|
| 1117 |
+
"logps/chosen": -34.53691864013672,
|
| 1118 |
+
"logps/rejected": -35.59252166748047,
|
| 1119 |
+
"loss": 0.5253,
|
| 1120 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1121 |
+
"rewards/chosen": 0.2933480739593506,
|
| 1122 |
+
"rewards/margins": 0.38882699608802795,
|
| 1123 |
+
"rewards/rejected": -0.09547892212867737,
|
| 1124 |
+
"step": 690
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 1.82,
|
| 1128 |
+
"grad_norm": 94.5,
|
| 1129 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1130 |
+
"logits/chosen": -1.9118010997772217,
|
| 1131 |
+
"logits/rejected": -1.9148480892181396,
|
| 1132 |
+
"logps/chosen": -31.930761337280273,
|
| 1133 |
+
"logps/rejected": -35.87628173828125,
|
| 1134 |
+
"loss": 0.8252,
|
| 1135 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1136 |
+
"rewards/chosen": 0.2688351273536682,
|
| 1137 |
+
"rewards/margins": 0.36267900466918945,
|
| 1138 |
+
"rewards/rejected": -0.09384389221668243,
|
| 1139 |
+
"step": 700
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 1.82,
|
| 1143 |
+
"eval_logits/chosen": -2.2346324920654297,
|
| 1144 |
+
"eval_logits/rejected": -2.2297565937042236,
|
| 1145 |
+
"eval_logps/chosen": -33.97812271118164,
|
| 1146 |
+
"eval_logps/rejected": -37.50526428222656,
|
| 1147 |
+
"eval_loss": 1.6334478855133057,
|
| 1148 |
+
"eval_rewards/accuracies": 0.5215947031974792,
|
| 1149 |
+
"eval_rewards/chosen": 0.022571848705410957,
|
| 1150 |
+
"eval_rewards/margins": 0.018028803169727325,
|
| 1151 |
+
"eval_rewards/rejected": 0.004543047398328781,
|
| 1152 |
+
"eval_runtime": 145.3597,
|
| 1153 |
+
"eval_samples_per_second": 2.36,
|
| 1154 |
+
"eval_steps_per_second": 0.296,
|
| 1155 |
+
"step": 700
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 1.84,
|
| 1159 |
+
"grad_norm": 62.75,
|
| 1160 |
+
"learning_rate": 3.2639145321045933e-06,
|
| 1161 |
+
"logits/chosen": -2.0267205238342285,
|
| 1162 |
+
"logits/rejected": -2.0294930934906006,
|
| 1163 |
+
"logps/chosen": -34.24456024169922,
|
| 1164 |
+
"logps/rejected": -34.45062255859375,
|
| 1165 |
+
"loss": 0.8568,
|
| 1166 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1167 |
+
"rewards/chosen": 0.2604122459888458,
|
| 1168 |
+
"rewards/margins": 0.3325996696949005,
|
| 1169 |
+
"rewards/rejected": -0.0721873790025711,
|
| 1170 |
+
"step": 710
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 1.87,
|
| 1174 |
+
"grad_norm": 105.0,
|
| 1175 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1176 |
+
"logits/chosen": -1.8829209804534912,
|
| 1177 |
+
"logits/rejected": -1.877000093460083,
|
| 1178 |
+
"logps/chosen": -34.198673248291016,
|
| 1179 |
+
"logps/rejected": -32.5337028503418,
|
| 1180 |
+
"loss": 0.9437,
|
| 1181 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1182 |
+
"rewards/chosen": 0.25382402539253235,
|
| 1183 |
+
"rewards/margins": 0.3265684247016907,
|
| 1184 |
+
"rewards/rejected": -0.07274442911148071,
|
| 1185 |
+
"step": 720
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.9,
|
| 1189 |
+
"grad_norm": 35.75,
|
| 1190 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1191 |
+
"logits/chosen": -2.0574429035186768,
|
| 1192 |
+
"logits/rejected": -2.0544135570526123,
|
| 1193 |
+
"logps/chosen": -30.041122436523438,
|
| 1194 |
+
"logps/rejected": -31.79475975036621,
|
| 1195 |
+
"loss": 0.5771,
|
| 1196 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1197 |
+
"rewards/chosen": 0.24474552273750305,
|
| 1198 |
+
"rewards/margins": 0.38337013125419617,
|
| 1199 |
+
"rewards/rejected": -0.13862457871437073,
|
| 1200 |
+
"step": 730
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 1.92,
|
| 1204 |
+
"grad_norm": 46.75,
|
| 1205 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1206 |
+
"logits/chosen": -1.8992414474487305,
|
| 1207 |
+
"logits/rejected": -1.8990885019302368,
|
| 1208 |
+
"logps/chosen": -32.76268768310547,
|
| 1209 |
+
"logps/rejected": -37.8664436340332,
|
| 1210 |
+
"loss": 0.5278,
|
| 1211 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1212 |
+
"rewards/chosen": 0.28019440174102783,
|
| 1213 |
+
"rewards/margins": 0.3756488263607025,
|
| 1214 |
+
"rewards/rejected": -0.09545440971851349,
|
| 1215 |
+
"step": 740
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"epoch": 1.95,
|
| 1219 |
+
"grad_norm": 87.0,
|
| 1220 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1221 |
+
"logits/chosen": -1.7812280654907227,
|
| 1222 |
+
"logits/rejected": -1.7864673137664795,
|
| 1223 |
+
"logps/chosen": -36.23480987548828,
|
| 1224 |
+
"logps/rejected": -34.09480667114258,
|
| 1225 |
+
"loss": 0.6358,
|
| 1226 |
+
"rewards/accuracies": 0.875,
|
| 1227 |
+
"rewards/chosen": 0.2790614366531372,
|
| 1228 |
+
"rewards/margins": 0.32804808020591736,
|
| 1229 |
+
"rewards/rejected": -0.048986706882715225,
|
| 1230 |
+
"step": 750
|
| 1231 |
+
},
|
| 1232 |
+
{
|
| 1233 |
+
"epoch": 1.97,
|
| 1234 |
+
"grad_norm": 37.5,
|
| 1235 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1236 |
+
"logits/chosen": -1.9918644428253174,
|
| 1237 |
+
"logits/rejected": -1.9932422637939453,
|
| 1238 |
+
"logps/chosen": -31.911479949951172,
|
| 1239 |
+
"logps/rejected": -32.96862030029297,
|
| 1240 |
+
"loss": 0.5173,
|
| 1241 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1242 |
+
"rewards/chosen": 0.2776804268360138,
|
| 1243 |
+
"rewards/margins": 0.3534439206123352,
|
| 1244 |
+
"rewards/rejected": -0.0757635161280632,
|
| 1245 |
+
"step": 760
|
| 1246 |
+
},
|
| 1247 |
+
{
|
| 1248 |
+
"epoch": 2.0,
|
| 1249 |
+
"grad_norm": 101.0,
|
| 1250 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1251 |
+
"logits/chosen": -1.9615776538848877,
|
| 1252 |
+
"logits/rejected": -1.960758924484253,
|
| 1253 |
+
"logps/chosen": -31.484752655029297,
|
| 1254 |
+
"logps/rejected": -35.08818817138672,
|
| 1255 |
+
"loss": 0.5913,
|
| 1256 |
+
"rewards/accuracies": 0.8708333969116211,
|
| 1257 |
+
"rewards/chosen": 0.27634382247924805,
|
| 1258 |
+
"rewards/margins": 0.30035752058029175,
|
| 1259 |
+
"rewards/rejected": -0.024013686925172806,
|
| 1260 |
+
"step": 770
|
| 1261 |
+
},
|
| 1262 |
+
{
|
| 1263 |
+
"epoch": 2.03,
|
| 1264 |
+
"grad_norm": 33.25,
|
| 1265 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1266 |
+
"logits/chosen": -1.9611775875091553,
|
| 1267 |
+
"logits/rejected": -1.960415244102478,
|
| 1268 |
+
"logps/chosen": -33.879024505615234,
|
| 1269 |
+
"logps/rejected": -31.674047470092773,
|
| 1270 |
+
"loss": 0.3971,
|
| 1271 |
+
"rewards/accuracies": 0.9375,
|
| 1272 |
+
"rewards/chosen": 0.30777519941329956,
|
| 1273 |
+
"rewards/margins": 0.4095862805843353,
|
| 1274 |
+
"rewards/rejected": -0.10181108862161636,
|
| 1275 |
+
"step": 780
|
| 1276 |
+
},
|
| 1277 |
+
{
|
| 1278 |
+
"epoch": 2.05,
|
| 1279 |
+
"grad_norm": 40.75,
|
| 1280 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1281 |
+
"logits/chosen": -1.9905250072479248,
|
| 1282 |
+
"logits/rejected": -1.989000916481018,
|
| 1283 |
+
"logps/chosen": -28.83389663696289,
|
| 1284 |
+
"logps/rejected": -32.96330261230469,
|
| 1285 |
+
"loss": 0.3133,
|
| 1286 |
+
"rewards/accuracies": 0.9375,
|
| 1287 |
+
"rewards/chosen": 0.29171431064605713,
|
| 1288 |
+
"rewards/margins": 0.4225086271762848,
|
| 1289 |
+
"rewards/rejected": -0.13079434633255005,
|
| 1290 |
+
"step": 790
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"epoch": 2.08,
|
| 1294 |
+
"grad_norm": 76.0,
|
| 1295 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1296 |
+
"logits/chosen": -2.0426809787750244,
|
| 1297 |
+
