Commit
•
3be5042
1
Parent(s):
73046c1
Model save
Browse files- README.md +29 -19
- adapter_model.safetensors +1 -1
- all_results.json +5 -5
- train_results.json +5 -5
- trainer_state.json +1903 -87
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 |
|
|
@@ -57,15 +55,27 @@ 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 |
-
| 0.4749 | 0.26 | 100 |
|
| 67 |
-
| 0.4376 | 0.52 | 200 |
|
| 68 |
-
| 0.3835 | 0.78 | 300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
|
|
|
| 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 an unknown dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.4684
|
| 22 |
+
- Rewards/chosen: 0.3669
|
| 23 |
+
- Rewards/rejected: 0.2161
|
| 24 |
+
- Rewards/accuracies: 0.5743
|
| 25 |
+
- Rewards/margins: 0.1508
|
| 26 |
+
- Logps/rejected: -37.2465
|
| 27 |
+
- Logps/chosen: -33.5759
|
| 28 |
+
- Logits/rejected: -2.1622
|
| 29 |
+
- Logits/chosen: -2.1669
|
| 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 |
+
| 0.4749 | 0.26 | 100 | -2.2375 | -2.2327 | -33.8512 | -37.3537 | 0.4963 | 0.5336 | 0.1467 | 0.0164 | 0.1303 |
|
| 65 |
+
| 0.4376 | 0.52 | 200 | -2.2339 | -2.2291 | -33.7896 | -37.2955 | 0.4956 | 0.5486 | 0.1959 | 0.0191 | 0.1769 |
|
| 66 |
+
| 0.3835 | 0.78 | 300 | -2.2312 | -2.2264 | -33.7789 | -37.2872 | 0.4950 | 0.5245 | 0.2045 | 0.0210 | 0.1836 |
|
| 67 |
+
| 0.3117 | 1.04 | 400 | 0.4891 | 0.3054 | 0.2586 | 0.5652 | 0.0468 | -37.1934 | -33.6528 | -2.2112 | -2.2160 |
|
| 68 |
+
| 0.2459 | 1.3 | 500 | 0.4885 | 0.3186 | 0.2671 | 0.5623 | 0.0514 | -37.1827 | -33.6364 | -2.1858 | -2.1906 |
|
| 69 |
+
| 0.2639 | 1.56 | 600 | 0.4750 | 0.3623 | 0.2503 | 0.5855 | 0.1120 | -37.2038 | -33.5817 | -2.1784 | -2.1832 |
|
| 70 |
+
| 0.2437 | 1.82 | 700 | 0.4742 | 0.3483 | 0.2298 | 0.5748 | 0.1184 | -37.2294 | -33.5992 | -2.1739 | -2.1786 |
|
| 71 |
+
| 0.1567 | 2.08 | 800 | 0.4695 | 0.3879 | 0.2480 | 0.5826 | 0.1399 | -37.2066 | -33.5496 | -2.1755 | -2.1803 |
|
| 72 |
+
| 0.131 | 2.34 | 900 | 0.4716 | 0.3533 | 0.2206 | 0.5860 | 0.1326 | -37.2408 | -33.5930 | -2.1658 | -2.1705 |
|
| 73 |
+
| 0.1784 | 2.6 | 1000 | 0.4673 | 0.3677 | 0.2130 | 0.5860 | 0.1548 | -37.2504 | -33.5749 | -2.1646 | -2.1693 |
|
| 74 |
+
| 0.1956 | 2.86 | 1100 | 0.4706 | 0.3580 | 0.2180 | 0.5860 | 0.1400 | -37.2442 | -33.5871 | -2.1622 | -2.1669 |
|
| 75 |
+
| 0.137 | 3.12 | 1200 | 0.4680 | 0.3694 | 0.2182 | 0.6063 | 0.1511 | -37.2438 | -33.5728 | -2.1625 | -2.1672 |
|
| 76 |
+
| 0.1211 | 3.38 | 1300 | 0.4705 | 0.3633 | 0.2219 | 0.5918 | 0.1414 | -37.2393 | -33.5805 | -2.1622 | -2.1669 |
|
| 77 |
+
| 0.1553 | 3.64 | 1400 | 0.4654 | 0.3698 | 0.2068 | 0.6034 | 0.1630 | -37.2582 | -33.5723 | -2.1621 | -2.1668 |
|
| 78 |
+
| 0.1447 | 3.9 | 1500 | 0.4684 | 0.3669 | 0.2161 | 0.5743 | 0.1508 | -37.2465 | -33.5759 | -2.1622 | -2.1669 |
|
| 79 |
|
| 80 |
|
| 81 |
### Framework versions
|
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 sha256:3847fa5053c205f61f041cbc4e44ddec25eb91ff0ac76b5bc16fcf571960eb4b
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.2309982776641846,
|
| 4 |
"eval_logits/rejected": -2.2261931896209717,
|
| 5 |
"eval_logps/chosen": -33.767173767089844,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.356,
|
| 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": -2.2309982776641846,
|
| 4 |
"eval_logits/rejected": -2.2261931896209717,
|
| 5 |
"eval_logps/chosen": -33.767173767089844,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.356,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.16076272354497537,
|
| 17 |
+
"train_runtime": 10802.3033,
|
| 18 |
"train_samples": 3079,
|
| 19 |
+
"train_samples_per_second": 1.14,
|
| 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": 0.16076272354497537,
|
| 4 |
+
"train_runtime": 10802.3033,
|
| 5 |
"train_samples": 3079,
|
| 6 |
+
"train_samples_per_second": 1.14,
|
| 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 +492,1946 @@
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
-
"learning_rate":
|
| 496 |
-
"logits/chosen": -1.
