Commit
•
13707dc
1
Parent(s):
9e98df7
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: 0.
|
| 24 |
-
- Rewards/chosen: 0.
|
| 25 |
-
- Rewards/rejected: 0.
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -33.
|
| 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 |
-
| 0.4799 | 0.26 | 100 |
|
| 67 |
-
| 0.4453 | 0.52 | 200 |
|
| 68 |
-
| 0.3947 | 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: 0.4747
|
| 22 |
+
- Rewards/chosen: 0.2822
|
| 23 |
+
- Rewards/rejected: 0.1614
|
| 24 |
+
- Rewards/accuracies: 0.5569
|
| 25 |
+
- Rewards/margins: 0.1208
|
| 26 |
+
- Logps/rejected: -37.2860
|
| 27 |
+
- Logps/chosen: -33.6314
|
| 28 |
+
- Logits/rejected: -2.1560
|
| 29 |
+
- Logits/chosen: -2.1607
|
| 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 |
+
| 0.4799 | 0.26 | 100 | -2.2381 | -2.2333 | -33.8607 | -37.3570 | 0.4978 | 0.5341 | 0.1217 | 0.0100 | 0.1117 |
|
| 65 |
+
| 0.4453 | 0.52 | 200 | -2.2347 | -2.2299 | -33.7685 | -37.2937 | 0.4928 | 0.5370 | 0.1862 | 0.0302 | 0.1561 |
|
| 66 |
+
| 0.3947 | 0.78 | 300 | -2.2322 | -2.2274 | -33.7551 | -37.2894 | 0.4910 | 0.5565 | 0.1956 | 0.0365 | 0.1591 |
|
| 67 |
+
| 0.3136 | 1.04 | 400 | 0.4857 | 0.2846 | 0.2244 | 0.5797 | 0.0602 | -37.1961 | -33.6280 | -2.2032 | -2.2080 |
|
| 68 |
+
| 0.2784 | 1.3 | 500 | 0.4891 | 0.2959 | 0.2519 | 0.5220 | 0.0439 | -37.1567 | -33.6119 | -2.2050 | -2.2098 |
|
| 69 |
+
| 0.2593 | 1.56 | 600 | 0.4795 | 0.3345 | 0.2439 | 0.5743 | 0.0906 | -37.1682 | -33.5567 | -2.1866 | -2.1914 |
|
| 70 |
+
| 0.2606 | 1.82 | 700 | 0.4764 | 0.3188 | 0.2158 | 0.6063 | 0.1031 | -37.2084 | -33.5791 | -2.1788 | -2.1836 |
|
| 71 |
+
| 0.1758 | 2.08 | 800 | 0.4767 | 0.2840 | 0.1749 | 0.5860 | 0.1091 | -37.2668 | -33.6289 | -2.1680 | -2.1727 |
|
| 72 |
+
| 0.1687 | 2.34 | 900 | 0.4770 | 0.2898 | 0.1833 | 0.5486 | 0.1065 | -37.2547 | -33.6205 | -2.1626 | -2.1674 |
|
| 73 |
+
| 0.1826 | 2.6 | 1000 | 0.4764 | 0.2700 | 0.1574 | 0.5831 | 0.1126 | -37.2917 | -33.6489 | -2.1578 | -2.1625 |
|
| 74 |
+
| 0.1541 | 2.86 | 1100 | 0.4751 | 0.2864 | 0.1692 | 0.5777 | 0.1171 | -37.2748 | -33.6254 | -2.1561 | -2.1608 |
|
| 75 |
+
| 0.194 | 3.12 | 1200 | 0.4748 | 0.2856 | 0.1654 | 0.5801 | 0.1202 | -37.2803 | -33.6265 | -2.1565 | -2.1612 |
|
| 76 |
+
| 0.1414 | 3.38 | 1300 | 0.4753 | 0.2859 | 0.1690 | 0.5831 | 0.1169 | -37.2751 | -33.6261 | -2.1558 | -2.1605 |
|
| 77 |
+
| 0.1492 | 3.64 | 1400 | 0.4744 | 0.2846 | 0.1627 | 0.5918 | 0.1220 | -37.2842 | -33.6279 | -2.1556 | -2.1603 |
|
| 78 |
+
| 0.1694 | 3.9 | 1500 | 0.4747 | 0.2822 | 0.1614 | 0.5569 | 0.1208 | -37.2860 | -33.6314 | -2.1560 | -2.1607 |
|
| 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:e4f8e306a73a999a41c8de964894eaa93994481de6722767ddb43b3c17557965
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.2319846153259277,
|
| 4 |
"eval_logits/rejected": -2.227167844772339,
|
| 5 |
"eval_logps/chosen": -33.76113510131836,
|
|
@@ -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": 0.
|
| 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": -2.2319846153259277,
|
| 4 |
"eval_logits/rejected": -2.227167844772339,
|
| 5 |
"eval_logps/chosen": -33.76113510131836,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.355,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.17319794591371115,
|
| 17 |
+
"train_runtime": 10768.0107,
|
| 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": 0.
|
| 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": 0.17319794591371115,
|
| 4 |
+
"train_runtime": 10768.0107,
|
| 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": 0.
|
| 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": 5.65625,
|
| 496 |
+
"learning_rate": 4.84533120650964e-06,
|
| 497 |
+
"logits/chosen": -2.069279909133911,
|
| 498 |
+
"logits/rejected": -2.0565598011016846,
|
| 499 |
+
"logps/chosen": -31.766155242919922,
|
| 500 |
+
"logps/rejected": -32.60447311401367,
|
| 501 |
+
"loss": 0.3879,
|
| 502 |
+
"rewards/accuracies": 0.875,
|
| 503 |
+
"rewards/chosen": 0.5335645079612732,
|
| 504 |
+
"rewards/margins": 0.5079580545425415,
|
| 505 |
+
"rewards/rejected": 0.025606412440538406,
|
| 506 |
"step": 310
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.83,
|
| 510 |
+
"grad_norm": 6.9375,
|
| 511 |
+
"learning_rate": 4.825108134172131e-06,
|
| 512 |
+
"logits/chosen": -1.9792858362197876,
|
| 513 |
+
"logits/rejected": -1.9706569910049438,
|
| 514 |
+
"logps/chosen": -31.41892433166504,
|
| 515 |
+
"logps/rejected": -30.155033111572266,
|
| 516 |
+
"loss": 0.3743,
|
| 517 |
+
"rewards/accuracies": 0.800000011920929,
|
| 518 |
+
"rewards/chosen": 0.6552442312240601,
|
| 519 |
+
"rewards/margins": 0.5947929620742798,
|
| 520 |
+
"rewards/rejected": 0.060451291501522064,
|
| 521 |
"step": 320
|
| 522 |
},
|
| 523 |
{
|
| 524 |
"epoch": 0.86,
|
| 525 |
+
"grad_norm": 8.1875,
|
| 526 |
+
"learning_rate": 4.80369052967602e-06,
|
| 527 |
+
"logits/chosen": -1.9177625179290771,
|
| 528 |
+
"logits/rejected": -1.9297670125961304,
|
| 529 |
+
"logps/chosen": -29.455638885498047,
|
| 530 |
+
"logps/rejected": -33.40266036987305,
|
| 531 |
+
"loss": 0.3501,
|
| 532 |
+
"rewards/accuracies": 0.762499988079071,
|
| 533 |
+
"rewards/chosen": 0.7165408730506897,
|
| 534 |
+
"rewards/margins": 0.7168257832527161,
|
| 535 |
+
"rewards/rejected": -0.00028486884548328817,
|
| 536 |
"step": 330
|
| 537 |
},
|
| 538 |
{
|
| 539 |
"epoch": 0.88,
|
| 540 |
+
"grad_norm": 9.25,
|
| 541 |
+
"learning_rate": 4.781089396387968e-06,
|
| 542 |
+
"logits/chosen": -1.880617380142212,
|
| 543 |
+
"logits/rejected": -1.8715527057647705,
|
| 544 |
+
"logps/chosen": -33.60810089111328,
|
| 545 |
+
"logps/rejected": -35.90662384033203,
|
| 546 |
+
"loss": 0.3405,
|
| 547 |
+
"rewards/accuracies": 0.887499988079071,
|
| 548 |
+
"rewards/chosen": 0.7635797262191772,
|
| 549 |
+
"rewards/margins": 0.7985055446624756,
|
| 550 |
+
"rewards/rejected": -0.03492575138807297,
|
| 551 |
"step": 340
|
| 552 |
},
|
| 553 |
{
|
| 554 |
"epoch": 0.91,
|
| 555 |
+
"grad_norm": 5.1875,
|
| 556 |
+
"learning_rate": 4.757316345716554e-06,
|
| 557 |
+
"logits/chosen": -1.9337537288665771,
|
| 558 |
+
"logits/rejected": -1.9343976974487305,
|
| 559 |
+
"logps/chosen": -33.176937103271484,
|
| 560 |
+
"logps/rejected": -33.779075622558594,
|
| 561 |
+
"loss": 0.3349,
|
| 562 |
+
"rewards/accuracies": 0.800000011920929,
|
| 563 |
+
"rewards/chosen": 0.8598405718803406,
|
| 564 |
+
"rewards/margins": 0.8109108805656433,
|
| 565 |
+
"rewards/rejected": 0.048929743468761444,
|
| 566 |
"step": 350
|
| 567 |
},
|
| 568 |
{
|
| 569 |
"epoch": 0.94,
|
| 570 |
+
"grad_norm": 6.3125,
|
| 571 |
+
"learning_rate": 4.73238359114687e-06,
|
| 572 |
+
"logits/chosen": -2.060123920440674,
|
| 573 |
+
"logits/rejected": -2.066301107406616,
|
| 574 |
+
"logps/chosen": -30.665380477905273,
|
| 575 |
+
"logps/rejected": -32.581119537353516,
|
| 576 |
+
"loss": 0.3792,
|
| 577 |
+
"rewards/accuracies": 0.800000011920929,
|
| 578 |
+
"rewards/chosen": 0.6355056762695312,
|
| 579 |
+
"rewards/margins": 0.5744917988777161,
|
| 580 |
+
"rewards/rejected": 0.061013948172330856,
|
| 581 |
"step": 360
|
| 582 |
},
|
| 583 |
{
|
| 584 |
"epoch": 0.96,
|
| 585 |
+
"grad_norm": 7.78125,
|
| 586 |
+
"learning_rate": 4.706303941965804e-06,
|
| 587 |
+
"logits/chosen": -1.9897676706314087,
|
| 588 |
+
"logits/rejected": -1.989344835281372,
|
| 589 |
+
"logps/chosen": -32.362205505371094,
|
| 590 |
+
"logps/rejected": -35.91802215576172,
|
| 591 |
+
"loss": 0.3617,
|
| 592 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 593 |
+
"rewards/chosen": 0.7579163908958435,
|
| 594 |
+
"rewards/margins": 0.6656854748725891,
|
| 595 |
+
"rewards/rejected": 0.09223100543022156,
|
| 596 |
"step": 370
|
| 597 |
},
|
| 598 |
{
|
| 599 |
"epoch": 0.99,
|
| 600 |
+
"grad_norm": 5.71875,
|
| 601 |
+
"learning_rate": 4.679090796681225e-06,
|
| 602 |
+
"logits/chosen": -2.022163152694702,
|
| 603 |
+
"logits/rejected": -2.017575740814209,
|
| 604 |
+
"logps/chosen": -29.644603729248047,
|
| 605 |
+
"logps/rejected": -29.16855239868164,
|
| 606 |
+
"loss": 0.3514,
|
| 607 |
+
"rewards/accuracies": 0.8125,
|
| 608 |
+
"rewards/chosen": 0.746194064617157,
|
| 609 |
+
"rewards/margins": 0.714457094669342,
|
| 610 |
+
"rewards/rejected": 0.03173698112368584,
|
| 611 |
"step": 380
|
| 612 |
},
|
| 613 |
{
|
| 614 |
+
"epoch": 1.01,
|
| 615 |
+
"grad_norm": 8.0625,
|
| 616 |
+
"learning_rate": 4.650758136138454e-06,
|
| 617 |
+
"logits/chosen": -1.7921892404556274,
|
| 618 |
+
"logits/rejected": -1.7986253499984741,
|
| 619 |
+
"logps/chosen": -31.069808959960938,
|
| 620 |
+
"logps/rejected": -36.217376708984375,
|
| 621 |
+
"loss": 0.3052,
|
| 622 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 623 |
+
"rewards/chosen": 0.9211344718933105,
|
| 624 |
+
"rewards/margins": 0.9904948472976685,
|
| 625 |
+
"rewards/rejected": -0.06936033815145493,
|
| 626 |
+
"step": 390
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 1.04,
|
| 630 |
+
"grad_norm": 5.84375,
|
| 631 |
+
"learning_rate": 4.621320516337559e-06,
|
| 632 |
+
"logits/chosen": -1.946427583694458,
|
| 633 |
+
"logits/rejected": -1.9403005838394165,
|
| 634 |
+
"logps/chosen": -32.449764251708984,
|
| 635 |
+
"logps/rejected": -32.27871322631836,
|
| 636 |
+
"loss": 0.3136,
|
| 637 |
+
"rewards/accuracies": 0.800000011920929,
|
| 638 |
+
"rewards/chosen": 0.968052864074707,
|
| 639 |
+
"rewards/margins": 0.9469360113143921,
|
| 640 |
+
"rewards/rejected": 0.021116863936185837,
|
| 641 |
+
"step": 400
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"epoch": 1.04,
|
| 645 |
+
"eval_logits/chosen": -2.2079687118530273,
|
| 646 |
+
"eval_logits/rejected": -2.2031569480895996,
|
| 647 |
+
"eval_logps/chosen": -33.62800598144531,
|
| 648 |
+
"eval_logps/rejected": -37.196067810058594,
|
| 649 |
+
"eval_loss": 0.48570817708969116,
|
| 650 |
+
"eval_rewards/accuracies": 0.579734206199646,
|
| 651 |
+
"eval_rewards/chosen": 0.2845819294452667,
|
| 652 |
+
"eval_rewards/margins": 0.06019504740834236,
|
| 653 |
+
"eval_rewards/rejected": 0.22438688576221466,
|
| 654 |
+
"eval_runtime": 145.4274,
|
| 655 |
+
"eval_samples_per_second": 2.359,
|
| 656 |
+
"eval_steps_per_second": 0.296,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 1.06,
|
| 661 |
+
"grad_norm": 5.96875,
|
| 662 |
+
"learning_rate": 4.590793060955158e-06,
|
| 663 |
+
"logits/chosen": -1.9444068670272827,
|
| 664 |
+
"logits/rejected": -1.951677918434143,
|
| 665 |
+
"logps/chosen": -28.08418846130371,
|
| 666 |
+
"logps/rejected": -29.256505966186523,
|
| 667 |
+
"loss": 0.3391,
|
| 668 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 669 |
+
"rewards/chosen": 0.7253325581550598,
|
| 670 |
+
"rewards/margins": 0.836132824420929,
|
| 671 |
+
"rewards/rejected": -0.11080024391412735,
|
| 672 |
+
"step": 410
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 1.09,
|
| 676 |
+
"grad_norm": 5.4375,
|
| 677 |
+
"learning_rate": 4.559191453574582e-06,
|
| 678 |
+
"logits/chosen": -1.966160774230957,
|
| 679 |
+
"logits/rejected": -1.965192437171936,
|
| 680 |
+
"logps/chosen": -32.974876403808594,
|