"logits/rejected": -2.039111375808716,
|
| 1298 |
+
"logps/chosen": -31.69073486328125,
|
| 1299 |
+
"logps/rejected": -34.53063201904297,
|
| 1300 |
+
"loss": 0.2848,
|
| 1301 |
+
"rewards/accuracies": 0.9375,
|
| 1302 |
+
"rewards/chosen": 0.2945896089076996,
|
| 1303 |
+
"rewards/margins": 0.42722758650779724,
|
| 1304 |
+
"rewards/rejected": -0.13263803720474243,
|
| 1305 |
+
"step": 800
|
| 1306 |
+
},
|
| 1307 |
+
{
|
| 1308 |
+
"epoch": 2.08,
|
| 1309 |
+
"eval_logits/chosen": -2.2361207008361816,
|
| 1310 |
+
"eval_logits/rejected": -2.231266736984253,
|
| 1311 |
+
"eval_logps/chosen": -33.99641799926758,
|
| 1312 |
+
"eval_logps/rejected": -37.524444580078125,
|
| 1313 |
+
"eval_loss": 1.603250503540039,
|
| 1314 |
+
"eval_rewards/accuracies": 0.52491694688797,
|
| 1315 |
+
"eval_rewards/chosen": 0.015252894721925259,
|
| 1316 |
+
"eval_rewards/margins": 0.018381815403699875,
|
| 1317 |
+
"eval_rewards/rejected": -0.003128920216113329,
|
| 1318 |
+
"eval_runtime": 145.2046,
|
| 1319 |
+
"eval_samples_per_second": 2.362,
|
| 1320 |
+
"eval_steps_per_second": 0.296,
|
| 1321 |
+
"step": 800
|
| 1322 |
+
},
|
| 1323 |
+
{
|
| 1324 |
+
"epoch": 2.1,
|
| 1325 |
+
"grad_norm": 60.5,
|
| 1326 |
+
"learning_rate": 2.7094204786572254e-06,
|
| 1327 |
+
"logits/chosen": -1.8801984786987305,
|
| 1328 |
+
"logits/rejected": -1.8722867965698242,
|
| 1329 |
+
"logps/chosen": -33.65568542480469,
|
| 1330 |
+
"logps/rejected": -34.81199645996094,
|
| 1331 |
+
"loss": 0.3333,
|
| 1332 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1333 |
+
"rewards/chosen": 0.32022127509117126,
|
| 1334 |
+
"rewards/margins": 0.42097288370132446,
|
| 1335 |
+
"rewards/rejected": -0.10075156390666962,
|
| 1336 |
+
"step": 810
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 2.13,
|
| 1340 |
+
"grad_norm": 61.25,
|
| 1341 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1342 |
+
"logits/chosen": -1.955896019935608,
|
| 1343 |
+
"logits/rejected": -1.9664968252182007,
|
| 1344 |
+
"logps/chosen": -34.896785736083984,
|
| 1345 |
+
"logps/rejected": -32.63465118408203,
|
| 1346 |
+
"loss": 0.4459,
|
| 1347 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1348 |
+
"rewards/chosen": 0.31956198811531067,
|
| 1349 |
+
"rewards/margins": 0.4008457660675049,
|
| 1350 |
+
"rewards/rejected": -0.0812838152050972,
|
| 1351 |
+
"step": 820
|
| 1352 |
+
},
|
| 1353 |
+
{
|
| 1354 |
+
"epoch": 2.16,
|
| 1355 |
+
"grad_norm": 84.0,
|
| 1356 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1357 |
+
"logits/chosen": -2.0049898624420166,
|
| 1358 |
+
"logits/rejected": -2.0092310905456543,
|
| 1359 |
+
"logps/chosen": -34.0383415222168,
|
| 1360 |
+
"logps/rejected": -29.4811954498291,
|
| 1361 |
+
"loss": 0.2762,
|
| 1362 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1363 |
+
"rewards/chosen": 0.34577542543411255,
|
| 1364 |
+
"rewards/margins": 0.43095555901527405,
|
| 1365 |
+
"rewards/rejected": -0.0851801335811615,
|
| 1366 |
+
"step": 830
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 2.18,
|
| 1370 |
+
"grad_norm": 82.0,
|
| 1371 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1372 |
+
"logits/chosen": -2.0014491081237793,
|
| 1373 |
+
"logits/rejected": -2.009239912033081,
|
| 1374 |
+
"logps/chosen": -34.01618194580078,
|
| 1375 |
+
"logps/rejected": -30.190011978149414,
|
| 1376 |
+
"loss": 0.496,
|
| 1377 |
+
"rewards/accuracies": 0.9375,
|
| 1378 |
+
"rewards/chosen": 0.30874118208885193,
|
| 1379 |
+
"rewards/margins": 0.4456659257411957,
|
| 1380 |
+
"rewards/rejected": -0.13692477345466614,
|
| 1381 |
+
"step": 840
|
| 1382 |
+
},
|
| 1383 |
+
{
|
| 1384 |
+
"epoch": 2.21,
|
| 1385 |
+
"grad_norm": 40.0,
|
| 1386 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1387 |
+
"logits/chosen": -1.930978536605835,
|
| 1388 |
+
"logits/rejected": -1.9214372634887695,
|
| 1389 |
+
"logps/chosen": -31.05685043334961,
|
| 1390 |
+
"logps/rejected": -32.36315155029297,
|
| 1391 |
+
"loss": 0.2819,
|
| 1392 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1393 |
+
"rewards/chosen": 0.29453787207603455,
|
| 1394 |
+
"rewards/margins": 0.43427667021751404,
|
| 1395 |
+
"rewards/rejected": -0.1397387832403183,
|
| 1396 |
+
"step": 850
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 2.23,
|
| 1400 |
+
"grad_norm": 34.75,
|
| 1401 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1402 |
+
"logits/chosen": -2.072263717651367,
|
| 1403 |
+
"logits/rejected": -2.062443971633911,
|
| 1404 |
+
"logps/chosen": -25.44430923461914,
|
| 1405 |
+
"logps/rejected": -30.206756591796875,
|
| 1406 |
+
"loss": 0.4507,
|
| 1407 |
+
"rewards/accuracies": 0.9375,
|
| 1408 |
+
"rewards/chosen": 0.3119046986103058,
|
| 1409 |
+
"rewards/margins": 0.44567227363586426,
|
| 1410 |
+
"rewards/rejected": -0.13376764953136444,
|
| 1411 |
+
"step": 860
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.26,
|
| 1415 |
+
"grad_norm": 89.0,
|
| 1416 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1417 |
+
"logits/chosen": -1.9230930805206299,
|
| 1418 |
+
"logits/rejected": -1.9242877960205078,
|
| 1419 |
+
"logps/chosen": -33.56352233886719,
|
| 1420 |
+
"logps/rejected": -30.285165786743164,
|
| 1421 |
+
"loss": 0.3499,
|
| 1422 |
+
"rewards/accuracies": 0.9375,
|
| 1423 |
+
"rewards/chosen": 0.3077928125858307,
|
| 1424 |
+
"rewards/margins": 0.42885905504226685,
|
| 1425 |
+
"rewards/rejected": -0.12106621265411377,
|
| 1426 |
+
"step": 870
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"epoch": 2.29,
|
| 1430 |
+
"grad_norm": 29.5,
|
| 1431 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1432 |
+
"logits/chosen": -1.9387619495391846,
|
| 1433 |
+
"logits/rejected": -1.9395242929458618,
|
| 1434 |
+
"logps/chosen": -31.70212745666504,
|
| 1435 |
+
"logps/rejected": -33.355247497558594,
|
| 1436 |
+
"loss": 0.5159,
|
| 1437 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1438 |
+
"rewards/chosen": 0.27484866976737976,
|
| 1439 |
+
"rewards/margins": 0.3965802788734436,
|
| 1440 |
+
"rewards/rejected": -0.12173161655664444,
|
| 1441 |
+
"step": 880
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.31,
|
| 1445 |
+
"grad_norm": 63.5,
|
| 1446 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1447 |
+
"logits/chosen": -1.8424198627471924,
|
| 1448 |
+
"logits/rejected": -1.8406569957733154,
|
| 1449 |
+
"logps/chosen": -34.8422966003418,
|
| 1450 |
+
"logps/rejected": -36.311790466308594,
|
| 1451 |
+
"loss": 0.4473,
|
| 1452 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1453 |
+
"rewards/chosen": 0.3290487825870514,
|
| 1454 |
+
"rewards/margins": 0.5028699636459351,
|
| 1455 |
+
"rewards/rejected": -0.17382115125656128,
|
| 1456 |
+
"step": 890
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"epoch": 2.34,
|
| 1460 |
+
"grad_norm": 120.5,
|
| 1461 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1462 |
+
"logits/chosen": -1.9881842136383057,
|
| 1463 |
+
"logits/rejected": -1.9813495874404907,
|
| 1464 |
+
"logps/chosen": -31.175704956054688,
|
| 1465 |
+
"logps/rejected": -32.99733352661133,
|
| 1466 |
+
"loss": 0.3671,
|
| 1467 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1468 |
+
"rewards/chosen": 0.3307729959487915,
|
| 1469 |
+
"rewards/margins": 0.42647916078567505,
|
| 1470 |
+
"rewards/rejected": -0.09570614993572235,
|
| 1471 |
+
"step": 900
|
| 1472 |
+
},
|
| 1473 |
+
{
|
| 1474 |
+
"epoch": 2.34,
|
| 1475 |
+
"eval_logits/chosen": -2.235804796218872,
|
| 1476 |
+
"eval_logits/rejected": -2.230922222137451,
|
| 1477 |
+
"eval_logps/chosen": -33.96369552612305,
|