|
| 497 |
-
"logits/rejected": -1.
|
| 498 |
-
"logps/chosen": -31.
|
| 499 |
-
"logps/rejected": -33.
|
| 500 |
-
"loss": 0.
|
| 501 |
"rewards/accuracies": 0.7124999761581421,
|
| 502 |
-
"rewards/chosen": 0.
|
| 503 |
-
"rewards/margins": 0.
|
| 504 |
-
"rewards/rejected": 0.
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
-
"learning_rate": 4.
|
| 510 |
-
"logits/chosen": -1.
|
| 511 |
-
"logits/rejected": -1.
|
| 512 |
-
"logps/chosen": -34.
|
| 513 |
-
"logps/rejected": -33.
|
| 514 |
-
"loss": 0.
|
| 515 |
"rewards/accuracies": 0.7124999761581421,
|
| 516 |
-
"rewards/chosen": 0.
|
| 517 |
-
"rewards/margins": 0.
|
| 518 |
-
"rewards/rejected": 0.
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
-
"learning_rate":
|
| 524 |
-
"logits/chosen": -2.
|
| 525 |
-
"logits/rejected": -2.
|
| 526 |
-
"logps/chosen": -32.
|
| 527 |
-
"logps/rejected": -32.
|
| 528 |
-
"loss": 0.
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": 0.
|
| 531 |
-
"rewards/margins": 0.
|
| 532 |
-
"rewards/rejected": 0.
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -33.
|
| 541 |
-
"logps/rejected": -32.
|
| 542 |
-
"loss": 0.
|
| 543 |
-
"rewards/accuracies": 0.
|
| 544 |
-
"rewards/chosen": 0.
|
| 545 |
-
"rewards/margins": 0.
|
| 546 |
-
"rewards/rejected": 0.
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
-
"learning_rate":
|
| 552 |
-
"logits/chosen": -1.
|
| 553 |
-
"logits/rejected": -1.
|
| 554 |
-
"logps/chosen": -32.
|
| 555 |
-
"logps/rejected": -32.
|
| 556 |
-
"loss": 0.
|
| 557 |
-
"rewards/accuracies": 0.
|
| 558 |
-
"rewards/chosen": 0.
|
| 559 |
-
"rewards/margins": 0.
|
| 560 |
-
"rewards/rejected": 0.
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
-
"learning_rate":
|
| 566 |
-
"logits/chosen": -1.
|
| 567 |
-
"logits/rejected": -1.
|
| 568 |
-
"logps/chosen": -31.
|
| 569 |
-
"logps/rejected": -35.
|
| 570 |
-
"loss": 0.
|
| 571 |
"rewards/accuracies": 0.675000011920929,
|
| 572 |
-
"rewards/chosen": 0.
|
| 573 |
-
"rewards/margins": 0.
|
| 574 |
-
"rewards/rejected": 0.
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
-
"learning_rate":
|
| 580 |
-
"logits/chosen": -2.
|
| 581 |
-
"logits/rejected": -2.
|
| 582 |
-
"logps/chosen": -
|
| 583 |
-
"logps/rejected": -28.
|
| 584 |
-
"loss": 0.
|
| 585 |
-
"rewards/accuracies": 0.
|
| 586 |
-
"rewards/chosen": 0.
|
| 587 |
-
"rewards/margins": 0.
|
| 588 |
-
"rewards/rejected": 0.