| 681 |
+
"logps/rejected": -30.77309226989746,
|
| 682 |
+
"loss": 0.3464,
|
| 683 |
+
"rewards/accuracies": 0.862500011920929,
|
| 684 |
+
"rewards/chosen": 0.8930822610855103,
|
| 685 |
+
"rewards/margins": 0.774114727973938,
|
| 686 |
+
"rewards/rejected": 0.11896754801273346,
|
| 687 |
+
"step": 420
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 1.12,
|
| 691 |
+
"grad_norm": 6.09375,
|
| 692 |
+
"learning_rate": 4.52653192962838e-06,
|
| 693 |
+
"logits/chosen": -1.9592077732086182,
|
| 694 |
+
"logits/rejected": -1.9421688318252563,
|
| 695 |
+
"logps/chosen": -29.91006851196289,
|
| 696 |
+
"logps/rejected": -33.13286590576172,
|
| 697 |
+
"loss": 0.3285,
|
| 698 |
+
"rewards/accuracies": 0.887499988079071,
|
| 699 |
+
"rewards/chosen": 0.7469960451126099,
|
| 700 |
+
"rewards/margins": 0.8916017413139343,
|
| 701 |
+
"rewards/rejected": -0.14460588991641998,
|
| 702 |
+
"step": 430
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 1.14,
|
| 706 |
+
"grad_norm": 5.3125,
|
| 707 |
+
"learning_rate": 4.492831268057307e-06,
|
| 708 |
+
"logits/chosen": -1.989438772201538,
|
| 709 |
+
"logits/rejected": -1.9913489818572998,
|
| 710 |
+
"logps/chosen": -35.05497741699219,
|
| 711 |
+
"logps/rejected": -34.920108795166016,
|
| 712 |
+
"loss": 0.2832,
|
| 713 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 714 |
+
"rewards/chosen": 1.0901695489883423,
|
| 715 |
+
"rewards/margins": 1.1322667598724365,
|
| 716 |
+
"rewards/rejected": -0.0420973002910614,
|
| 717 |
+
"step": 440
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.17,
|
| 721 |
+
"grad_norm": 6.5,
|
| 722 |
+
"learning_rate": 4.458106782690094e-06,
|
| 723 |
+
"logits/chosen": -2.0655436515808105,
|
| 724 |
+
"logits/rejected": -2.0654349327087402,
|
| 725 |
+
"logps/chosen": -31.32794761657715,
|
| 726 |
+
"logps/rejected": -33.56629180908203,
|
| 727 |
+
"loss": 0.3287,
|
| 728 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 729 |
+
"rewards/chosen": 0.9137799143791199,
|
| 730 |
+
"rewards/margins": 0.9181026220321655,
|
| 731 |
+
"rewards/rejected": -0.004322645254433155,
|
| 732 |
+
"step": 450
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"epoch": 1.19,
|
| 736 |
+
"grad_norm": 5.28125,
|
| 737 |
+
"learning_rate": 4.422376313348405e-06,
|
| 738 |
+
"logits/chosen": -2.0107359886169434,
|
| 739 |
+
"logits/rejected": -2.0033280849456787,
|
| 740 |
+
"logps/chosen": -30.83009910583496,
|
| 741 |
+
"logps/rejected": -36.22407150268555,
|
| 742 |
+
"loss": 0.2765,
|
| 743 |
+
"rewards/accuracies": 0.862500011920929,
|
| 744 |
+
"rewards/chosen": 1.1031007766723633,
|
| 745 |
+
"rewards/margins": 1.2221224308013916,
|
| 746 |
+
"rewards/rejected": -0.11902163922786713,
|
| 747 |
+
"step": 460
|
| 748 |
+
},
|
| 749 |
+
{
|
| 750 |
+
"epoch": 1.22,
|
| 751 |
+
"grad_norm": 6.28125,
|
| 752 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 753 |
+
"logits/chosen": -1.9149672985076904,
|
| 754 |
+
"logits/rejected": -1.9115540981292725,
|
| 755 |
+
"logps/chosen": -32.367496490478516,
|
| 756 |
+
"logps/rejected": -32.911216735839844,
|
| 757 |
+
"loss": 0.2962,
|
| 758 |
+
"rewards/accuracies": 0.824999988079071,
|
| 759 |
+
"rewards/chosen": 1.0953609943389893,
|
| 760 |
+
"rewards/margins": 1.0864982604980469,
|
| 761 |
+
"rewards/rejected": 0.008862635120749474,
|
| 762 |
+
"step": 470
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 1.25,
|
| 766 |
+
"grad_norm": 5.5,
|
| 767 |
+
"learning_rate": 4.347971356735789e-06,
|
| 768 |
+
"logits/chosen": -2.040391445159912,
|
| 769 |
+
"logits/rejected": -2.033470630645752,
|
| 770 |
+
"logps/chosen": -29.63046646118164,
|
| 771 |
+
"logps/rejected": -32.03357696533203,
|
| 772 |
+
"loss": 0.3336,
|
| 773 |
+
"rewards/accuracies": 0.8125,
|
| 774 |
+
"rewards/chosen": 0.8386493921279907,
|
| 775 |
+
"rewards/margins": 0.8667130470275879,
|
| 776 |
+
"rewards/rejected": -0.028063708916306496,
|
| 777 |
+
"step": 480
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 1.27,
|
| 781 |
+
"grad_norm": 5.59375,
|
| 782 |
+
"learning_rate": 4.309335095262675e-06,
|
| 783 |
+
"logits/chosen": -1.9858951568603516,
|
| 784 |
+
"logits/rejected": -1.987535834312439,
|
| 785 |
+
"logps/chosen": -33.76915740966797,
|
| 786 |
+
"logps/rejected": -33.62469482421875,
|
| 787 |
+
"loss": 0.2834,
|
| 788 |
+
"rewards/accuracies": 0.862500011920929,
|
| 789 |
+
"rewards/chosen": 1.108822226524353,
|
| 790 |
+
"rewards/margins": 1.0970509052276611,
|
| 791 |
+
"rewards/rejected": 0.011771360412240028,
|
| 792 |
+
"step": 490
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 1.3,
|
| 796 |
+
"grad_norm": 6.375,
|
| 797 |
+
"learning_rate": 4.269769281772082e-06,
|
| 798 |
+
"logits/chosen": -1.8744853734970093,
|
| 799 |
+
"logits/rejected": -1.8722003698349,
|
| 800 |
+
"logps/chosen": -31.54471206665039,
|
| 801 |
+
"logps/rejected": -36.603973388671875,
|
| 802 |
+
"loss": 0.2784,
|
| 803 |
+
"rewards/accuracies": 0.862500011920929,
|
| 804 |
+
"rewards/chosen": 1.1393038034439087,
|
| 805 |
+
"rewards/margins": 1.2370599508285522,
|
| 806 |
+
"rewards/rejected": -0.09775615483522415,
|
| 807 |
+
"step": 500
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 1.3,
|
| 811 |
+
"eval_logits/chosen": -2.2098300457000732,
|
| 812 |
+
"eval_logits/rejected": -2.205024480819702,
|
| 813 |
+
"eval_logps/chosen": -33.61191177368164,
|
| 814 |
+
"eval_logps/rejected": -37.15669250488281,
|
| 815 |
+
"eval_loss": 0.48908427357673645,
|
| 816 |
+
"eval_rewards/accuracies": 0.5220099687576294,
|
| 817 |
+
"eval_rewards/chosen": 0.29585084319114685,
|
| 818 |
+
"eval_rewards/margins": 0.04390101507306099,
|
| 819 |
+
"eval_rewards/rejected": 0.25194987654685974,
|
| 820 |
+
"eval_runtime": 145.3523,
|
| 821 |
+
"eval_samples_per_second": 2.36,
|
| 822 |
+
"eval_steps_per_second": 0.296,
|
| 823 |
+
"step": 500
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 1.32,
|
| 827 |
+
"grad_norm": 4.75,
|
| 828 |
+
"learning_rate": 4.22929424333435e-06,
|
| 829 |
+
"logits/chosen": -1.978362798690796,
|
| 830 |
+
"logits/rejected": -1.98308527469635,
|
| 831 |
+
"logps/chosen": -31.69674301147461,
|
| 832 |
+
"logps/rejected": -31.60154151916504,
|
| 833 |
+
"loss": 0.2762,
|
| 834 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 835 |
+
"rewards/chosen": 1.2324289083480835,
|
| 836 |
+
"rewards/margins": 1.2012327909469604,
|
| 837 |
+
"rewards/rejected": 0.03119627758860588,
|
| 838 |
+
"step": 510
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 1.35,
|
| 842 |
+
"grad_norm": 5.5625,
|
| 843 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 844 |
+
"logits/chosen": -2.0070841312408447,
|
| 845 |
+
"logits/rejected": -2.01804256439209,
|
| 846 |
+
"logps/chosen": -29.7496280670166,
|
| 847 |
+
"logps/rejected": -31.527536392211914,
|
| 848 |
+
"loss": 0.2821,
|
| 849 |
+
"rewards/accuracies": 0.875,
|
| 850 |
+
"rewards/chosen": 1.148550271987915,
|
| 851 |
+
"rewards/margins": 1.16244375705719,
|
| 852 |
+
"rewards/rejected": -0.013893580064177513,
|
| 853 |
+
"step": 520
|
| 854 |
+
},
|
| 855 |
+
{
|
| 856 |
+
"epoch": 1.38,
|
| 857 |
+
"grad_norm": 4.09375,
|
| 858 |
+
"learning_rate": 4.145700124802693e-06,
|
| 859 |
+
"logits/chosen": -1.9333133697509766,
|
| 860 |
+
"logits/rejected": -1.9299663305282593,
|
| 861 |
+
"logps/chosen": -30.964946746826172,
|
| 862 |
+
"logps/rejected": -32.423377990722656,
|
| 863 |
+
"loss": 0.2962,
|
| 864 |
+
"rewards/accuracies": 0.887499988079071,
|
| 865 |
+
"rewards/chosen": 1.0188950300216675,
|
| 866 |
+
"rewards/margins": 1.061314344406128,
|
| 867 |
+
"rewards/rejected": -0.0424191839993,
|
| 868 |
+
"step": 530
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"epoch": 1.4,
|
| 872 |
+
"grad_norm": 6.15625,
|
| 873 |
+
"learning_rate": 4.102623991469562e-06,
|
| 874 |
+
"logits/chosen": -1.7982631921768188,
|
| 875 |
+
"logits/rejected": -1.8074582815170288,
|
| 876 |
+
"logps/chosen": -30.847604751586914,
|
| 877 |
+
"logps/rejected": -31.964679718017578,
|
| 878 |
+
"loss": 0.2773,
|
| 879 |
+
"rewards/accuracies": 0.862500011920929,
|
| 880 |
+
"rewards/chosen": 1.205248475074768,
|
| 881 |
+
"rewards/margins": 1.2813327312469482,
|
| 882 |
+
"rewards/rejected": -0.07608427852392197,
|
| 883 |
+
"step": 540
|
| 884 |
+
},
|
| 885 |
+
{
|
| 886 |
+
"epoch": 1.43,
|
| 887 |
+
"grad_norm": 5.40625,
|
| 888 |
+
"learning_rate": 4.058724504646834e-06,
|
| 889 |
+
"logits/chosen": -1.8933544158935547,
|
| 890 |
+
"logits/rejected": -1.8870652914047241,
|
| 891 |
+
"logps/chosen": -31.835384368896484,
|
| 892 |
+
"logps/rejected": -30.871362686157227,
|
| 893 |
+
"loss": 0.2807,
|
| 894 |
+
"rewards/accuracies": 0.875,
|
| 895 |
+
"rewards/chosen": 1.3110452890396118,
|
| 896 |
+
"rewards/margins": 1.2270557880401611,
|
| 897 |
+
"rewards/rejected": 0.08398932963609695,
|
| 898 |
+
"step": 550
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 1.45,
|
| 902 |
+
"grad_norm": 5.15625,
|
| 903 |
+
"learning_rate": 4.014024217844167e-06,
|
| 904 |
+
"logits/chosen": -1.9839521646499634,
|
| 905 |
+
"logits/rejected": -1.9819648265838623,
|
| 906 |
+
"logps/chosen": -32.755332946777344,
|
| 907 |
+
"logps/rejected": -31.447025299072266,
|
| 908 |
+
"loss": 0.2809,
|
| 909 |
+
"rewards/accuracies": 0.824999988079071,
|
| 910 |
+
"rewards/chosen": 1.271106481552124,
|
| 911 |
+
"rewards/margins": 1.2358547449111938,
|
| 912 |
+
"rewards/rejected": 0.035251684486866,
|
| 913 |
+
"step": 560
|
| 914 |
+
},
|
| 915 |
+
{
|
| 916 |
+
"epoch": 1.48,
|
| 917 |
+
"grad_norm": 6.34375,
|
| 918 |
+
"learning_rate": 3.968546095984911e-06,
|
| 919 |
+
"logits/chosen": -1.8162040710449219,
|
| 920 |
+
"logits/rejected": -1.814014196395874,
|
| 921 |
+
"logps/chosen": -31.14450454711914,
|
| 922 |
+
"logps/rejected": -30.988576889038086,
|
| 923 |
+
"loss": 0.2898,
|
| 924 |
+
"rewards/accuracies": 0.862500011920929,
|
| 925 |
+
"rewards/chosen": 1.1563327312469482,
|
| 926 |
+
"rewards/margins": 1.146722435951233,
|
| 927 |
+
"rewards/rejected": 0.009610283188521862,
|
| 928 |
+
"step": 570
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"epoch": 1.51,
|
| 932 |
+
"grad_norm": 5.5625,
|
| 933 |
+
"learning_rate": 3.922313503607806e-06,
|
| 934 |
+
"logits/chosen": -1.95193612575531,
|
| 935 |
+
"logits/rejected": -1.948587417602539,
|
| 936 |
+
"logps/chosen": -29.390304565429688,
|
| 937 |
+
"logps/rejected": -34.86743927001953,
|
| 938 |
+
"loss": 0.2599,
|
| 939 |
+
"rewards/accuracies": 0.862500011920929,
|
| 940 |
+
"rewards/chosen": 1.2000386714935303,
|
| 941 |
+
"rewards/margins": 1.3689311742782593,
|
| 942 |
+
"rewards/rejected": -0.16889244318008423,
|
| 943 |
+
"step": 580
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.53,
|
| 947 |
+
"grad_norm": 5.1875,
|
| 948 |
+
"learning_rate": 3.875350192863368e-06,
|
| 949 |
+
"logits/chosen": -1.8900426626205444,
|
| 950 |
+
"logits/rejected": -1.8936386108398438,
|
| 951 |
+
"logps/chosen": -28.26003646850586,
|
| 952 |
+
"logps/rejected": -30.714879989624023,
|
| 953 |
+
"loss": 0.3068,
|
| 954 |
+
"rewards/accuracies": 0.800000011920929,