| 1478 |
+
"eval_logps/rejected": -37.47232437133789,
|
| 1479 |
+
"eval_loss": 1.6569337844848633,
|
| 1480 |
+
"eval_rewards/accuracies": 0.5161960124969482,
|
| 1481 |
+
"eval_rewards/chosen": 0.028342105448246002,
|
| 1482 |
+
"eval_rewards/margins": 0.010623876005411148,
|
| 1483 |
+
"eval_rewards/rejected": 0.017718229442834854,
|
| 1484 |
+
"eval_runtime": 145.1552,
|
| 1485 |
+
"eval_samples_per_second": 2.363,
|
| 1486 |
+
"eval_steps_per_second": 0.296,
|
| 1487 |
+
"step": 900
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 2.36,
|
| 1491 |
+
"grad_norm": 49.25,
|
| 1492 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 1493 |
+
"logits/chosen": -2.0350537300109863,
|
| 1494 |
+
"logits/rejected": -2.0302882194519043,
|
| 1495 |
+
"logps/chosen": -32.76664733886719,
|
| 1496 |
+
"logps/rejected": -34.81679916381836,
|
| 1497 |
+
"loss": 0.309,
|
| 1498 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1499 |
+
"rewards/chosen": 0.3164145350456238,
|
| 1500 |
+
"rewards/margins": 0.4316856861114502,
|
| 1501 |
+
"rewards/rejected": -0.1152711883187294,
|
| 1502 |
+
"step": 910
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 2.39,
|
| 1506 |
+
"grad_norm": 59.5,
|
| 1507 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1508 |
+
"logits/chosen": -1.9527454376220703,
|
| 1509 |
+
"logits/rejected": -1.9576810598373413,
|
| 1510 |
+
"logps/chosen": -34.50981903076172,
|
| 1511 |
+
"logps/rejected": -32.73914337158203,
|
| 1512 |
+
"loss": 0.4281,
|
| 1513 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1514 |
+
"rewards/chosen": 0.3355061411857605,
|
| 1515 |
+
"rewards/margins": 0.43119993805885315,
|
| 1516 |
+
"rewards/rejected": -0.09569380432367325,
|
| 1517 |
+
"step": 920
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 2.42,
|
| 1521 |
+
"grad_norm": 58.25,
|
| 1522 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1523 |
+
"logits/chosen": -2.0501203536987305,
|
| 1524 |
+
"logits/rejected": -2.0407142639160156,
|
| 1525 |
+
"logps/chosen": -31.465560913085938,
|
| 1526 |
+
"logps/rejected": -34.783172607421875,
|
| 1527 |
+
"loss": 0.3184,
|
| 1528 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1529 |
+
"rewards/chosen": 0.3277474045753479,
|
| 1530 |
+
"rewards/margins": 0.4638980031013489,
|
| 1531 |
+
"rewards/rejected": -0.13615058362483978,
|
| 1532 |
+
"step": 930
|
| 1533 |
+
},
|
| 1534 |
+
{
|
| 1535 |
+
"epoch": 2.44,
|
| 1536 |
+
"grad_norm": 45.5,
|
| 1537 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1538 |
+
"logits/chosen": -1.9754626750946045,
|
| 1539 |
+
"logits/rejected": -1.9723625183105469,
|
| 1540 |
+
"logps/chosen": -30.506328582763672,
|
| 1541 |
+
"logps/rejected": -32.73186111450195,
|
| 1542 |
+
"loss": 0.3318,
|
| 1543 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1544 |
+
"rewards/chosen": 0.3128509521484375,
|
| 1545 |
+
"rewards/margins": 0.41944509744644165,
|
| 1546 |
+
"rewards/rejected": -0.10659414529800415,
|
| 1547 |
+
"step": 940
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 2.47,
|
| 1551 |
+
"grad_norm": 21.875,
|
| 1552 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1553 |
+
"logits/chosen": -1.9184865951538086,
|
| 1554 |
+
"logits/rejected": -1.9259870052337646,
|
| 1555 |
+
"logps/chosen": -32.09298324584961,
|
| 1556 |
+
"logps/rejected": -35.809776306152344,
|
| 1557 |
+
"loss": 0.5459,
|
| 1558 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1559 |
+
"rewards/chosen": 0.3112277388572693,
|
| 1560 |
+
"rewards/margins": 0.4151647686958313,
|
| 1561 |
+
"rewards/rejected": -0.1039370447397232,
|
| 1562 |
+
"step": 950
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 2.49,
|
| 1566 |
+
"grad_norm": 67.5,
|
| 1567 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1568 |
+
"logits/chosen": -1.9273126125335693,
|
| 1569 |
+
"logits/rejected": -1.9204585552215576,
|
| 1570 |
+
"logps/chosen": -33.03001403808594,
|
| 1571 |
+
"logps/rejected": -36.92626190185547,
|
| 1572 |
+
"loss": 0.4169,
|
| 1573 |
+
"rewards/accuracies": 0.9375,
|
| 1574 |
+
"rewards/chosen": 0.31939974427223206,
|
| 1575 |
+
"rewards/margins": 0.45038121938705444,
|
| 1576 |
+
"rewards/rejected": -0.13098154962062836,
|
| 1577 |
+
"step": 960
|
| 1578 |
+
},
|
| 1579 |
+
{
|
| 1580 |
+
"epoch": 2.52,
|
| 1581 |
+
"grad_norm": 18.375,
|
| 1582 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1583 |
+
"logits/chosen": -1.9290339946746826,
|
| 1584 |
+
"logits/rejected": -1.9297443628311157,
|
| 1585 |
+
"logps/chosen": -30.42807960510254,
|
| 1586 |
+
"logps/rejected": -33.275390625,
|
| 1587 |
+
"loss": 0.3536,
|
| 1588 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1589 |
+
"rewards/chosen": 0.32332831621170044,
|
| 1590 |
+
"rewards/margins": 0.44098013639450073,
|
| 1591 |
+
"rewards/rejected": -0.11765184253454208,
|
| 1592 |
+
"step": 970
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 2.55,
|
| 1596 |
+
"grad_norm": 45.5,
|
| 1597 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1598 |
+
"logits/chosen": -1.9345155954360962,
|
| 1599 |
+
"logits/rejected": -1.9458023309707642,
|
| 1600 |
+
"logps/chosen": -32.152503967285156,
|
| 1601 |
+
"logps/rejected": -34.349021911621094,
|
| 1602 |
+
"loss": 0.346,
|
| 1603 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1604 |
+
"rewards/chosen": 0.36967429518699646,
|
| 1605 |
+
"rewards/margins": 0.4567973017692566,
|
| 1606 |
+
"rewards/rejected": -0.08712298423051834,
|
| 1607 |
+
"step": 980
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 2.57,
|
| 1611 |
+
"grad_norm": 79.0,
|
| 1612 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1613 |
+
"logits/chosen": -2.0078938007354736,
|
| 1614 |
+
"logits/rejected": -2.0053062438964844,
|
| 1615 |
+
"logps/chosen": -35.25431823730469,
|
| 1616 |
+
"logps/rejected": -32.59493637084961,
|
| 1617 |
+
"loss": 0.4808,
|
| 1618 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1619 |
+
"rewards/chosen": 0.34669190645217896,
|
| 1620 |
+
"rewards/margins": 0.42497485876083374,
|
| 1621 |
+
"rewards/rejected": -0.07828302681446075,
|
| 1622 |
+
"step": 990
|
| 1623 |
+
},
|
| 1624 |
+
{
|
| 1625 |
+
"epoch": 2.6,
|
| 1626 |
+
"grad_norm": 44.0,
|
| 1627 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1628 |
+
"logits/chosen": -1.9761663675308228,
|
| 1629 |
+
"logits/rejected": -1.9800007343292236,
|
| 1630 |
+
"logps/chosen": -33.687164306640625,
|
| 1631 |
+
"logps/rejected": -34.747779846191406,
|
| 1632 |
+
"loss": 0.3936,
|
| 1633 |
+
"rewards/accuracies": 0.9375,
|
| 1634 |
+
"rewards/chosen": 0.3104504346847534,
|
| 1635 |
+
"rewards/margins": 0.41514307260513306,
|
| 1636 |
+
"rewards/rejected": -0.10469266027212143,
|
| 1637 |
+
"step": 1000
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 2.6,
|
| 1641 |
+
"eval_logits/chosen": -2.2373571395874023,
|
| 1642 |
+
"eval_logits/rejected": -2.2324929237365723,
|
| 1643 |
+
"eval_logps/chosen": -33.947471618652344,
|
| 1644 |
+
"eval_logps/rejected": -37.469818115234375,
|
| 1645 |
+
"eval_loss": 1.6203227043151855,
|
| 1646 |
+
"eval_rewards/accuracies": 0.5427741408348083,
|
| 1647 |
+
"eval_rewards/chosen": 0.03483137488365173,
|
| 1648 |
+
"eval_rewards/margins": 0.0161101296544075,
|
| 1649 |
+
"eval_rewards/rejected": 0.018721245229244232,
|
| 1650 |
+
"eval_runtime": 145.2271,
|
| 1651 |
+
"eval_samples_per_second": 2.362,
|
| 1652 |
+
"eval_steps_per_second": 0.296,
|
| 1653 |
+
"step": 1000
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.62,
|
| 1657 |
+
"grad_norm": 103.5,
|
| 1658 |