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
-
"learning_rate":
|
| 594 |
-
"logits/chosen": -1.
|
| 595 |
-
"logits/rejected": -1.
|
| 596 |
-
"logps/chosen": -33.
|
| 597 |
-
"logps/rejected": -30.
|
| 598 |
-
"loss": 0.
|
| 599 |
-
"rewards/accuracies": 0.
|
| 600 |
-
"rewards/chosen": 0.
|
| 601 |
-
"rewards/margins": 0.
|
| 602 |
-
"rewards/rejected": 0.
|
| 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 |
+
"learning_rate": 4.84533120650964e-06,
|
| 496 |
+
"logits/chosen": -1.9174331426620483,
|
| 497 |
+
"logits/rejected": -1.9142847061157227,
|
| 498 |
+
"logps/chosen": -31.049779891967773,
|
| 499 |
+
"logps/rejected": -33.54362487792969,
|
| 500 |
+
"loss": 0.4366,
|
| 501 |
"rewards/accuracies": 0.7124999761581421,
|
| 502 |
+
"rewards/chosen": 0.4428572654724121,
|
| 503 |
+
"rewards/margins": 0.28403452038764954,
|
| 504 |
+
"rewards/rejected": 0.15882274508476257,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.9670381546020508,
|
| 511 |
+
"logits/rejected": -1.9549226760864258,
|
| 512 |
+
"logps/chosen": -34.00459671020508,
|
| 513 |
+
"logps/rejected": -33.429378509521484,
|
| 514 |
+
"loss": 0.4145,
|
| 515 |
"rewards/accuracies": 0.7124999761581421,
|
| 516 |
+
"rewards/chosen": 0.4731821119785309,
|
| 517 |
+
"rewards/margins": 0.38561564683914185,
|
| 518 |
+
"rewards/rejected": 0.08756652474403381,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -2.001861095428467,
|
| 525 |
+
"logits/rejected": -2.0005335807800293,
|
| 526 |
+
"logps/chosen": -32.78104019165039,
|
| 527 |
+
"logps/rejected": -32.248443603515625,
|
| 528 |
+
"loss": 0.4183,
|
| 529 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 530 |
+
"rewards/chosen": 0.5638058185577393,
|
| 531 |
+
"rewards/margins": 0.37889450788497925,
|
| 532 |
+
"rewards/rejected": 0.18491129577159882,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.0892767906188965,
|
| 539 |
+
"logits/rejected": -2.073693037033081,
|
| 540 |
+
"logps/chosen": -33.377323150634766,
|
| 541 |
+
"logps/rejected": -32.779090881347656,
|
| 542 |
+
"loss": 0.4167,
|
| 543 |
+
"rewards/accuracies": 0.6875,
|
| 544 |
+
"rewards/chosen": 0.641654908657074,
|
| 545 |
+
"rewards/margins": 0.37049978971481323,
|
| 546 |
+
"rewards/rejected": 0.2711551785469055,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.9618675708770752,
|
| 553 |
+
"logits/rejected": -1.9610551595687866,
|
| 554 |
+
"logps/chosen": -32.51572799682617,
|
| 555 |
+
"logps/rejected": -32.19974899291992,
|
| 556 |
+
"loss": 0.4137,
|
| 557 |
+
"rewards/accuracies": 0.699999988079071,
|
| 558 |
+
"rewards/chosen": 0.6130937933921814,
|
| 559 |
+
"rewards/margins": 0.4007722735404968,
|
| 560 |
+
"rewards/rejected": 0.21232159435749054,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.916259765625,
|
| 567 |
+
"logits/rejected": -1.9264857769012451,
|
| 568 |
+
"logps/chosen": -31.4000244140625,
|
| 569 |
+
"logps/rejected": -35.018882751464844,
|
| 570 |
+
"loss": 0.4065,
|
| 571 |
"rewards/accuracies": 0.675000011920929,
|
| 572 |
+
"rewards/chosen": 0.6437316536903381,
|
| 573 |
+
"rewards/margins": 0.427828311920166,
|
| 574 |
+
"rewards/rejected": 0.21590332686901093,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.050283908843994,
|
| 581 |
+
"logits/rejected": -2.043849468231201,
|
| 582 |
+
"logps/chosen": -32.88298797607422,
|
| 583 |
+
"logps/rejected": -28.950063705444336,
|
| 584 |
+
"loss": 0.4093,
|
| 585 |
+
"rewards/accuracies": 0.699999988079071,
|
| 586 |
+
"rewards/chosen": 0.6151864528656006,
|
| 587 |
+
"rewards/margins": 0.3940979838371277,
|
| 588 |
+
"rewards/rejected": 0.2210884839296341,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.9082043170928955,
|
| 595 |
+
"logits/rejected": -1.9103626012802124,
|
| 596 |
+
"logps/chosen": -33.2851676940918,
|
| 597 |
+
"logps/rejected": -30.665771484375,
|
| 598 |
+
"loss": 0.3759,
|
| 599 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 600 |
+
"rewards/chosen": 0.7631045579910278,
|
| 601 |
+
"rewards/margins": 0.5976192355155945,