|
| 955 |
+
"rewards/chosen": 0.9484975934028625,
|
| 956 |
+
"rewards/margins": 0.9877050518989563,
|
| 957 |
+
"rewards/rejected": -0.03920745849609375,
|
| 958 |
+
"step": 590
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 1.56,
|
| 962 |
+
"grad_norm": 6.03125,
|
| 963 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 964 |
+
"logits/chosen": -1.9284902811050415,
|
| 965 |
+
"logits/rejected": -1.9283053874969482,
|
| 966 |
+
"logps/chosen": -30.269977569580078,
|
| 967 |
+
"logps/rejected": -31.168771743774414,
|
| 968 |
+
"loss": 0.2593,
|
| 969 |
+
"rewards/accuracies": 0.875,
|
| 970 |
+
"rewards/chosen": 1.3660885095596313,
|
| 971 |
+
"rewards/margins": 1.3491170406341553,
|
| 972 |
+
"rewards/rejected": 0.016971405595541,
|
| 973 |
+
"step": 600
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.56,
|
| 977 |
+
"eval_logits/chosen": -2.1913671493530273,
|
| 978 |
+
"eval_logits/rejected": -2.186584949493408,
|
| 979 |
+
"eval_logps/chosen": -33.55670166015625,
|
| 980 |
+
"eval_logps/rejected": -37.16816711425781,
|
| 981 |
+
"eval_loss": 0.4795128405094147,
|
| 982 |
+
"eval_rewards/accuracies": 0.574335515499115,
|
| 983 |
+
"eval_rewards/chosen": 0.3344981372356415,
|
| 984 |
+
"eval_rewards/margins": 0.09058075398206711,
|
| 985 |
+
"eval_rewards/rejected": 0.24391737580299377,
|
| 986 |
+
"eval_runtime": 145.1862,
|
| 987 |
+
"eval_samples_per_second": 2.362,
|
| 988 |
+
"eval_steps_per_second": 0.296,
|
| 989 |
+
"step": 600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 1.58,
|
| 993 |
+
"grad_norm": 5.625,
|
| 994 |
+
"learning_rate": 3.7793282895240927e-06,
|
| 995 |
+
"logits/chosen": -1.9897689819335938,
|
| 996 |
+
"logits/rejected": -1.9904531240463257,
|
| 997 |
+
"logps/chosen": -32.8941650390625,
|
| 998 |
+
"logps/rejected": -33.11201477050781,
|
| 999 |
+
"loss": 0.2422,
|
| 1000 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1001 |
+
"rewards/chosen": 1.4951863288879395,
|
| 1002 |
+
"rewards/margins": 1.5146669149398804,
|
| 1003 |
+
"rewards/rejected": -0.019480543211102486,
|
| 1004 |
+
"step": 610
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 1.61,
|
| 1008 |
+
"grad_norm": 3.875,
|
| 1009 |
+
"learning_rate": 3.730319028506478e-06,
|
| 1010 |
+
"logits/chosen": -1.953918695449829,
|
| 1011 |
+
"logits/rejected": -1.9515819549560547,
|
| 1012 |
+
"logps/chosen": -31.27206039428711,
|
| 1013 |
+
"logps/rejected": -32.174869537353516,
|
| 1014 |
+
"loss": 0.2593,
|
| 1015 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1016 |
+
"rewards/chosen": 1.297335147857666,
|
| 1017 |
+
"rewards/margins": 1.3708014488220215,
|
| 1018 |
+
"rewards/rejected": -0.07346634566783905,
|
| 1019 |
+
"step": 620
|
| 1020 |
+
},
|
| 1021 |
+
{
|
| 1022 |
+
"epoch": 1.64,
|
| 1023 |
+
"grad_norm": 6.9375,
|
| 1024 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 1025 |
+
"logits/chosen": -1.967087745666504,
|
| 1026 |
+
"logits/rejected": -1.9582030773162842,
|
| 1027 |
+
"logps/chosen": -30.80006980895996,
|
| 1028 |
+
"logps/rejected": -31.055767059326172,
|
| 1029 |
+
"loss": 0.2573,
|
| 1030 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1031 |
+
"rewards/chosen": 1.3832311630249023,
|
| 1032 |
+
"rewards/margins": 1.4675111770629883,
|
| 1033 |
+
"rewards/rejected": -0.08427979052066803,
|
| 1034 |
+
"step": 630
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 1.66,
|
| 1038 |
+
"grad_norm": 6.21875,
|
| 1039 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1040 |
+
"logits/chosen": -1.9678598642349243,
|
| 1041 |
+
"logits/rejected": -1.9646453857421875,
|
| 1042 |
+
"logps/chosen": -30.486230850219727,
|
| 1043 |
+
"logps/rejected": -32.526588439941406,
|
| 1044 |
+
"loss": 0.2662,
|
| 1045 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1046 |
+
"rewards/chosen": 1.1985342502593994,
|
| 1047 |
+
"rewards/margins": 1.2735540866851807,
|
| 1048 |
+
"rewards/rejected": -0.07501962780952454,
|
| 1049 |
+
"step": 640
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 1.69,
|
| 1053 |
+
"grad_norm": 5.25,
|
| 1054 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1055 |
+
"logits/chosen": -1.9753246307373047,
|
| 1056 |
+
"logits/rejected": -1.9776771068572998,
|
| 1057 |
+
"logps/chosen": -31.804119110107422,
|
| 1058 |
+
"logps/rejected": -33.76448440551758,
|
| 1059 |
+
"loss": 0.2474,
|
| 1060 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1061 |
+
"rewards/chosen": 1.2947121858596802,
|
| 1062 |
+
"rewards/margins": 1.3514697551727295,
|
| 1063 |
+
"rewards/rejected": -0.056757647544145584,
|
| 1064 |
+
"step": 650
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 1.71,
|
| 1068 |
+
"grad_norm": 7.84375,
|
| 1069 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1070 |
+
"logits/chosen": -1.8884271383285522,
|
| 1071 |
+
"logits/rejected": -1.8889974355697632,
|
| 1072 |
+
"logps/chosen": -31.999134063720703,
|
| 1073 |
+
"logps/rejected": -31.639989852905273,
|
| 1074 |
+
"loss": 0.2419,
|
| 1075 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1076 |
+
"rewards/chosen": 1.363389253616333,
|
| 1077 |
+
"rewards/margins": 1.3863054513931274,
|
| 1078 |
+
"rewards/rejected": -0.022916005924344063,
|
| 1079 |
+
"step": 660
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 1.74,
|
| 1083 |
+
"grad_norm": 6.15625,
|
| 1084 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1085 |
+
"logits/chosen": -1.9295670986175537,
|
| 1086 |
+
"logits/rejected": -1.920019507408142,
|
| 1087 |
+
"logps/chosen": -32.041465759277344,
|
| 1088 |
+
"logps/rejected": -32.48603057861328,
|
| 1089 |
+
"loss": 0.2352,
|
| 1090 |
+
"rewards/accuracies": 0.9375,
|
| 1091 |
+
"rewards/chosen": 1.3718717098236084,
|
| 1092 |
+
"rewards/margins": 1.4966375827789307,
|
| 1093 |
+
"rewards/rejected": -0.1247657760977745,
|
| 1094 |
+
"step": 670
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 1.77,
|
| 1098 |
+
"grad_norm": 3.453125,
|
| 1099 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1100 |
+
"logits/chosen": -1.827956199645996,
|
| 1101 |
+
"logits/rejected": -1.8249555826187134,
|
| 1102 |
+
"logps/chosen": -29.395336151123047,
|
| 1103 |
+
"logps/rejected": -34.349639892578125,
|
| 1104 |
+
"loss": 0.2442,
|
| 1105 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1106 |
+
"rewards/chosen": 1.445934534072876,
|
| 1107 |
+
"rewards/margins": 1.4676737785339355,
|
| 1108 |
+
"rewards/rejected": -0.021739140152931213,
|
| 1109 |
+
"step": 680
|
| 1110 |
+
},
|
| 1111 |
+
{
|
| 1112 |
+
"epoch": 1.79,
|
| 1113 |
+
"grad_norm": 4.75,
|
| 1114 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1115 |
+
"logits/chosen": -1.8876937627792358,
|
| 1116 |
+
"logits/rejected": -1.8869459629058838,
|
| 1117 |
+
"logps/chosen": -32.98918151855469,
|
| 1118 |
+
"logps/rejected": -35.476680755615234,
|
| 1119 |
+
"loss": 0.2165,
|
| 1120 |
+
"rewards/accuracies": 0.9375,
|
| 1121 |
+
"rewards/chosen": 1.5967752933502197,
|
| 1122 |
+
"rewards/margins": 1.6827762126922607,
|
| 1123 |
+
"rewards/rejected": -0.08600117266178131,
|
| 1124 |
+
"step": 690
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 1.82,
|
| 1128 |
+
"grad_norm": 4.15625,
|
| 1129 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1130 |
+
"logits/chosen": -1.862781286239624,
|
| 1131 |
+
"logits/rejected": -1.8658040761947632,
|
| 1132 |
+
"logps/chosen": -30.54743003845215,
|
| 1133 |
+
"logps/rejected": -35.61197280883789,
|
| 1134 |
+
"loss": 0.2606,
|
| 1135 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1136 |
+
"rewards/chosen": 1.4387942552566528,
|
| 1137 |
+
"rewards/margins": 1.418005108833313,
|
| 1138 |
+
"rewards/rejected": 0.020789090543985367,
|
| 1139 |
+
"step": 700
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 1.82,
|
| 1143 |
+
"eval_logits/chosen": -2.1835741996765137,
|
| 1144 |
+
"eval_logits/rejected": -2.1788299083709717,
|
| 1145 |
+
"eval_logps/chosen": -33.57905578613281,
|
| 1146 |
+
"eval_logps/rejected": -37.20835876464844,
|
| 1147 |
+
"eval_loss": 0.47641801834106445,
|
| 1148 |
+
"eval_rewards/accuracies": 0.6063122749328613,
|
| 1149 |
+
"eval_rewards/chosen": 0.3188474476337433,
|
| 1150 |
+
"eval_rewards/margins": 0.10306696593761444,
|
| 1151 |
+
"eval_rewards/rejected": 0.21578048169612885,
|
| 1152 |
+
"eval_runtime": 145.3074,
|
| 1153 |
+
"eval_samples_per_second": 2.361,
|
| 1154 |
+
"eval_steps_per_second": 0.296,
|
| 1155 |
+
"step": 700
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 1.84,
|
| 1159 |
+
"grad_norm": 5.8125,
|
| 1160 |
+
"learning_rate": 3.2639145321045933e-06,
|
| 1161 |
+
"logits/chosen": -1.970207929611206,
|
| 1162 |
+
"logits/rejected": -1.972961187362671,
|
| 1163 |
+
"logps/chosen": -32.8425407409668,
|
| 1164 |
+
"logps/rejected": -34.31882858276367,
|
| 1165 |
+
"loss": 0.2562,
|
| 1166 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1167 |
+
"rewards/chosen": 1.4371341466903687,
|
| 1168 |
+
"rewards/margins": 1.471206784248352,
|
| 1169 |
+
"rewards/rejected": -0.034072600305080414,
|
| 1170 |
+
"step": 710
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 1.87,
|
| 1174 |
+
"grad_norm": 6.8125,
|
| 1175 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1176 |
+
"logits/chosen": -1.8238359689712524,
|
| 1177 |
+
"logits/rejected": -1.8179490566253662,
|
| 1178 |
+
"logps/chosen": -32.64445877075195,
|
| 1179 |
+
"logps/rejected": -32.34913635253906,
|
| 1180 |
+
"loss": 0.2397,
|
| 1181 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1182 |
+
"rewards/chosen": 1.5321400165557861,
|
| 1183 |
+
"rewards/margins": 1.530247449874878,
|
| 1184 |
+
"rewards/rejected": 0.0018926591146737337,
|
| 1185 |
+
"step": 720
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.9,
|
| 1189 |
+
"grad_norm": 3.765625,
|
| 1190 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1191 |
+
"logits/chosen": -1.996241569519043,
|
| 1192 |
+
"logits/rejected": -1.9933366775512695,
|
| 1193 |
+
"logps/chosen": -28.67278480529785,
|
| 1194 |
+
"logps/rejected": -31.668869018554688,
|
| 1195 |
+
"loss": 0.2416,
|
| 1196 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1197 |
+
"rewards/chosen": 1.3861420154571533,
|
| 1198 |
+
"rewards/margins": 1.5406121015548706,
|
| 1199 |
+
"rewards/rejected": -0.1544700562953949,
|
| 1200 |
+
"step": 730
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 1.92,
|
| 1204 |
+
"grad_norm": 3.875,
|
| 1205 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1206 |
+
"logits/chosen": -1.8369905948638916,
|
| 1207 |
+
"logits/rejected": -1.8370345830917358,
|
| 1208 |
+
"logps/chosen": -31.223901748657227,
|
| 1209 |
+
"logps/rejected": -37.633975982666016,
|
| 1210 |
+
"loss": 0.2245,
|
| 1211 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1212 |
+
"rewards/chosen": 1.5674915313720703,
|
| 1213 |
+
"rewards/margins": 1.571808099746704,
|
| 1214 |
+
"rewards/rejected": -0.004316648934036493,
|
| 1215 |
+
"step": 740
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"epoch": 1.95,
|
| 1219 |
+
"grad_norm": 3.1875,
|
| 1220 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1221 |
+