+
"learning_rate": 1.5972066659083796e-06,
|
| 1659 |
+
"logits/chosen": -1.9862302541732788,
|
| 1660 |
+
"logits/rejected": -1.9912649393081665,
|
| 1661 |
+
"logps/chosen": -32.02253341674805,
|
| 1662 |
+
"logps/rejected": -32.78570556640625,
|
| 1663 |
+
"loss": 0.3762,
|
| 1664 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1665 |
+
"rewards/chosen": 0.3111560046672821,
|
| 1666 |
+
"rewards/margins": 0.4169352650642395,
|
| 1667 |
+
"rewards/rejected": -0.10577923059463501,
|
| 1668 |
+
"step": 1010
|
| 1669 |
+
},
|
| 1670 |
+
{
|
| 1671 |
+
"epoch": 2.65,
|
| 1672 |
+
"grad_norm": 54.25,
|
| 1673 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1674 |
+
"logits/chosen": -1.948102355003357,
|
| 1675 |
+
"logits/rejected": -1.9508020877838135,
|
| 1676 |
+
"logps/chosen": -29.719890594482422,
|
| 1677 |
+
"logps/rejected": -31.992984771728516,
|
| 1678 |
+
"loss": 0.3461,
|
| 1679 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1680 |
+
"rewards/chosen": 0.29029980301856995,
|
| 1681 |
+
"rewards/margins": 0.407827228307724,
|
| 1682 |
+
"rewards/rejected": -0.11752742528915405,
|
| 1683 |
+
"step": 1020
|
| 1684 |
+
},
|
| 1685 |
+
{
|
| 1686 |
+
"epoch": 2.68,
|
| 1687 |
+
"grad_norm": 60.0,
|
| 1688 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1689 |
+
"logits/chosen": -2.005060911178589,
|
| 1690 |
+
"logits/rejected": -1.9991967678070068,
|
| 1691 |
+
"logps/chosen": -33.16535568237305,
|
| 1692 |
+
"logps/rejected": -33.78788757324219,
|
| 1693 |
+
"loss": 0.2867,
|
| 1694 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1695 |
+
"rewards/chosen": 0.32950255274772644,
|
| 1696 |
+
"rewards/margins": 0.43745851516723633,
|
| 1697 |
+
"rewards/rejected": -0.10795596987009048,
|
| 1698 |
+
"step": 1030
|
| 1699 |
+
},
|
| 1700 |
+
{
|
| 1701 |
+
"epoch": 2.7,
|
| 1702 |
+
"grad_norm": 69.0,
|
| 1703 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1704 |
+
"logits/chosen": -1.9332466125488281,
|
| 1705 |
+
"logits/rejected": -1.936981201171875,
|
| 1706 |
+
"logps/chosen": -34.967830657958984,
|
| 1707 |
+
"logps/rejected": -34.762996673583984,
|
| 1708 |
+
"loss": 0.3317,
|
| 1709 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1710 |
+
"rewards/chosen": 0.314359575510025,
|
| 1711 |
+
"rewards/margins": 0.4517282545566559,
|
| 1712 |
+
"rewards/rejected": -0.13736867904663086,
|
| 1713 |
+
"step": 1040
|
| 1714 |
+
},
|
| 1715 |
+
{
|
| 1716 |
+
"epoch": 2.73,
|
| 1717 |
+
"grad_norm": 26.875,
|
| 1718 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1719 |
+
"logits/chosen": -1.9331417083740234,
|
| 1720 |
+
"logits/rejected": -1.9462738037109375,
|
| 1721 |
+
"logps/chosen": -31.087848663330078,
|
| 1722 |
+
"logps/rejected": -33.28511428833008,
|
| 1723 |
+
"loss": 0.3932,
|
| 1724 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1725 |
+
"rewards/chosen": 0.31001484394073486,
|
| 1726 |
+
"rewards/margins": 0.4160439074039459,
|
| 1727 |
+
"rewards/rejected": -0.10602907836437225,
|
| 1728 |
+
"step": 1050
|
| 1729 |
+
},
|
| 1730 |
+
{
|
| 1731 |
+
"epoch": 2.75,
|
| 1732 |
+
"grad_norm": 97.5,
|
| 1733 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1734 |
+
"logits/chosen": -1.9240528345108032,
|
| 1735 |
+
"logits/rejected": -1.917449951171875,
|
| 1736 |
+
"logps/chosen": -31.343292236328125,
|
| 1737 |
+
"logps/rejected": -33.56601333618164,
|
| 1738 |
+
"loss": 0.3354,
|
| 1739 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1740 |
+
"rewards/chosen": 0.3264150619506836,
|
| 1741 |
+
"rewards/margins": 0.4256357252597809,
|
| 1742 |
+
"rewards/rejected": -0.0992206335067749,
|
| 1743 |
+
"step": 1060
|
| 1744 |
+
},
|
| 1745 |
+
{
|
| 1746 |
+
"epoch": 2.78,
|
| 1747 |
+
"grad_norm": 62.75,
|
| 1748 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1749 |
+
"logits/chosen": -2.051302433013916,
|
| 1750 |
+
"logits/rejected": -2.0521297454833984,
|
| 1751 |
+
"logps/chosen": -32.764705657958984,
|
| 1752 |
+
"logps/rejected": -33.20236587524414,
|
| 1753 |
+
"loss": 0.2782,
|
| 1754 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1755 |
+
"rewards/chosen": 0.312948077917099,
|
| 1756 |
+
"rewards/margins": 0.45601654052734375,
|
| 1757 |
+
"rewards/rejected": -0.14306847751140594,
|
| 1758 |
+
"step": 1070
|
| 1759 |
+
},
|
| 1760 |
+
{
|
| 1761 |
+
"epoch": 2.81,
|
| 1762 |
+
"grad_norm": 36.5,
|
| 1763 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1764 |
+
"logits/chosen": -1.982011079788208,
|
| 1765 |
+
"logits/rejected": -1.9808781147003174,
|
| 1766 |
+
"logps/chosen": -33.330299377441406,
|
| 1767 |
+
"logps/rejected": -33.3953971862793,
|
| 1768 |
+
"loss": 0.5045,
|
| 1769 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1770 |
+
"rewards/chosen": 0.3038138747215271,
|
| 1771 |
+
"rewards/margins": 0.38524001836776733,
|
| 1772 |
+
"rewards/rejected": -0.08142612874507904,
|
| 1773 |
+
"step": 1080
|
| 1774 |
+
},
|
| 1775 |
+
{
|
| 1776 |
+
"epoch": 2.83,
|
| 1777 |
+
"grad_norm": 31.875,
|
| 1778 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1779 |
+
"logits/chosen": -2.01176118850708,
|
| 1780 |
+
"logits/rejected": -2.004225969314575,
|
| 1781 |
+
"logps/chosen": -33.374778747558594,
|
| 1782 |
+
"logps/rejected": -31.283130645751953,
|
| 1783 |
+
"loss": 0.2661,
|
| 1784 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1785 |
+
"rewards/chosen": 0.34043893218040466,
|
| 1786 |
+
"rewards/margins": 0.4370775818824768,
|
| 1787 |
+
"rewards/rejected": -0.09663861989974976,
|
| 1788 |
+
"step": 1090
|
| 1789 |
+
},
|
| 1790 |
+
{
|
| 1791 |
+
"epoch": 2.86,
|
| 1792 |
+
"grad_norm": 102.5,
|
| 1793 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1794 |
+
"logits/chosen": -1.9781728982925415,
|
| 1795 |
+
"logits/rejected": -1.97735595703125,
|
| 1796 |
+
"logps/chosen": -34.61248779296875,
|
| 1797 |
+
"logps/rejected": -36.63856887817383,
|
| 1798 |
+
"loss": 0.3156,
|
| 1799 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1800 |
+
"rewards/chosen": 0.3223697543144226,
|
| 1801 |
+
"rewards/margins": 0.4367140233516693,
|
| 1802 |
+
"rewards/rejected": -0.11434426158666611,
|
| 1803 |
+
"step": 1100
|
| 1804 |
+
},
|
| 1805 |
+
{
|
| 1806 |
+
"epoch": 2.86,
|
| 1807 |
+
"eval_logits/chosen": -2.237501621246338,
|
| 1808 |
+
"eval_logits/rejected": -2.232618570327759,
|
| 1809 |
+
"eval_logps/chosen": -33.959171295166016,
|
| 1810 |
+
"eval_logps/rejected": -37.489627838134766,
|
| 1811 |
+
"eval_loss": 1.6012389659881592,
|
| 1812 |
+
"eval_rewards/accuracies": 0.5606312155723572,
|
| 1813 |
+
"eval_rewards/chosen": 0.030150815844535828,
|
| 1814 |
+
"eval_rewards/margins": 0.01935432478785515,
|
| 1815 |
+
"eval_rewards/rejected": 0.01079649105668068,
|
| 1816 |
+
"eval_runtime": 145.2896,
|
| 1817 |
+
"eval_samples_per_second": 2.361,
|
| 1818 |
+
"eval_steps_per_second": 0.296,
|
| 1819 |
+
"step": 1100
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"epoch": 2.88,
|
| 1823 |
+
"grad_norm": 66.0,
|
| 1824 |
+
"learning_rate": 1.0963854773524548e-06,
|
| 1825 |
+
"logits/chosen": -2.0257740020751953,
|
| 1826 |
+
"logits/rejected": -2.0311310291290283,
|
| 1827 |
+
"logps/chosen": -34.69486999511719,
|
| 1828 |
+
"logps/rejected": -35.754554748535156,
|
| 1829 |
+
"loss": 0.4157,
|
| 1830 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1831 |
+
"rewards/chosen": 0.30889827013015747,
|
| 1832 |
+
"rewards/margins": 0.39328455924987793,
|
| 1833 |
+
"rewards/rejected": -0.08438632637262344,
|
| 1834 |
+
"step": 1110
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"epoch": 2.91,
|
| 1838 |
+
"grad_norm": 90.0,
|
| 1839 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1840 |
+
"logits/chosen": -1.9978212118148804,
|
| 1841 |
+
"logits/rejected": -1.995883584022522,
|
| 1842 |
+
"logps/chosen": -31.900583267211914,
|
| 1843 |
+
"logps/rejected": -33.278594970703125,
|
| 1844 |
+
"loss": 0.4591,
|
| 1845 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1846 |
+
"rewards/chosen": 0.2862483263015747,
|
| 1847 |
+
"rewards/margins": 0.4045858383178711,
|
| 1848 |
+
"rewards/rejected": -0.11833752691745758,
|
| 1849 |
+
"step": 1120
|
| 1850 |
+
},
|
| 1851 |
+
{
|
| 1852 |
+
"epoch": 2.94,
|
| 1853 |
+
"grad_norm": 42.25,
|
| 1854 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1855 |
+
"logits/chosen": -1.9011417627334595,
|
| 1856 |
+
"logits/rejected": -1.9039758443832397,
|
| 1857 |
+
"logps/chosen": -28.964651107788086,
|
| 1858 |
+
"logps/rejected": -30.513931274414062,
|
| 1859 |
+
"loss": 0.2561,
|
| 1860 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1861 |
+
"rewards/chosen": 0.31359297037124634,
|
| 1862 |
+
"rewards/margins": 0.42114973068237305,
|
| 1863 |
+
"rewards/rejected": -0.1075567752122879,
|
| 1864 |
+
"step": 1130
|
| 1865 |
+
},
|
| 1866 |
+
{
|
| 1867 |
+
"epoch": 2.96,
|
| 1868 |
+
"grad_norm": 19.875,
|
| 1869 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1870 |
+
"logits/chosen": -1.9979721307754517,
|
| 1871 |
+
"logits/rejected": -1.994855284690857,
|
| 1872 |
+
"logps/chosen": -32.40828323364258,
|
| 1873 |
+
"logps/rejected": -32.802223205566406,
|
| 1874 |
+
"loss": 0.3667,
|
| 1875 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1876 |
+
"rewards/chosen": 0.3181864619255066,
|
| 1877 |
+
"rewards/margins": 0.4283228814601898,
|
| 1878 |
+
"rewards/rejected": -0.11013643443584442,
|
| 1879 |
+
"step": 1140
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 2.99,
|
| 1883 |
+
"grad_norm": 53.25,
|
| 1884 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1885 |
+
"logits/chosen": -1.9201629161834717,
|
| 1886 |
+
"logits/rejected": -1.9138599634170532,
|
| 1887 |
+
"logps/chosen": -27.41982078552246,
|
| 1888 |
+
"logps/rejected": -32.52082061767578,
|
| 1889 |
+
"loss": 0.282,
|
| 1890 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1891 |
+
"rewards/chosen": 0.28726693987846375,
|
| 1892 |
+
"rewards/margins": 0.43627747893333435,
|
| 1893 |
+
"rewards/rejected": -0.14901044964790344,
|
| 1894 |
+
"step": 1150
|
| 1895 |
+
},
|
| 1896 |
+
{
|
| 1897 |
+
"epoch": 3.01,
|
| 1898 |
+
"grad_norm": 52.75,
|
| 1899 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1900 |
+
"logits/chosen": -2.038132905960083,
|
| 1901 |
+
"logits/rejected": -2.0356593132019043,
|
| 1902 |
+
"logps/chosen": -32.35755920410156,
|
| 1903 |
+
"logps/rejected": -35.35866165161133,
|
| 1904 |
+
"loss": 0.2873,
|
| 1905 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1906 |
+
"rewards/chosen": 0.29540252685546875,
|
| 1907 |
+
"rewards/margins": 0.40225353837013245,
|
| 1908 |
+
"rewards/rejected": -0.10685104131698608,
|
| 1909 |
+
"step": 1160
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.04,
|
| 1913 |
+
"grad_norm": 18.625,
|
| 1914 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1915 |
+
"logits/chosen": -2.0191562175750732,
|
| 1916 |
+
"logits/rejected": -2.0172648429870605,
|
| 1917 |
+
"logps/chosen": -30.838241577148438,
|
| 1918 |
+
"logps/rejected": -32.716129302978516,
|
| 1919 |
+
"loss": 0.1723,
|
| 1920 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1921 |
+
"rewards/chosen": 0.3104664385318756,
|
| 1922 |
+
"rewards/margins": 0.4513087272644043,
|
| 1923 |
+
"rewards/rejected": -0.1408422887325287,
|
| 1924 |
+
"step": 1170
|
| 1925 |
+
},
|
| 1926 |
+
{
|
| 1927 |
+
"epoch": 3.06,
|
| 1928 |
+
"grad_norm": 22.0,
|
| 1929 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1930 |
+
"logits/chosen": -2.011491537094116,
|
| 1931 |
+
"logits/rejected": -1.999138593673706,
|
| 1932 |
+
"logps/chosen": -35.201385498046875,
|
| 1933 |
+
"logps/rejected": -34.34071731567383,
|
| 1934 |
+
"loss": 0.2661,
|
| 1935 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1936 |
+
"rewards/chosen": 0.33756545186042786,
|
| 1937 |
+
"rewards/margins": 0.44526106119155884,
|
| 1938 |
+
"rewards/rejected": -0.10769567638635635,
|
| 1939 |
+
"step": 1180
|
| 1940 |
+
},
|
| 1941 |
+
{
|
| 1942 |
+
"epoch": 3.09,
|
| 1943 |
+
"grad_norm": 36.75,
|
| 1944 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1945 |
+
"logits/chosen": -1.9028282165527344,
|
| 1946 |
+
"logits/rejected": -1.8974478244781494,
|
| 1947 |
+
"logps/chosen": -34.48936080932617,
|
| 1948 |
+
"logps/rejected": -34.49340057373047,
|
| 1949 |
+
"loss": 0.2986,
|
| 1950 |
+
"rewards/accuracies": 0.9375,
|
| 1951 |
+
"rewards/chosen": 0.34413009881973267,
|
| 1952 |
+
"rewards/margins": 0.44636183977127075,
|
| 1953 |
+
"rewards/rejected": -0.1022317185997963,
|
| 1954 |
+
"step": 1190
|
| 1955 |
+
},
|
| 1956 |
+
{
|
| 1957 |
+
"epoch": 3.12,
|
| 1958 |
+
"grad_norm": 62.25,
|
| 1959 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1960 |
+
"logits/chosen": -1.9445934295654297,
|
| 1961 |
+
"logits/rejected": -1.9476436376571655,
|
| 1962 |
+
"logps/chosen": -31.85054588317871,
|
| 1963 |
+
"logps/rejected": -31.752578735351562,
|
| 1964 |
+
"loss": 0.2893,
|
| 1965 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1966 |
+
"rewards/chosen": 0.30582019686698914,
|
| 1967 |
+
"rewards/margins": 0.41237980127334595,
|
| 1968 |
+
"rewards/rejected": -0.10655955970287323,
|
| 1969 |
+
"step": 1200
|
| 1970 |
+
},
|
| 1971 |
+
{
|
| 1972 |
+
"epoch": 3.12,
|
| 1973 |
+
"eval_logits/chosen": -2.238332748413086,
|
| 1974 |
+
"eval_logits/rejected": -2.233450174331665,
|
| 1975 |
+
"eval_logps/chosen": -33.94802474975586,
|
| 1976 |
+
"eval_logps/rejected": -37.49089431762695,
|
| 1977 |
+
"eval_loss": 1.5704575777053833,
|
| 1978 |
+
"eval_rewards/accuracies": 0.5365448594093323,
|
| 1979 |
+
"eval_rewards/chosen": 0.03461090475320816,
|
| 1980 |
+
"eval_rewards/margins": 0.024320118129253387,
|
| 1981 |
+
"eval_rewards/rejected": 0.010290787555277348,
|
| 1982 |
+
"eval_runtime": 145.3137,
|
| 1983 |
+
"eval_samples_per_second": 2.36,
|
| 1984 |
+
"eval_steps_per_second": 0.296,
|
| 1985 |
+
"step": 1200
|
| 1986 |
+
},
|
| 1987 |
+
{
|
| 1988 |
+
"epoch": 3.14,
|
| 1989 |
+
"grad_norm": 39.0,
|
| 1990 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1991 |
+
"logits/chosen": -1.976152777671814,
|
| 1992 |
+
"logits/rejected": -1.971604347229004,
|
| 1993 |
+
"logps/chosen": -30.516315460205078,
|
| 1994 |
+
"logps/rejected": -31.238773345947266,
|
| 1995 |
+
"loss": 0.2438,
|
| 1996 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1997 |
+
"rewards/chosen": 0.3029424548149109,
|
| 1998 |
+
"rewards/margins": 0.4352136552333832,
|
| 1999 |
+
"rewards/rejected": -0.1322712004184723,
|
| 2000 |
+
"step": 1210
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 3.17,
|
| 2004 |
+
"grad_norm": 28.875,
|
| 2005 |
+
"learning_rate": 6.293013489185315e-07,
|
| 2006 |
+
"logits/chosen": -1.988176703453064,
|
| 2007 |
+
"logits/rejected": -1.982642412185669,
|
| 2008 |
+
"logps/chosen": -34.4455680847168,
|
| 2009 |
+
"logps/rejected": -34.812355041503906,
|
| 2010 |
+
"loss": 0.2853,
|
| 2011 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2012 |
+
"rewards/chosen": 0.33134883642196655,
|
| 2013 |
+
"rewards/margins": 0.4283396303653717,
|
| 2014 |
+
"rewards/rejected": -0.09699074923992157,
|