"logits/chosen": -1.7186479568481445,
|
| 1222 |
+
"logits/rejected": -1.7238624095916748,
|
| 1223 |
+
"logps/chosen": -34.5988655090332,
|
| 1224 |
+
"logps/rejected": -33.955909729003906,
|
| 1225 |
+
"loss": 0.2323,
|
| 1226 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1227 |
+
"rewards/chosen": 1.6335166692733765,
|
| 1228 |
+
"rewards/margins": 1.6220157146453857,
|
| 1229 |
+
"rewards/rejected": 0.011500892229378223,
|
| 1230 |
+
"step": 750
|
| 1231 |
+
},
|
| 1232 |
+
{
|
| 1233 |
+
"epoch": 1.97,
|
| 1234 |
+
"grad_norm": 5.625,
|
| 1235 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1236 |
+
"logits/chosen": -1.9235740900039673,
|
| 1237 |
+
"logits/rejected": -1.9250673055648804,
|
| 1238 |
+
"logps/chosen": -30.627086639404297,
|
| 1239 |
+
"logps/rejected": -32.82661437988281,
|
| 1240 |
+
"loss": 0.2421,
|
| 1241 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1242 |
+
"rewards/chosen": 1.385018229484558,
|
| 1243 |
+
"rewards/margins": 1.4181984663009644,
|
| 1244 |
+
"rewards/rejected": -0.03318030759692192,
|
| 1245 |
+
"step": 760
|
| 1246 |
+
},
|
| 1247 |
+
{
|
| 1248 |
+
"epoch": 2.0,
|
| 1249 |
+
"grad_norm": 5.34375,
|
| 1250 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1251 |
+
"logits/chosen": -1.894831895828247,
|
| 1252 |
+
"logits/rejected": -1.8939815759658813,
|
| 1253 |
+
"logps/chosen": -30.067148208618164,
|
| 1254 |
+
"logps/rejected": -34.84111785888672,
|
| 1255 |
+
"loss": 0.2551,
|
| 1256 |
+
"rewards/accuracies": 0.85833340883255,
|
| 1257 |
+
"rewards/chosen": 1.4759252071380615,
|
| 1258 |
+
"rewards/margins": 1.3449987173080444,
|
| 1259 |
+
"rewards/rejected": 0.13092657923698425,
|
| 1260 |
+
"step": 770
|
| 1261 |
+
},
|
| 1262 |
+
{
|
| 1263 |
+
"epoch": 2.03,
|
| 1264 |
+
"grad_norm": 3.671875,
|
| 1265 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1266 |
+
"logits/chosen": -1.895990014076233,
|
| 1267 |
+
"logits/rejected": -1.8953578472137451,
|
| 1268 |
+
"logps/chosen": -32.30732727050781,
|
| 1269 |
+
"logps/rejected": -31.775619506835938,
|
| 1270 |
+
"loss": 0.1909,
|
| 1271 |
+
"rewards/accuracies": 0.9375,
|
| 1272 |
+
"rewards/chosen": 1.6387954950332642,
|
| 1273 |
+
"rewards/margins": 1.8880647420883179,
|
| 1274 |
+
"rewards/rejected": -0.24926936626434326,
|
| 1275 |
+
"step": 780
|
| 1276 |
+
},
|
| 1277 |
+
{
|
| 1278 |
+
"epoch": 2.05,
|
| 1279 |
+
"grad_norm": 4.90625,
|
| 1280 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1281 |
+
"logits/chosen": -1.9268324375152588,
|
| 1282 |
+
"logits/rejected": -1.9253270626068115,
|
| 1283 |
+
"logps/chosen": -27.4288387298584,
|
| 1284 |
+
"logps/rejected": -33.15345764160156,
|
| 1285 |
+
"loss": 0.1931,
|
| 1286 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1287 |
+
"rewards/chosen": 1.4940404891967773,
|
| 1288 |
+
"rewards/margins": 1.8560407161712646,
|
| 1289 |
+
"rewards/rejected": -0.36200040578842163,
|
| 1290 |
+
"step": 790
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"epoch": 2.08,
|
| 1294 |
+
"grad_norm": 3.78125,
|
| 1295 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1296 |
+
"logits/chosen": -1.9773414134979248,
|
| 1297 |
+
"logits/rejected": -1.974020004272461,
|
| 1298 |
+
"logps/chosen": -29.959461212158203,
|
| 1299 |
+
"logps/rejected": -34.62144470214844,
|
| 1300 |
+
"loss": 0.1758,
|
| 1301 |
+
"rewards/accuracies": 0.9375,
|
| 1302 |
+
"rewards/chosen": 1.7274243831634521,
|
| 1303 |
+
"rewards/margins": 2.023111581802368,
|
| 1304 |
+
"rewards/rejected": -0.2956870496273041,
|
| 1305 |
+
"step": 800
|
| 1306 |
+
},
|
| 1307 |
+
{
|
| 1308 |
+
"epoch": 2.08,
|
| 1309 |
+
"eval_logits/chosen": -2.172740936279297,
|
| 1310 |
+
"eval_logits/rejected": -2.1680266857147217,
|
| 1311 |
+
"eval_logps/chosen": -33.62887191772461,
|
| 1312 |
+
"eval_logps/rejected": -37.266761779785156,
|
| 1313 |
+
"eval_loss": 0.4767366945743561,
|
| 1314 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1315 |
+
"eval_rewards/chosen": 0.28397703170776367,
|
| 1316 |
+
"eval_rewards/margins": 0.10907654464244843,
|
| 1317 |
+
"eval_rewards/rejected": 0.17490047216415405,
|
| 1318 |
+
"eval_runtime": 145.1754,
|
| 1319 |
+
"eval_samples_per_second": 2.363,
|
| 1320 |
+
"eval_steps_per_second": 0.296,
|
| 1321 |
+
"step": 800
|
| 1322 |
+
},
|
| 1323 |
+
{
|
| 1324 |
+
"epoch": 2.1,
|
| 1325 |
+
"grad_norm": 2.46875,
|
| 1326 |
+
"learning_rate": 2.7094204786572254e-06,
|
| 1327 |
+
"logits/chosen": -1.8107726573944092,
|
| 1328 |
+
"logits/rejected": -1.8030484914779663,
|
| 1329 |
+
"logps/chosen": -31.506906509399414,
|
| 1330 |
+
"logps/rejected": -35.04960250854492,
|
| 1331 |
+
"loss": 0.1452,
|
| 1332 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1333 |
+
"rewards/chosen": 2.0645337104797363,
|
| 1334 |
+
"rewards/margins": 2.407174587249756,
|
| 1335 |
+
"rewards/rejected": -0.3426409363746643,
|
| 1336 |
+
"step": 810
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 2.13,
|
| 1340 |
+
"grad_norm": 4.40625,
|
| 1341 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1342 |
+
"logits/chosen": -1.887786865234375,
|
| 1343 |
+
"logits/rejected": -1.8982326984405518,
|
| 1344 |
+
"logps/chosen": -33.01645278930664,
|
| 1345 |
+
"logps/rejected": -32.74128341674805,
|
| 1346 |
+
"loss": 0.1899,
|
| 1347 |
+
"rewards/accuracies": 0.9375,
|
| 1348 |
+
"rewards/chosen": 1.8754699230194092,
|
| 1349 |
+
"rewards/margins": 2.0923609733581543,
|
| 1350 |
+
"rewards/rejected": -0.2168913334608078,
|
| 1351 |
+
"step": 820
|
| 1352 |
+
},
|
| 1353 |
+
{
|
| 1354 |
+
"epoch": 2.16,
|
| 1355 |
+
"grad_norm": 3.796875,
|
| 1356 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1357 |
+
"logits/chosen": -1.9369271993637085,
|
| 1358 |
+
"logits/rejected": -1.9413642883300781,
|
| 1359 |
+
"logps/chosen": -32.51754379272461,
|
| 1360 |
+
"logps/rejected": -29.556041717529297,
|
| 1361 |
+
"loss": 0.1869,
|
| 1362 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1363 |
+
"rewards/chosen": 1.6696659326553345,
|
| 1364 |
+
"rewards/margins": 1.8711254596710205,
|
| 1365 |
+
"rewards/rejected": -0.20145969092845917,
|
| 1366 |
+
"step": 830
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 2.18,
|
| 1370 |
+
"grad_norm": 4.46875,
|
| 1371 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1372 |
+
"logits/chosen": -1.9355465173721313,
|
| 1373 |
+
"logits/rejected": -1.9433647394180298,
|
| 1374 |
+
"logps/chosen": -32.05677795410156,
|
| 1375 |
+
"logps/rejected": -30.385608673095703,
|
| 1376 |
+
"loss": 0.1604,
|
| 1377 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1378 |
+
"rewards/chosen": 1.911877989768982,
|
| 1379 |
+
"rewards/margins": 2.2884128093719482,
|
| 1380 |
+
"rewards/rejected": -0.3765348792076111,
|
| 1381 |
+
"step": 840
|
| 1382 |
+
},
|
| 1383 |
+
{
|
| 1384 |
+
"epoch": 2.21,
|
| 1385 |
+
"grad_norm": 4.625,
|
| 1386 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1387 |
+
"logits/chosen": -1.866355538368225,
|
| 1388 |
+
"logits/rejected": -1.8570215702056885,
|
| 1389 |
+
"logps/chosen": -29.2946720123291,
|
| 1390 |
+
"logps/rejected": -32.352394104003906,
|
| 1391 |
+
"loss": 0.1771,
|
| 1392 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1393 |
+
"rewards/chosen": 1.7489668130874634,
|
| 1394 |
+
"rewards/margins": 1.9859797954559326,
|
| 1395 |
+
"rewards/rejected": -0.23701317608356476,
|
| 1396 |
+
"step": 850
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 2.23,
|
| 1400 |
+
"grad_norm": 3.6875,
|
| 1401 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1402 |
+
"logits/chosen": -2.0053553581237793,
|
| 1403 |
+
"logits/rejected": -1.9958158731460571,
|
| 1404 |
+
"logps/chosen": -23.93155288696289,
|
| 1405 |
+
"logps/rejected": -30.337451934814453,
|
| 1406 |
+
"loss": 0.1938,
|
| 1407 |
+
"rewards/accuracies": 0.9375,
|
| 1408 |
+
"rewards/chosen": 1.604762077331543,
|
| 1409 |
+
"rewards/margins": 1.9303429126739502,
|
| 1410 |
+
"rewards/rejected": -0.3255811333656311,
|
| 1411 |
+
"step": 860
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.26,
|
| 1415 |
+
"grad_norm": 3.4375,
|
| 1416 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1417 |
+
"logits/chosen": -1.8573694229125977,
|
| 1418 |
+
"logits/rejected": -1.8587795495986938,
|
| 1419 |
+
"logps/chosen": -31.69219398498535,
|
| 1420 |
+
"logps/rejected": -30.3853816986084,
|
| 1421 |
+
"loss": 0.1718,
|
| 1422 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1423 |
+
"rewards/chosen": 1.8485662937164307,
|
| 1424 |
+
"rewards/margins": 2.130580186843872,
|
| 1425 |
+
"rewards/rejected": -0.28201407194137573,
|
| 1426 |
+
"step": 870
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"epoch": 2.29,
|
| 1430 |
+
"grad_norm": 3.1875,
|
| 1431 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1432 |
+
"logits/chosen": -1.8693485260009766,
|
| 1433 |
+
"logits/rejected": -1.8701515197753906,
|
| 1434 |
+
"logps/chosen": -29.962982177734375,
|
| 1435 |
+
"logps/rejected": -33.51511764526367,
|
| 1436 |
+
"loss": 0.1879,
|
| 1437 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1438 |
+
"rewards/chosen": 1.6983861923217773,
|
| 1439 |
+
"rewards/margins": 2.0233235359191895,
|
| 1440 |
+
"rewards/rejected": -0.3249376118183136,
|
| 1441 |
+
"step": 880
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.31,
|
| 1445 |
+
"grad_norm": 3.25,
|
| 1446 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1447 |
+
"logits/chosen": -1.7684412002563477,
|
| 1448 |
+
"logits/rejected": -1.7669271230697632,
|
| 1449 |
+
"logps/chosen": -32.97483825683594,
|
| 1450 |
+
"logps/rejected": -36.623138427734375,
|
| 1451 |
+
"loss": 0.1579,
|
| 1452 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1453 |
+
"rewards/chosen": 1.8830540180206299,
|
| 1454 |
+
"rewards/margins": 2.405186176300049,
|
| 1455 |
+
"rewards/rejected": -0.522132396697998,
|
| 1456 |
+
"step": 890
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"epoch": 2.34,
|
| 1460 |
+
"grad_norm": 3.296875,
|
| 1461 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1462 |
+
"logits/chosen": -1.9180320501327515,
|
| 1463 |
+
"logits/rejected": -1.9112794399261475,
|
| 1464 |
+
"logps/chosen": -29.328502655029297,
|
| 1465 |
+
"logps/rejected": -33.22010803222656,
|
| 1466 |
+
"loss": 0.1687,
|
| 1467 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1468 |
+
"rewards/chosen": 1.8718944787979126,
|
| 1469 |
+
"rewards/margins": 2.1953208446502686,
|
| 1470 |
+
"rewards/rejected": -0.3234265446662903,
|
| 1471 |
+
"step": 900
|
| 1472 |
+
},
|
| 1473 |
+
{
|
| 1474 |
+
"epoch": 2.34,
|
| 1475 |
+
"eval_logits/chosen": -2.1673672199249268,
|
| 1476 |
+
"eval_logits/rejected": -2.1626341342926025,
|
| 1477 |
+
"eval_logps/chosen": -33.620479583740234,
|
| 1478 |
+
"eval_logps/rejected": -37.25470733642578,
|
| 1479 |
+
"eval_loss": 0.4770047068595886,
|
| 1480 |
+
"eval_rewards/accuracies": 0.5485880374908447,
|
| 1481 |
+
"eval_rewards/chosen": 0.2898494601249695,
|
| 1482 |
+
"eval_rewards/margins": 0.10651160031557083,
|
| 1483 |
+
"eval_rewards/rejected": 0.18333788216114044,
|
| 1484 |
+
"eval_runtime": 145.3634,