| 2015 |
+
"step": 1220
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 3.19,
|
| 2019 |
+
"grad_norm": 40.0,
|
| 2020 |
+
"learning_rate": 5.921934551632086e-07,
|
| 2021 |
+
"logits/chosen": -1.9901288747787476,
|
| 2022 |
+
"logits/rejected": -1.9769742488861084,
|
| 2023 |
+
"logps/chosen": -32.32646560668945,
|
| 2024 |
+
"logps/rejected": -34.40354537963867,
|
| 2025 |
+
"loss": 0.3093,
|
| 2026 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2027 |
+
"rewards/chosen": 0.30702072381973267,
|
| 2028 |
+
"rewards/margins": 0.4204995036125183,
|
| 2029 |
+
"rewards/rejected": -0.11347881704568863,
|
| 2030 |
+
"step": 1230
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 3.22,
|
| 2034 |
+
"grad_norm": 44.0,
|
| 2035 |
+
"learning_rate": 5.560657034652405e-07,
|
| 2036 |
+
"logits/chosen": -2.034132957458496,
|
| 2037 |
+
"logits/rejected": -2.0313124656677246,
|
| 2038 |
+
"logps/chosen": -34.251712799072266,
|
| 2039 |
+
"logps/rejected": -32.091835021972656,
|
| 2040 |
+
"loss": 0.2462,
|
| 2041 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2042 |
+
"rewards/chosen": 0.3124735653400421,
|
| 2043 |
+
"rewards/margins": 0.41325968503952026,
|
| 2044 |
+
"rewards/rejected": -0.10078616440296173,
|
| 2045 |
+
"step": 1240
|
| 2046 |
+
},
|
| 2047 |
+
{
|
| 2048 |
+
"epoch": 3.25,
|
| 2049 |
+
"grad_norm": 113.0,
|
| 2050 |
+
"learning_rate": 5.2093665457911e-07,
|
| 2051 |
+
"logits/chosen": -1.9658873081207275,
|
| 2052 |
+
"logits/rejected": -1.9630600214004517,
|
| 2053 |
+
"logps/chosen": -33.601524353027344,
|
| 2054 |
+
"logps/rejected": -35.11139678955078,
|
| 2055 |
+
"loss": 0.2643,
|
| 2056 |
+
"rewards/accuracies": 0.9375,
|
| 2057 |
+
"rewards/chosen": 0.3178528845310211,
|
| 2058 |
+
"rewards/margins": 0.44355902075767517,
|
| 2059 |
+
"rewards/rejected": -0.12570618093013763,
|
| 2060 |
+
"step": 1250
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 3.27,
|
| 2064 |
+
"grad_norm": 19.75,
|
| 2065 |
+
"learning_rate": 4.868243561723535e-07,
|
| 2066 |
+
"logits/chosen": -2.0280373096466064,
|
| 2067 |
+
"logits/rejected": -2.022881031036377,
|
| 2068 |
+
"logps/chosen": -29.97477149963379,
|
| 2069 |
+
"logps/rejected": -32.641361236572266,
|
| 2070 |
+
"loss": 0.2187,
|
| 2071 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2072 |
+
"rewards/chosen": 0.31440991163253784,
|
| 2073 |
+
"rewards/margins": 0.41793423891067505,
|
| 2074 |
+
"rewards/rejected": -0.103524349629879,
|
| 2075 |
+
"step": 1260
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 3.3,
|
| 2079 |
+
"grad_norm": 25.25,
|
| 2080 |
+
"learning_rate": 4.537463335535161e-07,
|
| 2081 |
+
"logits/chosen": -2.0543882846832275,
|
| 2082 |
+
"logits/rejected": -2.0590672492980957,
|
| 2083 |
+
"logps/chosen": -31.920276641845703,
|
| 2084 |
+
"logps/rejected": -31.91879653930664,
|
| 2085 |
+
"loss": 0.204,
|
| 2086 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2087 |
+
"rewards/chosen": 0.31628847122192383,
|
| 2088 |
+
"rewards/margins": 0.41126084327697754,
|
| 2089 |
+
"rewards/rejected": -0.09497231990098953,
|
| 2090 |
+
"step": 1270
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"epoch": 3.32,
|
| 2094 |
+
"grad_norm": 28.25,
|
| 2095 |
+
"learning_rate": 4.217195806684629e-07,
|
| 2096 |
+
"logits/chosen": -1.9492976665496826,
|
| 2097 |
+
"logits/rejected": -1.9564584493637085,
|
| 2098 |
+
"logps/chosen": -34.14177703857422,
|
| 2099 |
+
"logps/rejected": -32.934364318847656,
|
| 2100 |
+
"loss": 0.2121,
|
| 2101 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2102 |
+
"rewards/chosen": 0.3061862885951996,
|
| 2103 |
+
"rewards/margins": 0.4406258463859558,
|
| 2104 |
+
"rewards/rejected": -0.13443957269191742,
|
| 2105 |
+
"step": 1280
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 3.35,
|
| 2109 |
+
"grad_norm": 30.75,
|
| 2110 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2111 |
+
"logits/chosen": -1.8527300357818604,
|
| 2112 |
+
"logits/rejected": -1.8549416065216064,
|
| 2113 |
+
"logps/chosen": -32.315616607666016,
|
| 2114 |
+
"logps/rejected": -36.80881118774414,
|
| 2115 |
+
"loss": 0.2597,
|
| 2116 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2117 |
+
"rewards/chosen": 0.3265129327774048,
|
| 2118 |
+
"rewards/margins": 0.4283279776573181,
|
| 2119 |
+
"rewards/rejected": -0.10181504487991333,
|
| 2120 |
+
"step": 1290
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 3.38,
|
| 2124 |
+
"grad_norm": 157.0,
|
| 2125 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2126 |
+
"logits/chosen": -1.908172369003296,
|
| 2127 |
+
"logits/rejected": -1.9115606546401978,
|
| 2128 |
+
"logps/chosen": -32.43517303466797,
|
| 2129 |
+
"logps/rejected": -32.76791763305664,
|
| 2130 |
+
"loss": 0.277,
|
| 2131 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2132 |
+
"rewards/chosen": 0.3125966489315033,
|
| 2133 |
+
"rewards/margins": 0.4340974688529968,
|
| 2134 |
+
"rewards/rejected": -0.12150083482265472,
|
| 2135 |
+
"step": 1300
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.38,
|
| 2139 |
+
"eval_logits/chosen": -2.238203287124634,
|
| 2140 |
+
"eval_logits/rejected": -2.233325719833374,
|
| 2141 |
+
"eval_logps/chosen": -33.95594787597656,
|
| 2142 |
+
"eval_logps/rejected": -37.486488342285156,
|
| 2143 |
+
"eval_loss": 1.6102017164230347,
|
| 2144 |
+
"eval_rewards/accuracies": 0.5402824282646179,
|
| 2145 |
+
"eval_rewards/chosen": 0.03144202008843422,
|
| 2146 |
+
"eval_rewards/margins": 0.019388414919376373,
|
| 2147 |
+
"eval_rewards/rejected": 0.012053603306412697,
|
| 2148 |
+
"eval_runtime": 145.3717,
|
| 2149 |
+
"eval_samples_per_second": 2.359,
|
| 2150 |
+
"eval_steps_per_second": 0.296,
|
| 2151 |
+
"step": 1300
|
| 2152 |
+
},
|
| 2153 |
+
{
|
| 2154 |
+
"epoch": 3.4,
|
| 2155 |
+
"grad_norm": 53.0,
|
| 2156 |
+
"learning_rate": 3.321087280364757e-07,
|
| 2157 |
+
"logits/chosen": -2.0169308185577393,
|
| 2158 |
+
"logits/rejected": -1.9967368841171265,
|
| 2159 |
+
"logps/chosen": -30.350967407226562,
|
| 2160 |
+
"logps/rejected": -35.431488037109375,
|
| 2161 |
+
"loss": 0.2812,
|
| 2162 |
+
"rewards/accuracies": 0.9375,
|
| 2163 |
+
"rewards/chosen": 0.30497312545776367,
|
| 2164 |
+
"rewards/margins": 0.41100645065307617,
|
| 2165 |
+
"rewards/rejected": -0.1060333400964737,
|
| 2166 |
+
"step": 1310
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"epoch": 3.43,
|
| 2170 |
+
"grad_norm": 93.0,
|
| 2171 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2172 |
+
"logits/chosen": -1.9958242177963257,
|
| 2173 |
+
"logits/rejected": -2.001598834991455,
|
| 2174 |
+
"logps/chosen": -32.20707702636719,
|
| 2175 |
+
"logps/rejected": -32.74049758911133,
|
| 2176 |
+
"loss": 0.2256,
|
| 2177 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2178 |
+
"rewards/chosen": 0.33068421483039856,
|
| 2179 |
+
"rewards/margins": 0.43058300018310547,
|
| 2180 |
+
"rewards/rejected": -0.09989883005619049,
|
| 2181 |
+
"step": 1320
|
| 2182 |
+
},
|
| 2183 |
+
{
|
| 2184 |
+
"epoch": 3.45,
|
| 2185 |
+
"grad_norm": 54.75,
|
| 2186 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2187 |
+
"logits/chosen": -1.964477300643921,
|
| 2188 |
+
"logits/rejected": -1.9595527648925781,
|
| 2189 |
+
"logps/chosen": -32.27384567260742,
|
| 2190 |
+
"logps/rejected": -34.89734649658203,
|
| 2191 |
+
"loss": 0.2232,
|
| 2192 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2193 |
+
"rewards/chosen": 0.2994646430015564,
|
| 2194 |
+
"rewards/margins": 0.42948657274246216,
|
| 2195 |
+
"rewards/rejected": -0.13002195954322815,