|
| 1485 |
+
"eval_samples_per_second": 2.36,
|
| 1486 |
+
"eval_steps_per_second": 0.296,
|
| 1487 |
+
"step": 900
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 2.36,
|
| 1491 |
+
"grad_norm": 4.0625,
|
| 1492 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 1493 |
+
"logits/chosen": -1.9585449695587158,
|
| 1494 |
+
"logits/rejected": -1.953966498374939,
|
| 1495 |
+
"logps/chosen": -31.010263442993164,
|
| 1496 |
+
"logps/rejected": -35.009552001953125,
|
| 1497 |
+
"loss": 0.164,
|
| 1498 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1499 |
+
"rewards/chosen": 1.783191442489624,
|
| 1500 |
+
"rewards/margins": 2.119842052459717,
|
| 1501 |
+
"rewards/rejected": -0.3366504907608032,
|
| 1502 |
+
"step": 910
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 2.39,
|
| 1506 |
+
"grad_norm": 3.453125,
|
| 1507 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1508 |
+
"logits/chosen": -1.8792098760604858,
|
| 1509 |
+
"logits/rejected": -1.8842833042144775,
|
| 1510 |
+
"logps/chosen": -32.50172805786133,
|
| 1511 |
+
"logps/rejected": -32.84755325317383,
|
| 1512 |
+
"loss": 0.1644,
|
| 1513 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1514 |
+
"rewards/chosen": 1.9927990436553955,
|
| 1515 |
+
"rewards/margins": 2.236151933670044,
|
| 1516 |
+
"rewards/rejected": -0.24335305392742157,
|
| 1517 |
+
"step": 920
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 2.42,
|
| 1521 |
+
"grad_norm": 4.15625,
|
| 1522 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1523 |
+
"logits/chosen": -1.9770915508270264,
|
| 1524 |
+
"logits/rejected": -1.9679561853408813,
|
| 1525 |
+
"logps/chosen": -29.701152801513672,
|
| 1526 |
+
"logps/rejected": -34.7978630065918,
|
| 1527 |
+
"loss": 0.1669,
|
| 1528 |
+
"rewards/accuracies": 1.0,
|
| 1529 |
+
"rewards/chosen": 1.8086459636688232,
|
| 1530 |
+
"rewards/margins": 2.0571932792663574,
|
| 1531 |
+
"rewards/rejected": -0.2485475093126297,
|
| 1532 |
+
"step": 930
|
| 1533 |
+
},
|
| 1534 |
+
{
|
| 1535 |
+
"epoch": 2.44,
|
| 1536 |
+
"grad_norm": 3.78125,
|
| 1537 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1538 |
+
"logits/chosen": -1.903071641921997,
|
| 1539 |
+
"logits/rejected": -1.8999197483062744,
|
| 1540 |
+
"logps/chosen": -29.023662567138672,
|
| 1541 |
+
"logps/rejected": -32.71151351928711,
|
| 1542 |
+
"loss": 0.2019,
|
| 1543 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1544 |
+
"rewards/chosen": 1.585354208946228,
|
| 1545 |
+
"rewards/margins": 1.757653832435608,
|
| 1546 |
+
"rewards/rejected": -0.17229948937892914,
|
| 1547 |
+
"step": 940
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 2.47,
|
| 1551 |
+
"grad_norm": 3.4375,
|
| 1552 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1553 |
+
"logits/chosen": -1.842889428138733,
|
| 1554 |
+
"logits/rejected": -1.8503259420394897,
|
| 1555 |
+
"logps/chosen": -30.36769676208496,
|
| 1556 |
+
"logps/rejected": -35.7653923034668,
|
| 1557 |
+
"loss": 0.2098,
|
| 1558 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1559 |
+
"rewards/chosen": 1.7523466348648071,
|
| 1560 |
+
"rewards/margins": 1.903167963027954,
|
| 1561 |
+
"rewards/rejected": -0.15082135796546936,
|
| 1562 |
+
"step": 950
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 2.49,
|
| 1566 |
+
"grad_norm": 4.15625,
|
| 1567 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1568 |
+
"logits/chosen": -1.8521137237548828,
|
| 1569 |
+
"logits/rejected": -1.8454768657684326,
|
| 1570 |
+
"logps/chosen": -31.145715713500977,
|
| 1571 |
+
"logps/rejected": -37.1761589050293,
|
| 1572 |
+
"loss": 0.1613,
|
| 1573 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1574 |
+
"rewards/chosen": 1.877960205078125,
|
| 1575 |
+
"rewards/margins": 2.282109260559082,
|
| 1576 |
+
"rewards/rejected": -0.4041489064693451,
|
| 1577 |
+
"step": 960
|
| 1578 |
+
},
|
| 1579 |
+
{
|
| 1580 |
+
"epoch": 2.52,
|
| 1581 |
+
"grad_norm": 4.1875,
|
| 1582 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1583 |
+
"logits/chosen": -1.8534562587738037,
|
| 1584 |
+
"logits/rejected": -1.8540384769439697,
|
| 1585 |
+
"logps/chosen": -28.50092124938965,
|
| 1586 |
+
"logps/rejected": -33.28984832763672,
|
| 1587 |
+
"loss": 0.1593,
|
| 1588 |
+
"rewards/accuracies": 0.9375,
|
| 1589 |
+
"rewards/chosen": 1.9148372411727905,
|
| 1590 |
+
"rewards/margins": 2.1308465003967285,
|
| 1591 |
+
"rewards/rejected": -0.2160094678401947,
|
| 1592 |
+
"step": 970
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 2.55,
|
| 1596 |
+
"grad_norm": 4.375,
|
| 1597 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1598 |
+
"logits/chosen": -1.8584295511245728,
|
| 1599 |
+
"logits/rejected": -1.8696168661117554,
|
| 1600 |
+
"logps/chosen": -30.254512786865234,
|
| 1601 |
+
"logps/rejected": -34.45421600341797,
|
| 1602 |
+
"loss": 0.1653,
|
| 1603 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1604 |
+
"rewards/chosen": 1.97552490234375,
|
| 1605 |
+
"rewards/margins": 2.201627731323242,
|
| 1606 |
+
"rewards/rejected": -0.22610275447368622,
|
| 1607 |
+
"step": 980
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 2.57,
|
| 1611 |
+
"grad_norm": 2.828125,
|
| 1612 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1613 |
+
"logits/chosen": -1.926839828491211,
|
| 1614 |
+
"logits/rejected": -1.924412488937378,
|
| 1615 |
+
"logps/chosen": -33.426719665527344,
|
| 1616 |
+
"logps/rejected": -32.59941101074219,
|
| 1617 |
+
"loss": 0.1853,
|
| 1618 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1619 |
+
"rewards/chosen": 1.8860301971435547,
|
| 1620 |
+
"rewards/margins": 2.0261569023132324,
|
| 1621 |
+
"rewards/rejected": -0.14012634754180908,
|
| 1622 |
+
"step": 990
|
| 1623 |
+
},
|
| 1624 |
+
{
|
| 1625 |
+
"epoch": 2.6,
|
| 1626 |
+
"grad_norm": 4.125,
|
| 1627 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1628 |
+
"logits/chosen": -1.8978465795516968,
|
| 1629 |
+
"logits/rejected": -1.9018335342407227,
|
| 1630 |
+
"logps/chosen": -31.86910629272461,
|
| 1631 |
+
"logps/rejected": -34.74235153198242,
|
| 1632 |
+
"loss": 0.1826,
|
| 1633 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1634 |
+
"rewards/chosen": 1.8159281015396118,
|
| 1635 |
+
"rewards/margins": 1.9953454732894897,
|
| 1636 |
+
"rewards/rejected": -0.17941740155220032,
|
| 1637 |
+
"step": 1000
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 2.6,
|
| 1641 |
+
"eval_logits/chosen": -2.162538528442383,
|
| 1642 |
+
"eval_logits/rejected": -2.157820701599121,
|
| 1643 |
+
"eval_logps/chosen": -33.64886474609375,
|
| 1644 |
+
"eval_logps/rejected": -37.291744232177734,
|
| 1645 |
+
"eval_loss": 0.4763668477535248,
|
| 1646 |
+
"eval_rewards/accuracies": 0.5830564498901367,
|
| 1647 |
+
"eval_rewards/chosen": 0.26998022198677063,
|
| 1648 |
+
"eval_rewards/margins": 0.11256679147481918,
|
| 1649 |
+
"eval_rewards/rejected": 0.15741342306137085,
|
| 1650 |
+
"eval_runtime": 145.331,
|
| 1651 |
+
"eval_samples_per_second": 2.36,
|
| 1652 |
+
"eval_steps_per_second": 0.296,
|
| 1653 |
+
"step": 1000
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.62,
|
| 1657 |
+
"grad_norm": 5.9375,
|
| 1658 |
+
"learning_rate": 1.5972066659083796e-06,
|
| 1659 |
+
"logits/chosen": -1.9077873229980469,
|
| 1660 |
+
"logits/rejected": -1.9126176834106445,
|
| 1661 |
+
"logps/chosen": -30.2926082611084,
|
| 1662 |
+
"logps/rejected": -32.68244171142578,
|
| 1663 |
+
"loss": 0.2024,
|
| 1664 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1665 |
+
"rewards/chosen": 1.7554700374603271,
|
| 1666 |
+
"rewards/margins": 1.8682985305786133,
|
| 1667 |
+
"rewards/rejected": -0.11282841116189957,
|
| 1668 |
+
"step": 1010
|
| 1669 |
+
},
|
| 1670 |
+
{
|
| 1671 |
+
"epoch": 2.65,
|
| 1672 |
+
"grad_norm": 3.734375,
|
| 1673 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1674 |
+
"logits/chosen": -1.8675119876861572,
|
| 1675 |
+
"logits/rejected": -1.8702566623687744,
|
| 1676 |
+
"logps/chosen": -27.929931640625,
|
| 1677 |
+
"logps/rejected": -32.15419006347656,
|
| 1678 |
+
"loss": 0.1718,
|
| 1679 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1680 |
+
"rewards/chosen": 1.7609975337982178,
|
| 1681 |
+
"rewards/margins": 2.079514503479004,
|
| 1682 |
+
"rewards/rejected": -0.31851673126220703,
|
| 1683 |
+
"step": 1020
|
| 1684 |
+
},
|
| 1685 |
+
{
|
| 1686 |
+
"epoch": 2.68,
|
| 1687 |
+
"grad_norm": 2.671875,
|
| 1688 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1689 |
+
"logits/chosen": -1.9224603176116943,
|
| 1690 |
+
"logits/rejected": -1.916998267173767,
|
| 1691 |
+
"logps/chosen": -31.501073837280273,
|
| 1692 |
+
"logps/rejected": -33.88199996948242,
|
| 1693 |
+
"loss": 0.1823,
|
| 1694 |
+
"rewards/accuracies": 0.9375,
|
| 1695 |
+
"rewards/chosen": 1.7416280508041382,
|
| 1696 |
+
"rewards/margins": 1.9964303970336914,
|
| 1697 |
+
"rewards/rejected": -0.2548021972179413,
|
| 1698 |
+
"step": 1030
|
| 1699 |
+
},
|
| 1700 |
+
{
|
| 1701 |
+
"epoch": 2.7,
|
| 1702 |
+
"grad_norm": 4.71875,
|
| 1703 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1704 |
+
"logits/chosen": -1.8555800914764404,
|
| 1705 |
+
"logits/rejected": -1.8593838214874268,
|
| 1706 |
+
"logps/chosen": -33.17572021484375,
|
| 1707 |
+
"logps/rejected": -34.55617141723633,
|
| 1708 |
+
"loss": 0.19,
|
| 1709 |
+
"rewards/accuracies": 0.9375,
|
| 1710 |
+
"rewards/chosen": 1.8046057224273682,
|
| 1711 |
+
"rewards/margins": 1.9002208709716797,
|
| 1712 |
+
"rewards/rejected": -0.09561540186405182,
|
| 1713 |
+
"step": 1040
|
| 1714 |
+
},
|
| 1715 |
+
{
|
| 1716 |
+
"epoch": 2.73,
|
| 1717 |
+
"grad_norm": 3.5,
|
| 1718 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1719 |
+
"logits/chosen": -1.8540366888046265,
|
| 1720 |
+
"logits/rejected": -1.866837501525879,
|
| 1721 |
+
"logps/chosen": -29.41763687133789,
|
| 1722 |
+
"logps/rejected": -33.34986114501953,
|
| 1723 |
+
"loss": 0.1906,
|
| 1724 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1725 |
+
"rewards/chosen": 1.711678147315979,
|
| 1726 |
+
"rewards/margins": 1.9425528049468994,
|
| 1727 |
+
"rewards/rejected": -0.23087480664253235,
|
| 1728 |
+
"step": 1050
|
| 1729 |
+
},
|
| 1730 |
+
{
|
| 1731 |
+
"epoch": 2.75,
|
| 1732 |
+
"grad_norm": 4.0625,
|
| 1733 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1734 |
+
"logits/chosen": -1.8432916402816772,
|
| 1735 |
+
"logits/rejected": -1.8369777202606201,
|
| 1736 |
+
"logps/chosen": -29.536724090576172,
|
| 1737 |
+
"logps/rejected": -33.66853713989258,
|
| 1738 |
+
"loss": 0.1732,
|
| 1739 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1740 |
+
"rewards/chosen": 1.835822343826294,
|
| 1741 |
+
"rewards/margins": 2.0812220573425293,
|
| 1742 |
+
"rewards/rejected": -0.24539966881275177,
|
| 1743 |
+
"step": 1060
|
| 1744 |
+
},
|
| 1745 |
+
{
|
| 1746 |
+
"epoch": 2.78,
|
| 1747 |
+
"grad_norm": 3.59375,
|
| 1748 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1749 |
+
"logits/chosen": -1.9685659408569336,
|
| 1750 |
+
"logits/rejected": -1.9693739414215088,
|
| 1751 |
+
"logps/chosen": -30.985530853271484,
|
| 1752 |
+
"logps/rejected": -33.102413177490234,
|
| 1753 |
+
"loss": 0.1709,
|
| 1754 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1755 |
+
"rewards/chosen": 1.7930816411972046,
|
| 1756 |
+
"rewards/margins": 1.9734855890274048,
|
| 1757 |
+
"rewards/rejected": -0.180403932929039,
|
| 1758 |
+
"step": 1070
|
| 1759 |
+
},
|
| 1760 |
+
{
|
| 1761 |
+
"epoch": 2.81,
|
| 1762 |
+
"grad_norm": 5.28125,
|
| 1763 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1764 |
+
"logits/chosen": -1.9013326168060303,
|
| 1765 |
+
"logits/rejected": -1.9002430438995361,
|
| 1766 |
+
"logps/chosen": -31.509572982788086,
|
| 1767 |
+
"logps/rejected": -33.25670623779297,
|
| 1768 |
+
"loss": 0.2033,
|
| 1769 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1770 |
+
"rewards/chosen": 1.8061821460723877,
|
| 1771 |
+
"rewards/margins": 1.8515968322753906,
|
| 1772 |
+
"rewards/rejected": -0.045414675027132034,
|
| 1773 |
+
"step": 1080
|
| 1774 |
+
},
|
| 1775 |
+
{
|
| 1776 |
+
"epoch": 2.83,
|
| 1777 |
+
"grad_norm": 3.234375,
|
| 1778 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1779 |
+
"logits/chosen": -1.9260213375091553,
|
| 1780 |
+
"logits/rejected": -1.918617606163025,
|
| 1781 |
+
"logps/chosen": -31.691287994384766,
|
| 1782 |
+
"logps/rejected": -31.300344467163086,
|
| 1783 |
+
"loss": 0.1849,
|
| 1784 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1785 |
+
"rewards/chosen": 1.774212121963501,
|
| 1786 |
+
"rewards/margins": 1.955381155014038,
|
| 1787 |
+
"rewards/rejected": -0.1811690330505371,
|
| 1788 |
+
"step": 1090
|
| 1789 |
+
},
|
| 1790 |
+
{
|
| 1791 |
+
"epoch": 2.86,
|
| 1792 |
+
"grad_norm": 2.796875,
|
| 1793 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1794 |
+
"logits/chosen": -1.896405577659607,
|
| 1795 |
+
"logits/rejected": -1.8956992626190186,
|
| 1796 |
+
"logps/chosen": -32.56309509277344,
|
| 1797 |
+
"logps/rejected": -36.86756134033203,
|
| 1798 |
+
"loss": 0.1541,
|
| 1799 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1800 |
+
"rewards/chosen": 1.9987242221832275,
|
| 1801 |
+
"rewards/margins": 2.359121322631836,
|
| 1802 |
+
"rewards/rejected": -0.3603973388671875,
|
| 1803 |
+
"step": 1100
|
| 1804 |
+
},
|
| 1805 |
+
{
|
| 1806 |
+
"epoch": 2.86,
|
| 1807 |
+
"eval_logits/chosen": -2.1608357429504395,
|
| 1808 |
+
"eval_logits/rejected": -2.1561272144317627,
|
| 1809 |
+
"eval_logps/chosen": -33.625431060791016,
|
| 1810 |
+
"eval_logps/rejected": -37.27484893798828,
|
| 1811 |
+
"eval_loss": 0.47510504722595215,
|
| 1812 |
+
"eval_rewards/accuracies": 0.5776578187942505,
|
| 1813 |
+
"eval_rewards/chosen": 0.28638461232185364,
|
| 1814 |
+
"eval_rewards/margins": 0.11714447289705276,
|
| 1815 |
+
"eval_rewards/rejected": 0.16924014687538147,
|
| 1816 |
+
"eval_runtime": 145.2146,
|
| 1817 |
+
"eval_samples_per_second": 2.362,
|
| 1818 |
+
"eval_steps_per_second": 0.296,
|
| 1819 |
+
"step": 1100
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"epoch": 2.88,
|
| 1823 |
+
"grad_norm": 3.734375,
|
| 1824 |
+
"learning_rate": 1.0963854773524548e-06,
|
| 1825 |
+
"logits/chosen": -1.9410547018051147,
|
| 1826 |
+
"logits/rejected": -1.946345567703247,
|
| 1827 |
+
"logps/chosen": -32.88351821899414,
|
| 1828 |
+
"logps/rejected": -35.84196472167969,
|
| 1829 |
+
"loss": 0.1916,
|
| 1830 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1831 |
+
"rewards/chosen": 1.8085190057754517,
|
| 1832 |
+
"rewards/margins": 2.0173850059509277,
|
| 1833 |
+
"rewards/rejected": -0.2088661640882492,
|
| 1834 |
+
"step": 1110
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"epoch": 2.91,
|
| 1838 |
+
"grad_norm": 2.40625,
|
| 1839 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1840 |
+
"logits/chosen": -1.914781928062439,
|
| 1841 |
+
"logits/rejected": -1.912883996963501,
|
| 1842 |
+
"logps/chosen": -30.12225341796875,
|
| 1843 |
+
"logps/rejected": -33.4071044921875,
|
| 1844 |
+
"loss": 0.1829,
|
| 1845 |
+
"rewards/accuracies": 0.9375,
|
| 1846 |
+
"rewards/chosen": 1.74576735496521,
|
| 1847 |
+
"rewards/margins": 2.0428128242492676,
|
| 1848 |
+
"rewards/rejected": -0.29704537987709045,
|
| 1849 |
+
"step": 1120
|
| 1850 |
+
},
|
| 1851 |
+
{
|
| 1852 |
+
"epoch": 2.94,
|
| 1853 |
+
"grad_norm": 2.828125,
|
| 1854 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1855 |
+
"logits/chosen": -1.8179181814193726,
|
| 1856 |
+
"logits/rejected": -1.8206218481063843,
|
| 1857 |
+
"logps/chosen": -27.161823272705078,
|
| 1858 |
+
"logps/rejected": -30.888957977294922,
|
| 1859 |
+
"loss": 0.1529,
|
| 1860 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1861 |
+
"rewards/chosen": 1.8107696771621704,
|
| 1862 |
+
"rewards/margins": 2.2615151405334473,
|
| 1863 |
+
"rewards/rejected": -0.4507456421852112,
|
| 1864 |
+
"step": 1130
|
| 1865 |
+
},
|
| 1866 |
+
{
|
| 1867 |
+
"epoch": 2.96,
|
| 1868 |
+
"grad_norm": 3.171875,
|
| 1869 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1870 |
+
"logits/chosen": -1.916400671005249,
|
| 1871 |
+
"logits/rejected": -1.9134107828140259,
|
| 1872 |
+
"logps/chosen": -30.4929141998291,
|
| 1873 |
+
"logps/rejected": -32.6131477355957,
|
| 1874 |
+
"loss": 0.178,
|
| 1875 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1876 |
+
"rewards/chosen": 1.897585153579712,
|
| 1877 |
+
"rewards/margins": 1.957972764968872,
|
| 1878 |
+
"rewards/rejected": -0.060387589037418365,
|
| 1879 |
+
"step": 1140
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 2.99,
|
| 1883 |
+
"grad_norm": 4.5,
|
| 1884 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1885 |
+
"logits/chosen": -1.8358739614486694,
|
| 1886 |
+
"logits/rejected": -1.8296668529510498,
|
| 1887 |
+
"logps/chosen": -25.861547470092773,
|
| 1888 |
+
"logps/rejected": -32.6159553527832,
|
| 1889 |
+
"loss": 0.2002,
|
| 1890 |
+
"rewards/accuracies": 0.9375,
|
| 1891 |
+
"rewards/chosen": 1.5935077667236328,
|
| 1892 |
+
"rewards/margins": 1.9208663702011108,
|
| 1893 |
+
"rewards/rejected": -0.32735878229141235,
|
| 1894 |
+
"step": 1150
|
| 1895 |
+
},
|
| 1896 |
+
{
|
| 1897 |
+
"epoch": 3.01,
|
| 1898 |
+
"grad_norm": 2.78125,
|
| 1899 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1900 |
+
"logits/chosen": -1.9491764307022095,
|
| 1901 |
+
"logits/rejected": -1.9467414617538452,
|
| 1902 |
+
"logps/chosen": -30.44895362854004,
|
| 1903 |
+
"logps/rejected": -35.62247848510742,
|
| 1904 |
+
"loss": 0.1676,
|
| 1905 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1906 |
+
"rewards/chosen": 1.852980375289917,
|
| 1907 |
+
"rewards/margins": 2.224640369415283,
|
| 1908 |
+
"rewards/rejected": -0.3716599941253662,
|
| 1909 |
+
"step": 1160
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.04,
|
| 1913 |
+
"grad_norm": 2.578125,
|
| 1914 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1915 |
+
"logits/chosen": -1.932654619216919,
|
| 1916 |
+
"logits/rejected": -1.930672287940979,
|
| 1917 |
+
"logps/chosen": -28.86700439453125,
|
| 1918 |
+
"logps/rejected": -32.90315246582031,
|
| 1919 |
+
"loss": 0.1532,
|
| 1920 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1921 |
+
"rewards/chosen": 1.9231812953948975,
|
| 1922 |
+
"rewards/margins": 2.300574779510498,
|
| 1923 |
+
"rewards/rejected": -0.37739327549934387,
|
| 1924 |
+
"step": 1170
|
| 1925 |
+
},
|
| 1926 |
+
{
|
| 1927 |
+
"epoch": 3.06,
|
| 1928 |
+
"grad_norm": 3.125,
|
| 1929 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1930 |
+
"logits/chosen": -1.9274743795394897,
|
| 1931 |
+
"logits/rejected": -1.9154332876205444,
|
| 1932 |
+
"logps/chosen": -33.33545684814453,
|
| 1933 |
+
"logps/rejected": -34.530860900878906,
|
| 1934 |
+
"loss": 0.1634,
|
| 1935 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1936 |
+
"rewards/chosen": 1.8968908786773682,
|
| 1937 |
+
"rewards/margins": 2.218459367752075,
|
| 1938 |
+
"rewards/rejected": -0.3215683400630951,
|
| 1939 |
+
"step": 1180
|
| 1940 |
+
},
|
| 1941 |
+
{
|
| 1942 |
+
"epoch": 3.09,
|
| 1943 |
+
"grad_norm": 4.5,
|
| 1944 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1945 |
+
"logits/chosen": -1.819190263748169,
|
| 1946 |
+
"logits/rejected": -1.8140865564346313,
|
| 1947 |
+
"logps/chosen": -32.35272979736328,
|
| 1948 |
+
"logps/rejected": -34.45198440551758,
|
| 1949 |
+
"loss": 0.1541,
|
| 1950 |
+
"rewards/accuracies": 0.9375,
|
| 1951 |
+
"rewards/chosen": 2.097872734069824,
|
| 1952 |
+
"rewards/margins": 2.247788906097412,
|
| 1953 |
+
"rewards/rejected": -0.14991624653339386,
|
| 1954 |
+
"step": 1190
|
| 1955 |
+
},
|
| 1956 |
+
{
|
| 1957 |
+
"epoch": 3.12,
|
| 1958 |
+
"grad_norm": 3.046875,
|
| 1959 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1960 |
+
"logits/chosen": -1.8649425506591797,
|
| 1961 |
+
"logits/rejected": -1.8679530620574951,
|
| 1962 |
+
"logps/chosen": -30.119836807250977,
|
| 1963 |
+
"logps/rejected": -31.720102310180664,
|
| 1964 |
+
"loss": 0.194,
|
| 1965 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1966 |
+
"rewards/chosen": 1.746681809425354,
|
| 1967 |
+
"rewards/margins": 1.91042959690094,
|
| 1968 |
+
"rewards/rejected": -0.163747638463974,
|
| 1969 |
+
"step": 1200
|
| 1970 |
+
},
|
| 1971 |
+
{
|
| 1972 |
+
"epoch": 3.12,
|
| 1973 |
+
"eval_logits/chosen": -2.1612026691436768,
|
| 1974 |
+
"eval_logits/rejected": -2.156489610671997,
|
| 1975 |
+
"eval_logps/chosen": -33.626529693603516,
|
| 1976 |
+
"eval_logps/rejected": -37.28034210205078,
|
| 1977 |
+
"eval_loss": 0.47478199005126953,
|
| 1978 |
+
"eval_rewards/accuracies": 0.5801494717597961,
|
| 1979 |
+
"eval_rewards/chosen": 0.28561660647392273,
|
| 1980 |
+
"eval_rewards/margins": 0.12022235989570618,
|
| 1981 |
+
"eval_rewards/rejected": 0.16539424657821655,
|
| 1982 |
+
"eval_runtime": 145.3188,
|
| 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": 3.8125,
|
| 1990 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1991 |
+
"logits/chosen": -1.8913599252700806,
|
| 1992 |
+
"logits/rejected": -1.8870538473129272,
|
| 1993 |
+
"logps/chosen": -28.591426849365234,
|
| 1994 |
+
"logps/rejected": -31.342187881469727,
|
| 1995 |
+
"loss": 0.1557,
|
| 1996 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1997 |
+
"rewards/chosen": 1.8775701522827148,
|
| 1998 |
+
"rewards/margins": 2.181436538696289,
|
| 1999 |
+
"rewards/rejected": -0.30386656522750854,
|
| 2000 |
+
"step": 1210
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 3.17,
|
| 2004 |
+
"grad_norm": 3.296875,
|
| 2005 |
+
"learning_rate": 6.293013489185315e-07,
|
| 2006 |
+
"logits/chosen": -1.9030640125274658,
|
| 2007 |
+
"logits/rejected": -1.897794485092163,
|
| 2008 |
+
"logps/chosen": -32.43680953979492,
|
| 2009 |
+
"logps/rejected": -35.05671691894531,
|
| 2010 |
+
"loss": 0.1566,
|
| 2011 |
+
"rewards/accuracies": 0.9375,
|
| 2012 |
+
"rewards/chosen": 1.9859931468963623,
|
| 2013 |
+
"rewards/margins": 2.3267762660980225,
|
| 2014 |
+
"rewards/rejected": -0.3407830595970154,
|
| 2015 |
+
"step": 1220
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 3.19,
|
| 2019 |
+
"grad_norm": 2.828125,
|
| 2020 |
+
"learning_rate": 5.921934551632086e-07,
|
| 2021 |
+