|
| 2196 |
+
"step": 1330
|
| 2197 |
+
},
|
| 2198 |
+
{
|
| 2199 |
+
"epoch": 3.48,
|
| 2200 |
+
"grad_norm": 28.625,
|
| 2201 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2202 |
+
"logits/chosen": -1.8306457996368408,
|
| 2203 |
+
"logits/rejected": -1.840346097946167,
|
| 2204 |
+
"logps/chosen": -32.9451789855957,
|
| 2205 |
+
"logps/rejected": -31.277496337890625,
|
| 2206 |
+
"loss": 0.2011,
|
| 2207 |
+
"rewards/accuracies": 1.0,
|
| 2208 |
+
"rewards/chosen": 0.34176546335220337,
|
| 2209 |
+
"rewards/margins": 0.43911829590797424,
|
| 2210 |
+
"rewards/rejected": -0.09735285490751266,
|
| 2211 |
+
"step": 1340
|
| 2212 |
+
},
|
| 2213 |
+
{
|
| 2214 |
+
"epoch": 3.51,
|
| 2215 |
+
"grad_norm": 21.375,
|
| 2216 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2217 |
+
"logits/chosen": -1.8909282684326172,
|
| 2218 |
+
"logits/rejected": -1.8956327438354492,
|
| 2219 |
+
"logps/chosen": -33.05854797363281,
|
| 2220 |
+
"logps/rejected": -33.51481246948242,
|
| 2221 |
+
"loss": 0.152,
|
| 2222 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2223 |
+
"rewards/chosen": 0.35695692896842957,
|
| 2224 |
+
"rewards/margins": 0.480957567691803,
|
| 2225 |
+
"rewards/rejected": -0.12400064617395401,
|
| 2226 |
+
"step": 1350
|
| 2227 |
+
},
|
| 2228 |
+
{
|
| 2229 |
+
"epoch": 3.53,
|
| 2230 |
+
"grad_norm": 135.0,
|
| 2231 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2232 |
+
"logits/chosen": -1.917721152305603,
|
| 2233 |
+
"logits/rejected": -1.9233052730560303,
|
| 2234 |
+
"logps/chosen": -33.081817626953125,
|
| 2235 |
+
"logps/rejected": -34.85997772216797,
|
| 2236 |
+
"loss": 0.3369,
|
| 2237 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2238 |
+
"rewards/chosen": 0.29993897676467896,
|
| 2239 |
+
"rewards/margins": 0.41870397329330444,
|
| 2240 |
+
"rewards/rejected": -0.11876501142978668,
|
| 2241 |
+
"step": 1360
|
| 2242 |
+
},
|
| 2243 |
+
{
|
| 2244 |
+
"epoch": 3.56,
|
| 2245 |
+
"grad_norm": 40.0,
|
| 2246 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2247 |
+
"logits/chosen": -2.0079197883605957,
|
| 2248 |
+
"logits/rejected": -2.0034046173095703,
|
| 2249 |
+
"logps/chosen": -31.874841690063477,
|
| 2250 |
+
"logps/rejected": -31.1782169342041,
|
| 2251 |
+
"loss": 0.2298,
|
| 2252 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2253 |
+
"rewards/chosen": 0.30433863401412964,
|
| 2254 |
+
"rewards/margins": 0.4108657240867615,
|
| 2255 |
+
"rewards/rejected": -0.10652711242437363,
|
| 2256 |
+
"step": 1370
|
| 2257 |
+
},
|
| 2258 |
+
{
|
| 2259 |
+
"epoch": 3.58,
|
| 2260 |
+
"grad_norm": 39.5,
|
| 2261 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2262 |
+
"logits/chosen": -1.930163025856018,
|
| 2263 |
+
"logits/rejected": -1.9319047927856445,
|
| 2264 |
+
"logps/chosen": -28.90609359741211,
|
| 2265 |
+
"logps/rejected": -31.02655601501465,
|
| 2266 |
+
"loss": 0.2334,
|
| 2267 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2268 |
+
"rewards/chosen": 0.3012833595275879,
|
| 2269 |
+
"rewards/margins": 0.4142329692840576,
|
| 2270 |
+
"rewards/rejected": -0.11294958740472794,
|
| 2271 |
+
"step": 1380
|
| 2272 |
+
},
|
| 2273 |
+
{
|
| 2274 |
+
"epoch": 3.61,
|
| 2275 |
+
"grad_norm": 47.5,
|
| 2276 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2277 |
+
"logits/chosen": -1.9173237085342407,
|
| 2278 |
+
"logits/rejected": -1.9218599796295166,
|
| 2279 |
+
"logps/chosen": -31.18109703063965,
|
| 2280 |
+
"logps/rejected": -33.24029541015625,
|
| 2281 |
+
"loss": 0.1751,
|
| 2282 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2283 |
+
"rewards/chosen": 0.3353577256202698,
|
| 2284 |
+
"rewards/margins": 0.43625006079673767,
|
| 2285 |
+
"rewards/rejected": -0.10089237987995148,
|
| 2286 |
+
"step": 1390
|
| 2287 |
+
},
|
| 2288 |
+
{
|
| 2289 |
+
"epoch": 3.64,
|
| 2290 |
+
"grad_norm": 39.5,
|
| 2291 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2292 |
+
"logits/chosen": -1.9880527257919312,
|
| 2293 |
+
"logits/rejected": -1.9912986755371094,
|
| 2294 |
+
"logps/chosen": -34.084754943847656,
|
| 2295 |
+
"logps/rejected": -33.75811004638672,
|
| 2296 |
+
"loss": 0.139,
|
| 2297 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2298 |
+
"rewards/chosen": 0.33832791447639465,
|
| 2299 |
+
"rewards/margins": 0.47214603424072266,
|
| 2300 |
+
"rewards/rejected": -0.1338181048631668,
|
| 2301 |
+
"step": 1400
|
| 2302 |
+
},
|
| 2303 |
+
{
|
| 2304 |
+
"epoch": 3.64,
|
| 2305 |
+
"eval_logits/chosen": -2.2381296157836914,
|
| 2306 |
+
"eval_logits/rejected": -2.2332541942596436,
|
| 2307 |
+
"eval_logps/chosen": -33.96630096435547,
|
| 2308 |
+
"eval_logps/rejected": -37.49371337890625,
|
| 2309 |
+
"eval_loss": 1.6181015968322754,
|
| 2310 |
+
"eval_rewards/accuracies": 0.5307309031486511,
|
| 2311 |
+
"eval_rewards/chosen": 0.027301045134663582,
|
| 2312 |
+
"eval_rewards/margins": 0.01813744381070137,
|
| 2313 |
+
"eval_rewards/rejected": 0.009163597598671913,
|
| 2314 |
+
"eval_runtime": 145.1897,
|
| 2315 |
+
"eval_samples_per_second": 2.362,
|
| 2316 |
+
"eval_steps_per_second": 0.296,
|
| 2317 |
+
"step": 1400
|
| 2318 |
+
},
|
| 2319 |
+
{
|
| 2320 |
+
"epoch": 3.66,
|
| 2321 |
+
"grad_norm": 16.375,
|
| 2322 |
+
"learning_rate": 1.0775199359171346e-07,
|
| 2323 |
+
"logits/chosen": -2.0566494464874268,
|
| 2324 |
+
"logits/rejected": -2.049795150756836,
|
| 2325 |
+
"logps/chosen": -32.942054748535156,
|
| 2326 |
+
"logps/rejected": -34.50425338745117,
|
| 2327 |
+
"loss": 0.2,
|
| 2328 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2329 |
+
"rewards/chosen": 0.3441971242427826,
|
| 2330 |
+
"rewards/margins": 0.4480733871459961,
|
| 2331 |
+
"rewards/rejected": -0.1038762778043747,
|
| 2332 |
+
"step": 1410
|
| 2333 |
+
},
|
| 2334 |
+
{
|
| 2335 |
+
"epoch": 3.69,
|
| 2336 |
+
"grad_norm": 38.5,
|
| 2337 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2338 |
+
"logits/chosen": -1.942500352859497,
|
| 2339 |
+
"logits/rejected": -1.9532867670059204,
|
| 2340 |
+
"logps/chosen": -33.3380126953125,
|
| 2341 |
+
"logps/rejected": -33.94847869873047,
|
| 2342 |
+
"loss": 0.1618,
|
| 2343 |
+
"rewards/accuracies": 1.0,
|
| 2344 |
+
"rewards/chosen": 0.3245156705379486,
|
| 2345 |
+
"rewards/margins": 0.4674775004386902,
|
| 2346 |
+
"rewards/rejected": -0.1429617702960968,
|
| 2347 |
+
"step": 1420
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.71,
|
| 2351 |
+
"grad_norm": 35.0,
|
| 2352 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2353 |
+
"logits/chosen": -2.0275931358337402,
|
| 2354 |
+
"logits/rejected": -2.039466381072998,
|
| 2355 |
+
"logps/chosen": -34.284996032714844,
|
| 2356 |
+
"logps/rejected": -34.00452423095703,
|
| 2357 |
+
"loss": 0.262,
|
| 2358 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2359 |
+
"rewards/chosen": 0.3103240728378296,
|
| 2360 |
+
"rewards/margins": 0.45679956674575806,
|
| 2361 |
+
"rewards/rejected": -0.14647547900676727,
|
| 2362 |
+
"step": 1430
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"epoch": 3.74,
|
| 2366 |
+
"grad_norm": 47.25,
|
| 2367 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2368 |
+
"logits/chosen": -2.0178730487823486,
|
| 2369 |
+
"logits/rejected": -2.0198729038238525,
|
| 2370 |
+
"logps/chosen": -32.42511749267578,
|
| 2371 |
+
"logps/rejected": -34.111873626708984,
|
| 2372 |
+
"loss": 0.2048,
|
| 2373 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2374 |
+
"rewards/chosen": 0.31330591440200806,
|
| 2375 |
+
"rewards/margins": 0.4140680432319641,
|
| 2376 |
+