"logits/chosen": -1.9055864810943604,
|
| 2022 |
+
"logits/rejected": -1.892482042312622,
|
| 2023 |
+
"logps/chosen": -30.573678970336914,
|
| 2024 |
+
"logps/rejected": -34.583587646484375,
|
| 2025 |
+
"loss": 0.1759,
|
| 2026 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2027 |
+
"rewards/chosen": 1.7642362117767334,
|
| 2028 |
+
"rewards/margins": 2.0888543128967285,
|
| 2029 |
+
"rewards/rejected": -0.32461827993392944,
|
| 2030 |
+
"step": 1230
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 3.22,
|
| 2034 |
+
"grad_norm": 2.5,
|
| 2035 |
+
"learning_rate": 5.560657034652405e-07,
|
| 2036 |
+
"logits/chosen": -1.9504817724227905,
|
| 2037 |
+
"logits/rejected": -1.9479200839996338,
|
| 2038 |
+
"logps/chosen": -32.358009338378906,
|
| 2039 |
+
"logps/rejected": -32.21631622314453,
|
| 2040 |
+
"loss": 0.168,
|
| 2041 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2042 |
+
"rewards/chosen": 1.8724260330200195,
|
| 2043 |
+
"rewards/margins": 2.13594126701355,
|
| 2044 |
+
"rewards/rejected": -0.2635149657726288,
|
| 2045 |
+
"step": 1240
|
| 2046 |
+
},
|
| 2047 |
+
{
|
| 2048 |
+
"epoch": 3.25,
|
| 2049 |
+
"grad_norm": 2.421875,
|
| 2050 |
+
"learning_rate": 5.2093665457911e-07,
|
| 2051 |
+
"logits/chosen": -1.8791261911392212,
|
| 2052 |
+
"logits/rejected": -1.8764369487762451,
|
| 2053 |
+
"logps/chosen": -31.556161880493164,
|
| 2054 |
+
"logps/rejected": -35.48642349243164,
|
| 2055 |
+
"loss": 0.1379,
|
| 2056 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2057 |
+
"rewards/chosen": 1.987993597984314,
|
| 2058 |
+
"rewards/margins": 2.4705002307891846,
|
| 2059 |
+
"rewards/rejected": -0.48250651359558105,
|
| 2060 |
+
"step": 1250
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 3.27,
|
| 2064 |
+
"grad_norm": 2.640625,
|
| 2065 |
+
"learning_rate": 4.868243561723535e-07,
|
| 2066 |
+
"logits/chosen": -1.9446487426757812,
|
| 2067 |
+
"logits/rejected": -1.9393678903579712,
|
| 2068 |
+
"logps/chosen": -28.063180923461914,
|
| 2069 |
+
"logps/rejected": -32.76237487792969,
|
| 2070 |
+
"loss": 0.1586,
|
| 2071 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2072 |
+
"rewards/chosen": 1.8883317708969116,
|
| 2073 |
+
"rewards/margins": 2.1542108058929443,
|
| 2074 |
+
"rewards/rejected": -0.2658790349960327,
|
| 2075 |
+
"step": 1260
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 3.3,
|
| 2079 |
+
"grad_norm": 3.375,
|
| 2080 |
+
"learning_rate": 4.537463335535161e-07,
|
| 2081 |
+
"logits/chosen": -1.9708101749420166,
|
| 2082 |
+
"logits/rejected": -1.975584626197815,
|
| 2083 |
+
"logps/chosen": -29.93453025817871,
|
| 2084 |
+
"logps/rejected": -32.06245422363281,
|
| 2085 |
+
"loss": 0.1628,
|
| 2086 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2087 |
+
"rewards/chosen": 1.9435287714004517,
|
| 2088 |
+
"rewards/margins": 2.210292339324951,
|
| 2089 |
+
"rewards/rejected": -0.2667638957500458,
|
| 2090 |
+
"step": 1270
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"epoch": 3.32,
|
| 2094 |
+
"grad_norm": 4.78125,
|
| 2095 |
+
"learning_rate": 4.217195806684629e-07,
|
| 2096 |
+
"logits/chosen": -1.8651002645492554,
|
| 2097 |
+
"logits/rejected": -1.872433066368103,
|
| 2098 |
+
"logps/chosen": -32.172630310058594,
|
| 2099 |
+
"logps/rejected": -33.01711654663086,
|
| 2100 |
+
"loss": 0.1714,
|
| 2101 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2102 |
+
"rewards/chosen": 1.914229154586792,
|
| 2103 |
+
"rewards/margins": 2.2074227333068848,
|
| 2104 |
+
"rewards/rejected": -0.29319363832473755,
|
| 2105 |
+
"step": 1280
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 3.35,
|
| 2109 |
+
"grad_norm": 3.84375,
|
| 2110 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2111 |
+
"logits/chosen": -1.7700433731079102,
|
| 2112 |
+
"logits/rejected": -1.7724153995513916,
|
| 2113 |
+
"logps/chosen": -30.311264038085938,
|
| 2114 |
+
"logps/rejected": -36.89826202392578,
|
| 2115 |
+
"loss": 0.1748,
|
| 2116 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2117 |
+
"rewards/chosen": 1.9744415283203125,
|
| 2118 |
+
"rewards/margins": 2.2152342796325684,
|
| 2119 |
+
"rewards/rejected": -0.2407928705215454,
|
| 2120 |
+
"step": 1290
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 3.38,
|
| 2124 |
+
"grad_norm": 1.546875,
|
| 2125 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2126 |
+
"logits/chosen": -1.8248789310455322,
|
| 2127 |
+
"logits/rejected": -1.82831609249115,
|
| 2128 |
+
"logps/chosen": -30.20694923400879,
|
| 2129 |
+
"logps/rejected": -33.00934600830078,
|
| 2130 |
+
"loss": 0.1414,
|
| 2131 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2132 |
+
"rewards/chosen": 2.1068058013916016,
|
| 2133 |
+
"rewards/margins": 2.488433361053467,
|
| 2134 |
+
"rewards/rejected": -0.38162797689437866,
|
| 2135 |
+
"step": 1300
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.38,
|
| 2139 |
+
"eval_logits/chosen": -2.160539150238037,
|
| 2140 |
+
"eval_logits/rejected": -2.1558327674865723,
|
| 2141 |
+
"eval_logps/chosen": -33.62611389160156,
|
| 2142 |
+
"eval_logps/rejected": -37.27513122558594,
|
| 2143 |
+
"eval_loss": 0.47534918785095215,
|
| 2144 |
+
"eval_rewards/accuracies": 0.5830564498901367,
|
| 2145 |
+
"eval_rewards/chosen": 0.2859070897102356,
|
| 2146 |
+
"eval_rewards/margins": 0.11686603724956512,
|
| 2147 |
+
"eval_rewards/rejected": 0.16904108226299286,
|
| 2148 |
+
"eval_runtime": 145.3117,
|
| 2149 |
+
"eval_samples_per_second": 2.36,
|
| 2150 |
+
"eval_steps_per_second": 0.296,
|
| 2151 |
+
"step": 1300
|
| 2152 |
+
},
|
| 2153 |
+
{
|
| 2154 |
+
"epoch": 3.4,
|
| 2155 |
+
"grad_norm": 2.8125,
|
| 2156 |
+
"learning_rate": 3.321087280364757e-07,
|
| 2157 |
+
"logits/chosen": -1.9337997436523438,
|
| 2158 |
+
"logits/rejected": -1.9138942956924438,
|
| 2159 |
+
"logps/chosen": -28.2788143157959,
|
| 2160 |
+
"logps/rejected": -35.673282623291016,
|
| 2161 |
+
"loss": 0.1617,
|
| 2162 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2163 |
+
"rewards/chosen": 1.984209418296814,
|
| 2164 |
+
"rewards/margins": 2.339022159576416,
|
| 2165 |
+
"rewards/rejected": -0.3548130691051483,
|
| 2166 |
+
"step": 1310
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"epoch": 3.43,
|
| 2170 |
+
"grad_norm": 2.890625,
|
| 2171 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2172 |
+
"logits/chosen": -1.9120073318481445,
|
| 2173 |
+
"logits/rejected": -1.917673110961914,
|
| 2174 |
+
"logps/chosen": -29.9759578704834,
|
| 2175 |
+
"logps/rejected": -33.02309036254883,
|
| 2176 |
+
"loss": 0.1289,
|
| 2177 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2178 |
+
"rewards/chosen": 2.1404826641082764,
|
| 2179 |
+
"rewards/margins": 2.5131189823150635,
|
| 2180 |
+
"rewards/rejected": -0.3726363182067871,
|
| 2181 |
+
"step": 1320
|
| 2182 |
+
},
|
| 2183 |
+
{
|
| 2184 |
+
"epoch": 3.45,
|
| 2185 |
+
"grad_norm": 3.6875,
|
| 2186 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2187 |
+
"logits/chosen": -1.884010910987854,
|
| 2188 |
+
"logits/rejected": -1.879150390625,
|
| 2189 |
+
"logps/chosen": -30.3497257232666,
|
| 2190 |
+
"logps/rejected": -35.16407012939453,
|
| 2191 |
+
"loss": 0.154,
|
| 2192 |
+
"rewards/accuracies": 0.9375,
|
| 2193 |
+
"rewards/chosen": 1.8709485530853271,
|
| 2194 |
+
"rewards/margins": 2.2851948738098145,
|
| 2195 |
+
"rewards/rejected": -0.41424623131752014,
|
| 2196 |
+
"step": 1330
|
| 2197 |
+
},
|
| 2198 |
+
{
|
| 2199 |
+
"epoch": 3.48,
|
| 2200 |
+
"grad_norm": 2.046875,
|
| 2201 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2202 |
+
"logits/chosen": -1.7457748651504517,
|
| 2203 |
+
"logits/rejected": -1.755414366722107,
|
| 2204 |
+
"logps/chosen": -30.9390811920166,
|
| 2205 |
+
"logps/rejected": -31.315013885498047,
|
| 2206 |
+
"loss": 0.1642,
|
| 2207 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2208 |
+
"rewards/chosen": 2.0023577213287354,
|
| 2209 |
+
"rewards/margins": 2.198988676071167,
|
| 2210 |
+
"rewards/rejected": -0.19663105905056,
|
| 2211 |
+
"step": 1340
|
| 2212 |
+
},
|
| 2213 |
+
{
|
| 2214 |
+
"epoch": 3.51,
|
| 2215 |
+
"grad_norm": 3.21875,
|
| 2216 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2217 |
+
"logits/chosen": -1.8068130016326904,
|
| 2218 |
+
"logits/rejected": -1.8115451335906982,
|
| 2219 |
+
"logps/chosen": -30.966552734375,
|
| 2220 |
+
"logps/rejected": -33.47523880004883,
|
| 2221 |
+
"loss": 0.1355,
|
| 2222 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2223 |
+
"rewards/chosen": 2.0890681743621826,
|
| 2224 |
+
"rewards/margins": 2.2783665657043457,
|
| 2225 |
+
"rewards/rejected": -0.18929778039455414,
|
| 2226 |
+
"step": 1350
|
| 2227 |
+
},
|
| 2228 |
+
{
|
| 2229 |
+
"epoch": 3.53,
|
| 2230 |
+
"grad_norm": 3.96875,
|
| 2231 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2232 |
+
"logits/chosen": -1.8347194194793701,
|
| 2233 |
+
"logits/rejected": -1.840275764465332,
|
| 2234 |
+
"logps/chosen": -31.000539779663086,
|
| 2235 |
+
"logps/rejected": -35.11589813232422,
|
| 2236 |
+
"loss": 0.1568,
|
| 2237 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2238 |
+
"rewards/chosen": 1.981787085533142,
|
| 2239 |
+
"rewards/margins": 2.36877179145813,
|
| 2240 |
+
"rewards/rejected": -0.38698458671569824,
|
| 2241 |
+
"step": 1360
|
| 2242 |
+
},
|
| 2243 |
+
{
|
| 2244 |
+
"epoch": 3.56,
|
| 2245 |
+
"grad_norm": 2.125,
|
| 2246 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2247 |
+
"logits/chosen": -1.9270436763763428,
|
| 2248 |
+
"logits/rejected": -1.9225692749023438,
|
| 2249 |
+
"logps/chosen": -29.990549087524414,
|
| 2250 |
+
"logps/rejected": -31.350738525390625,
|
| 2251 |
+
"loss": 0.169,
|
| 2252 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2253 |
+
"rewards/chosen": 1.8515956401824951,
|
| 2254 |
+
"rewards/margins": 2.158782482147217,
|
| 2255 |
+
"rewards/rejected": -0.30718696117401123,
|
| 2256 |
+
"step": 1370
|
| 2257 |
+
},
|
| 2258 |
+
{
|
| 2259 |
+
"epoch": 3.58,
|
| 2260 |
+
"grad_norm": 3.859375,
|
| 2261 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2262 |
+
"logits/chosen": -1.8426074981689453,
|
| 2263 |
+
"logits/rejected": -1.8444961309432983,
|
| 2264 |
+
"logps/chosen": -27.056955337524414,
|
| 2265 |
+
"logps/rejected": -31.148794174194336,
|
| 2266 |
+
"loss": 0.1754,
|
| 2267 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2268 |
+
"rewards/chosen": 1.8216432332992554,
|
| 2269 |
+
"rewards/margins": 2.104875087738037,
|
| 2270 |
+
"rewards/rejected": -0.2832319140434265,
|
| 2271 |
+
"step": 1380
|
| 2272 |
+
},
|
| 2273 |
+
{
|
| 2274 |
+
"epoch": 3.61,
|
| 2275 |
+
"grad_norm": 4.9375,
|
| 2276 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2277 |
+
"logits/chosen": -1.836014986038208,
|
| 2278 |
+
"logits/rejected": -1.8406941890716553,
|
| 2279 |
+
"logps/chosen": -29.474157333374023,
|
| 2280 |
+
"logps/rejected": -33.005123138427734,
|
| 2281 |
+
"loss": 0.185,
|
| 2282 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2283 |
+
"rewards/chosen": 1.7817341089248657,
|
| 2284 |
+
"rewards/margins": 1.7936760187149048,
|
| 2285 |
+