"rewards/rejected": -0.10076216608285904,
|
| 2377 |
+
"step": 1440
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.77,
|
| 2381 |
+
"grad_norm": 31.75,
|
| 2382 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2383 |
+
"logits/chosen": -1.9993807077407837,
|
| 2384 |
+
"logits/rejected": -1.992196798324585,
|
| 2385 |
+
"logps/chosen": -34.766136169433594,
|
| 2386 |
+
"logps/rejected": -34.06538772583008,
|
| 2387 |
+
"loss": 0.2089,
|
| 2388 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2389 |
+
"rewards/chosen": 0.32087528705596924,
|
| 2390 |
+
"rewards/margins": 0.43886953592300415,
|
| 2391 |
+
"rewards/rejected": -0.1179942637681961,
|
| 2392 |
+
"step": 1450
|
| 2393 |
+
},
|
| 2394 |
+
{
|
| 2395 |
+
"epoch": 3.79,
|
| 2396 |
+
"grad_norm": 47.5,
|
| 2397 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2398 |
+
"logits/chosen": -1.9658348560333252,
|
| 2399 |
+
"logits/rejected": -1.9719345569610596,
|
| 2400 |
+
"logps/chosen": -30.47488021850586,
|
| 2401 |
+
"logps/rejected": -34.05385971069336,
|
| 2402 |
+
"loss": 0.2022,
|
| 2403 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2404 |
+
"rewards/chosen": 0.29531940817832947,
|
| 2405 |
+
"rewards/margins": 0.44886380434036255,
|
| 2406 |
+
"rewards/rejected": -0.15354439616203308,
|
| 2407 |
+
"step": 1460
|
| 2408 |
+
},
|
| 2409 |
+
{
|
| 2410 |
+
"epoch": 3.82,
|
| 2411 |
+
"grad_norm": 24.625,
|
| 2412 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2413 |
+
"logits/chosen": -1.9123966693878174,
|
| 2414 |
+
"logits/rejected": -1.910093903541565,
|
| 2415 |
+
"logps/chosen": -30.029949188232422,
|
| 2416 |
+
"logps/rejected": -30.82769775390625,
|
| 2417 |
+
"loss": 0.211,
|
| 2418 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2419 |
+
"rewards/chosen": 0.3262413442134857,
|
| 2420 |
+
"rewards/margins": 0.41962748765945435,
|
| 2421 |
+
"rewards/rejected": -0.09338615089654922,
|
| 2422 |
+
"step": 1470
|
| 2423 |
+
},
|
| 2424 |
+
{
|
| 2425 |
+
"epoch": 3.84,
|
| 2426 |
+
"grad_norm": 25.5,
|
| 2427 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2428 |
+
"logits/chosen": -2.0873568058013916,
|
| 2429 |
+
"logits/rejected": -2.0816843509674072,
|
| 2430 |
+
"logps/chosen": -34.48334884643555,
|
| 2431 |
+
"logps/rejected": -32.353607177734375,
|
| 2432 |
+
"loss": 0.1471,
|
| 2433 |
+
"rewards/accuracies": 1.0,
|
| 2434 |
+
"rewards/chosen": 0.32076385617256165,
|
| 2435 |
+
"rewards/margins": 0.43952280282974243,
|
| 2436 |
+
"rewards/rejected": -0.1187589168548584,
|
| 2437 |
+
"step": 1480
|
| 2438 |
+
},
|
| 2439 |
+
{
|
| 2440 |
+
"epoch": 3.87,
|
| 2441 |
+
"grad_norm": 20.125,
|
| 2442 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2443 |
+
"logits/chosen": -1.9781821966171265,
|
| 2444 |
+
"logits/rejected": -1.9803062677383423,
|
| 2445 |
+
"logps/chosen": -27.70733642578125,
|
| 2446 |
+
"logps/rejected": -28.307626724243164,
|
| 2447 |
+
"loss": 0.2222,
|
| 2448 |
+
"rewards/accuracies": 0.9375,
|
| 2449 |
+
"rewards/chosen": 0.3076176047325134,
|
| 2450 |
+
"rewards/margins": 0.4217551648616791,
|
| 2451 |
+
"rewards/rejected": -0.11413755267858505,
|
| 2452 |
+
"step": 1490
|
| 2453 |
+
},
|
| 2454 |
+
{
|
| 2455 |
+
"epoch": 3.9,
|
| 2456 |
+
"grad_norm": 43.0,
|
| 2457 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2458 |
+
"logits/chosen": -1.987160086631775,
|
| 2459 |
+
"logits/rejected": -1.98151433467865,
|
| 2460 |
+
"logps/chosen": -32.05828857421875,
|
| 2461 |
+
"logps/rejected": -32.30227279663086,
|
| 2462 |
+
"loss": 0.24,
|
| 2463 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2464 |
+
"rewards/chosen": 0.2970001995563507,
|
| 2465 |
+
"rewards/margins": 0.43032747507095337,
|
| 2466 |
+
"rewards/rejected": -0.13332727551460266,
|
| 2467 |
+
"step": 1500
|
| 2468 |
+
},
|
| 2469 |
+
{
|
| 2470 |
+
"epoch": 3.9,
|
| 2471 |
+
"eval_logits/chosen": -2.2380595207214355,
|
| 2472 |
+
"eval_logits/rejected": -2.2331745624542236,
|
| 2473 |
+
"eval_logps/chosen": -33.95270919799805,
|
| 2474 |
+
"eval_logps/rejected": -37.47934341430664,
|
| 2475 |
+
"eval_loss": 1.6201497316360474,
|
| 2476 |
+
"eval_rewards/accuracies": 0.52491694688797,
|
| 2477 |
+
"eval_rewards/chosen": 0.032736361026763916,
|
| 2478 |
+
"eval_rewards/margins": 0.017824459820985794,
|
| 2479 |
+
"eval_rewards/rejected": 0.014911895617842674,
|
| 2480 |
+
"eval_runtime": 145.1995,
|
| 2481 |
+
"eval_samples_per_second": 2.362,
|
| 2482 |
+
"eval_steps_per_second": 0.296,
|
| 2483 |
+
"step": 1500
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 3.92,
|
| 2487 |
+
"grad_norm": 26.25,
|
| 2488 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2489 |
+
"logits/chosen": -1.9137132167816162,
|
| 2490 |
+
"logits/rejected": -1.918095350265503,
|
| 2491 |
+
"logps/chosen": -33.41365051269531,
|
| 2492 |
+
"logps/rejected": -34.701316833496094,
|
| 2493 |
+
"loss": 0.2691,
|
| 2494 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2495 |
+
"rewards/chosen": 0.3029375374317169,
|
| 2496 |
+
"rewards/margins": 0.41931143403053284,
|
| 2497 |
+
"rewards/rejected": -0.11637385189533234,
|
| 2498 |
+
"step": 1510
|
| 2499 |
+
},
|
| 2500 |
+
{
|
| 2501 |
+
"epoch": 3.95,
|
| 2502 |
+
"grad_norm": 39.75,
|
| 2503 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2504 |
+
"logits/chosen": -1.973489761352539,
|
| 2505 |
+
"logits/rejected": -1.9717988967895508,
|
| 2506 |
+
"logps/chosen": -29.93434715270996,
|
| 2507 |
+
"logps/rejected": -32.88115692138672,
|
| 2508 |
+
"loss": 0.2589,
|
| 2509 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2510 |
+
"rewards/chosen": 0.31993287801742554,
|
| 2511 |
+
"rewards/margins": 0.4365101456642151,
|
| 2512 |
+
"rewards/rejected": -0.11657728254795074,
|
| 2513 |
+
"step": 1520
|
| 2514 |
+
},
|
| 2515 |
+
{
|
| 2516 |
+
"epoch": 3.97,
|
| 2517 |
+
"grad_norm": 44.75,
|
| 2518 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2519 |
+
"logits/chosen": -2.070810556411743,
|
| 2520 |
+
"logits/rejected": -2.063199043273926,
|
| 2521 |
+
"logps/chosen": -27.3148250579834,
|
| 2522 |
+
"logps/rejected": -28.99945068359375,
|
| 2523 |
+
"loss": 0.1829,
|
| 2524 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2525 |
+
"rewards/chosen": 0.2992464005947113,
|
| 2526 |
+
"rewards/margins": 0.4303843379020691,
|
| 2527 |
+
"rewards/rejected": -0.1311378926038742,
|
| 2528 |
+
"step": 1530
|
| 2529 |
+
},
|
| 2530 |
+
{
|
| 2531 |
+
"epoch": 4.0,
|
| 2532 |
+
"grad_norm": 28.0,
|
| 2533 |
+
"learning_rate": 0.0,
|
| 2534 |
+
"logits/chosen": -1.9703433513641357,
|
| 2535 |
+
"logits/rejected": -1.9601656198501587,
|
| 2536 |
+
"logps/chosen": -32.22865295410156,
|
| 2537 |
+
"logps/rejected": -35.73883056640625,
|
| 2538 |
+
"loss": 0.1937,
|
| 2539 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2540 |
+
"rewards/chosen": 0.33501744270324707,
|
| 2541 |
+
"rewards/margins": 0.4660801291465759,
|
| 2542 |
+
"rewards/rejected": -0.13106267154216766,
|
| 2543 |
+
"step": 1540
|
| 2544 |
+
},
|
| 2545 |
+
{
|
| 2546 |
+
"epoch": 4.0,
|
| 2547 |
+
"step": 1540,
|
| 2548 |
"total_flos": 0.0,
|
| 2549 |
+
"train_loss": 0.36088944433571457,
|
| 2550 |
+
"train_runtime": 10769.3493,
|
| 2551 |
+
"train_samples_per_second": 1.144,
|
| 2552 |
+
"train_steps_per_second": 0.143
|
| 2553 |
}
|
| 2554 |
],
|
| 2555 |
"logging_steps": 10,
|
| 2556 |
+
"max_steps": 1540,
|
| 2557 |
"num_input_tokens_seen": 0,
|
| 2558 |
+
"num_train_epochs": 4,
|
| 2559 |
"save_steps": 100,
|
| 2560 |
"total_flos": 0.0,
|
| 2561 |
"train_batch_size": 4,
|