"rewards/rejected": -0.011941825971007347,
|
| 2286 |
+
"step": 1390
|
| 2287 |
+
},
|
| 2288 |
+
{
|
| 2289 |
+
"epoch": 3.64,
|
| 2290 |
+
"grad_norm": 2.03125,
|
| 2291 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2292 |
+
"logits/chosen": -1.9031331539154053,
|
| 2293 |
+
"logits/rejected": -1.9063663482666016,
|
| 2294 |
+
"logps/chosen": -32.08364486694336,
|
| 2295 |
+
"logps/rejected": -33.883567810058594,
|
| 2296 |
+
"loss": 0.1492,
|
| 2297 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2298 |
+
"rewards/chosen": 1.9928478002548218,
|
| 2299 |
+
"rewards/margins": 2.3148512840270996,
|
| 2300 |
+
"rewards/rejected": -0.32200372219085693,
|
| 2301 |
+
"step": 1400
|
| 2302 |
+
},
|
| 2303 |
+
{
|
| 2304 |
+
"epoch": 3.64,
|
| 2305 |
+
"eval_logits/chosen": -2.1602745056152344,
|
| 2306 |
+
"eval_logits/rejected": -2.155562400817871,
|
| 2307 |
+
"eval_logps/chosen": -33.62794876098633,
|
| 2308 |
+
"eval_logps/rejected": -37.28424072265625,
|
| 2309 |
+
"eval_loss": 0.4743622839450836,
|
| 2310 |
+
"eval_rewards/accuracies": 0.5917773842811584,
|
| 2311 |
+
"eval_rewards/chosen": 0.2846195697784424,
|
| 2312 |
+
"eval_rewards/margins": 0.12195253372192383,
|
| 2313 |
+
"eval_rewards/rejected": 0.16266703605651855,
|
| 2314 |
+
"eval_runtime": 145.0405,
|
| 2315 |
+
"eval_samples_per_second": 2.365,
|
| 2316 |
+
"eval_steps_per_second": 0.296,
|
| 2317 |
+
"step": 1400
|
| 2318 |
+
},
|
| 2319 |
+
{
|
| 2320 |
+
"epoch": 3.66,
|
| 2321 |
+
"grad_norm": 3.1875,
|
| 2322 |
+
"learning_rate": 1.0775199359171346e-07,
|
| 2323 |
+
"logits/chosen": -1.9723188877105713,
|
| 2324 |
+
"logits/rejected": -1.9656269550323486,
|
| 2325 |
+
"logps/chosen": -30.971176147460938,
|
| 2326 |
+
"logps/rejected": -34.55162811279297,
|
| 2327 |
+
"loss": 0.1679,
|
| 2328 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2329 |
+
"rewards/chosen": 1.9819583892822266,
|
| 2330 |
+
"rewards/margins": 2.196903705596924,
|
| 2331 |
+
"rewards/rejected": -0.2149454653263092,
|
| 2332 |
+
"step": 1410
|
| 2333 |
+
},
|
| 2334 |
+
{
|
| 2335 |
+
"epoch": 3.69,
|
| 2336 |
+
"grad_norm": 4.28125,
|
| 2337 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2338 |
+
"logits/chosen": -1.8590764999389648,
|
| 2339 |
+
"logits/rejected": -1.8698148727416992,
|
| 2340 |
+
"logps/chosen": -31.296035766601562,
|
| 2341 |
+
"logps/rejected": -34.09710693359375,
|
| 2342 |
+
"loss": 0.1478,
|
| 2343 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2344 |
+
"rewards/chosen": 1.9972867965698242,
|
| 2345 |
+
"rewards/margins": 2.3515076637268066,
|
| 2346 |
+
"rewards/rejected": -0.35422074794769287,
|
| 2347 |
+
"step": 1420
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.71,
|
| 2351 |
+
"grad_norm": 3.78125,
|
| 2352 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2353 |
+
"logits/chosen": -1.9422740936279297,
|
| 2354 |
+
"logits/rejected": -1.9541749954223633,
|
| 2355 |
+
"logps/chosen": -32.07694625854492,
|
| 2356 |
+
"logps/rejected": -34.164581298828125,
|
| 2357 |
+
"loss": 0.1459,
|
| 2358 |
+
"rewards/accuracies": 0.9375,
|
| 2359 |
+
"rewards/chosen": 2.0887014865875244,
|
| 2360 |
+
"rewards/margins": 2.4570722579956055,
|
| 2361 |
+
"rewards/rejected": -0.3683708906173706,
|
| 2362 |
+
"step": 1430
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"epoch": 3.74,
|
| 2366 |
+
"grad_norm": 3.015625,
|
| 2367 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2368 |
+
"logits/chosen": -1.9317998886108398,
|
| 2369 |
+
"logits/rejected": -1.933950424194336,
|
| 2370 |
+
"logps/chosen": -30.444339752197266,
|
| 2371 |
+
"logps/rejected": -34.485233306884766,
|
| 2372 |
+
"loss": 0.1473,
|
| 2373 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2374 |
+
"rewards/chosen": 1.9348328113555908,
|
| 2375 |
+
"rewards/margins": 2.3725178241729736,
|
| 2376 |
+
"rewards/rejected": -0.4376850128173828,
|
| 2377 |
+
"step": 1440
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.77,
|
| 2381 |
+
"grad_norm": 3.859375,
|
| 2382 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2383 |
+
"logits/chosen": -1.9155280590057373,
|
| 2384 |
+
"logits/rejected": -1.9085607528686523,
|
| 2385 |
+
"logps/chosen": -32.7889289855957,
|
| 2386 |
+
"logps/rejected": -34.09833526611328,
|
| 2387 |
+
"loss": 0.172,
|
| 2388 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2389 |
+
"rewards/chosen": 1.9455804824829102,
|
| 2390 |
+
"rewards/margins": 2.1751322746276855,
|
| 2391 |
+
"rewards/rejected": -0.22955195605754852,
|
| 2392 |
+
"step": 1450
|
| 2393 |
+
},
|
| 2394 |
+
{
|
| 2395 |
+
"epoch": 3.79,
|
| 2396 |
+
"grad_norm": 2.875,
|
| 2397 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2398 |
+
"logits/chosen": -1.884874701499939,
|
| 2399 |
+
"logits/rejected": -1.890824556350708,
|
| 2400 |
+
"logps/chosen": -28.527862548828125,
|
| 2401 |
+
"logps/rejected": -34.241851806640625,
|
| 2402 |
+
"loss": 0.159,
|
| 2403 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2404 |
+
"rewards/chosen": 1.879720687866211,
|
| 2405 |
+
"rewards/margins": 2.2800161838531494,
|
| 2406 |
+
"rewards/rejected": -0.40029558539390564,
|
| 2407 |
+
"step": 1460
|
| 2408 |
+
},
|
| 2409 |
+
{
|
| 2410 |
+
"epoch": 3.82,
|
| 2411 |
+
"grad_norm": 3.25,
|
| 2412 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2413 |
+
"logits/chosen": -1.8300120830535889,
|
| 2414 |
+
"logits/rejected": -1.8280490636825562,
|
| 2415 |
+
"logps/chosen": -28.143136978149414,
|
| 2416 |
+
"logps/rejected": -30.871578216552734,
|
| 2417 |
+
"loss": 0.1654,
|
| 2418 |
+
"rewards/accuracies": 0.9375,
|
| 2419 |
+
"rewards/chosen": 1.8916912078857422,
|
| 2420 |
+
"rewards/margins": 2.0858335494995117,
|
| 2421 |
+
"rewards/rejected": -0.1941421777009964,
|
| 2422 |
+
"step": 1470
|
| 2423 |
+
},
|
| 2424 |
+
{
|
| 2425 |
+
"epoch": 3.84,
|
| 2426 |
+
"grad_norm": 3.6875,
|
| 2427 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2428 |
+
"logits/chosen": -2.004854440689087,
|
| 2429 |
+
"logits/rejected": -1.999352216720581,
|
| 2430 |
+
"logps/chosen": -32.60303497314453,
|
| 2431 |
+
"logps/rejected": -32.43815994262695,
|
| 2432 |
+
"loss": 0.16,
|
| 2433 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2434 |
+
"rewards/chosen": 1.8775575160980225,
|
| 2435 |
+
"rewards/margins": 2.14457368850708,
|
| 2436 |
+
"rewards/rejected": -0.26701638102531433,
|
| 2437 |
+
"step": 1480
|
| 2438 |
+
},
|
| 2439 |
+
{
|
| 2440 |
+
"epoch": 3.87,
|
| 2441 |
+
"grad_norm": 3.140625,
|
| 2442 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2443 |
+
"logits/chosen": -1.8964626789093018,
|
| 2444 |
+
"logits/rejected": -1.8985786437988281,
|
| 2445 |
+
"logps/chosen": -26.187755584716797,
|
| 2446 |
+
"logps/rejected": -28.33608627319336,
|
| 2447 |
+
"loss": 0.1956,
|
| 2448 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2449 |
+
"rewards/chosen": 1.6020386219024658,
|
| 2450 |
+
"rewards/margins": 1.821703314781189,
|
| 2451 |
+
"rewards/rejected": -0.21966469287872314,
|
| 2452 |
+
"step": 1490
|
| 2453 |
+
},
|
| 2454 |
+
{
|
| 2455 |
+
"epoch": 3.9,
|
| 2456 |
+
"grad_norm": 4.375,
|
| 2457 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2458 |
+
"logits/chosen": -1.905769944190979,
|
| 2459 |
+
"logits/rejected": -1.9003044366836548,
|
| 2460 |
+
"logps/chosen": -30.289718627929688,
|
| 2461 |
+
"logps/rejected": -32.631587982177734,
|
| 2462 |
+
"loss": 0.1694,
|
| 2463 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2464 |
+
"rewards/chosen": 1.7577499151229858,
|
| 2465 |
+
"rewards/margins": 2.221593141555786,
|
| 2466 |
+
"rewards/rejected": -0.46384334564208984,
|
| 2467 |
+
"step": 1500
|
| 2468 |
+
},
|
| 2469 |
+
{
|
| 2470 |
+
"epoch": 3.9,
|
| 2471 |
+
"eval_logits/chosen": -2.1607134342193604,
|
| 2472 |
+
"eval_logits/rejected": -2.1560044288635254,
|
| 2473 |
+
"eval_logps/chosen": -33.631370544433594,
|
| 2474 |
+
"eval_logps/rejected": -37.28603744506836,
|
| 2475 |
+
"eval_loss": 0.47470974922180176,
|
| 2476 |
+
"eval_rewards/accuracies": 0.5568937063217163,
|
| 2477 |
+
"eval_rewards/chosen": 0.2822282910346985,
|
| 2478 |
+
"eval_rewards/margins": 0.12082076817750931,
|
| 2479 |
+
"eval_rewards/rejected": 0.16140751540660858,
|
| 2480 |
+
"eval_runtime": 145.1244,
|
| 2481 |
+
"eval_samples_per_second": 2.363,
|
| 2482 |
+
"eval_steps_per_second": 0.296,
|
| 2483 |
+
"step": 1500
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 3.92,
|
| 2487 |
+
"grad_norm": 5.78125,
|
| 2488 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2489 |
+
"logits/chosen": -1.8295776844024658,
|
| 2490 |
+
"logits/rejected": -1.8337701559066772,
|
| 2491 |
+
"logps/chosen": -31.576879501342773,
|
| 2492 |
+
"logps/rejected": -34.6661262512207,
|
| 2493 |
+
"loss": 0.1827,
|
| 2494 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2495 |
+
"rewards/chosen": 1.8158810138702393,
|
| 2496 |
+
"rewards/margins": 1.994899034500122,
|
| 2497 |
+
"rewards/rejected": -0.17901787161827087,
|
| 2498 |
+
"step": 1510
|
| 2499 |
+
},
|
| 2500 |
+
{
|
| 2501 |
+
"epoch": 3.95,
|
| 2502 |
+
"grad_norm": 2.1875,
|
| 2503 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2504 |
+
"logits/chosen": -1.8878551721572876,
|
| 2505 |
+
"logits/rejected": -1.8862745761871338,
|
| 2506 |
+
"logps/chosen": -28.058696746826172,
|
| 2507 |
+
"logps/rejected": -33.05814743041992,
|
| 2508 |
+
"loss": 0.1623,
|
| 2509 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2510 |
+
"rewards/chosen": 1.872835397720337,
|
| 2511 |
+
"rewards/margins": 2.200737953186035,
|
| 2512 |
+
"rewards/rejected": -0.3279026448726654,
|
| 2513 |
+
"step": 1520
|
| 2514 |
+
},
|
| 2515 |
+
{
|
| 2516 |
+
"epoch": 3.97,
|
| 2517 |
+
"grad_norm": 4.0625,
|
| 2518 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2519 |
+
"logits/chosen": -1.9854892492294312,
|
| 2520 |
+
"logits/rejected": -1.9778953790664673,
|
| 2521 |
+
"logps/chosen": -25.75176429748535,
|
| 2522 |
+
"logps/rejected": -29.129873275756836,
|
| 2523 |
+
"loss": 0.1804,
|
| 2524 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2525 |
+
"rewards/chosen": 1.6178255081176758,
|
| 2526 |
+
"rewards/margins": 1.938612699508667,
|
| 2527 |
+
"rewards/rejected": -0.3207874298095703,
|
| 2528 |
+
"step": 1530
|
| 2529 |
+
},
|
| 2530 |
+
{
|
| 2531 |
+
"epoch": 4.0,
|
| 2532 |
+
"grad_norm": 2.921875,
|
| 2533 |
+
"learning_rate": 0.0,
|
| 2534 |
+
"logits/chosen": -1.8875421285629272,
|
| 2535 |
+
"logits/rejected": -1.877659797668457,
|
| 2536 |
+
"logps/chosen": -30.428552627563477,
|
| 2537 |
+
"logps/rejected": -36.047637939453125,
|
| 2538 |
+
"loss": 0.1484,
|
| 2539 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2540 |
+
"rewards/chosen": 1.8463512659072876,
|
| 2541 |
+
"rewards/margins": 2.291874408721924,
|
| 2542 |
+
"rewards/rejected": -0.445523202419281,
|
| 2543 |
+
"step": 1540
|
| 2544 |
+
},
|
| 2545 |
+
{
|
| 2546 |
+
"epoch": 4.0,
|
| 2547 |
+
"step": 1540,
|
| 2548 |
"total_flos": 0.0,
|
| 2549 |
+
"train_loss": 0.17319794591371115,
|
| 2550 |
+
"train_runtime": 10768.0107,
|
| 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,
|