|
| 602 |
+
"rewards/rejected": 0.1654852330684662,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9357115030288696,
|
| 609 |
+
"logits/rejected": -1.9344127178192139,
|
| 610 |
+
"logps/chosen": -33.36685562133789,
|
| 611 |
+
"logps/rejected": -35.62567138671875,
|
| 612 |
+
"loss": 0.3409,
|
| 613 |
+
"rewards/accuracies": 0.7583333253860474,
|
| 614 |
+
"rewards/chosen": 0.8143496513366699,
|
| 615 |
+
"rewards/margins": 0.790077805519104,
|
| 616 |
+
"rewards/rejected": 0.02427184209227562,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.8698232173919678,
|
| 623 |
+
"logits/rejected": -1.8616269826889038,
|
| 624 |
+
"logps/chosen": -30.623443603515625,
|
| 625 |
+
"logps/rejected": -36.03126907348633,
|
| 626 |
+
"loss": 0.3117,
|
| 627 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 628 |
+
"rewards/chosen": 0.920426070690155,
|
| 629 |
+
"rewards/margins": 0.9929585456848145,
|
| 630 |
+
"rewards/rejected": -0.07253243774175644,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.2160162925720215,
|
| 636 |
+
"eval_logits/rejected": -2.2111997604370117,
|
| 637 |
+
"eval_logps/chosen": -33.652793884277344,
|
| 638 |
+
"eval_logps/rejected": -37.193359375,
|
| 639 |
+
"eval_loss": 0.4891091287136078,
|
| 640 |
+
"eval_rewards/accuracies": 0.5651993155479431,
|
| 641 |
+
"eval_rewards/chosen": 0.30540817975997925,
|
| 642 |
+
"eval_rewards/margins": 0.04680241644382477,
|
| 643 |
+
"eval_rewards/rejected": 0.2586057484149933,
|
| 644 |
+
"eval_runtime": 146.2599,
|
| 645 |
+
"eval_samples_per_second": 2.345,
|
| 646 |
+
"eval_steps_per_second": 0.294,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 1.06,
|
| 651 |
+
"learning_rate": 4.590793060955158e-06,
|
| 652 |
+
"logits/chosen": -2.044127941131592,
|
| 653 |
+
"logits/rejected": -2.046945571899414,
|
| 654 |
+
"logps/chosen": -31.79901695251465,
|
| 655 |
+
"logps/rejected": -34.87114334106445,
|
| 656 |
+
"loss": 0.299,
|
| 657 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 658 |
+
"rewards/chosen": 1.003937005996704,
|
| 659 |
+
"rewards/margins": 1.0665733814239502,
|
| 660 |
+
"rewards/rejected": -0.0626363754272461,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.8841445446014404,
|
| 667 |
+
"logits/rejected": -1.8827756643295288,
|
| 668 |
+
"logps/chosen": -27.981430053710938,
|
| 669 |
+
"logps/rejected": -32.39961624145508,
|
| 670 |
+
"loss": 0.3151,
|
| 671 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 672 |
+
"rewards/chosen": 0.868902325630188,
|
| 673 |
+
"rewards/margins": 0.9344655871391296,
|
| 674 |
+
"rewards/rejected": -0.06556323915719986,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.8422701358795166,
|
| 681 |
+
"logits/rejected": -1.835268259048462,
|
| 682 |
+
"logps/chosen": -32.6194953918457,
|
| 683 |
+
"logps/rejected": -34.0600700378418,
|
| 684 |
+
"loss": 0.2904,
|
| 685 |
+
"rewards/accuracies": 0.9375,
|
| 686 |
+
"rewards/chosen": 1.1636759042739868,
|
| 687 |
+
"rewards/margins": 1.0155143737792969,
|
| 688 |
+
"rewards/rejected": 0.14816154539585114,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -2.0123181343078613,
|
| 695 |
+
"logits/rejected": -2.0072364807128906,
|
| 696 |
+
"logps/chosen": -30.41460609436035,
|
| 697 |
+
"logps/rejected": -32.033443450927734,
|
| 698 |
+
"loss": 0.284,
|
| 699 |
+
"rewards/accuracies": 0.925000011920929,
|
| 700 |
+
"rewards/chosen": 1.0949517488479614,
|
| 701 |
+
"rewards/margins": 1.1476168632507324,
|
| 702 |
+
"rewards/rejected": -0.05266512185335159,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.8972448110580444,
|
| 709 |
+
"logits/rejected": -1.9013574123382568,
|
| 710 |
+
"logps/chosen": -33.069793701171875,
|
| 711 |
+
"logps/rejected": -32.7395133972168,
|
| 712 |
+
"loss": 0.263,
|
| 713 |
+
"rewards/accuracies": 0.9375,
|
| 714 |
+
"rewards/chosen": 1.156842589378357,
|
| 715 |
+
"rewards/margins": 1.2775455713272095,
|
| 716 |
+
"rewards/rejected": -0.12070278078317642,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.9023644924163818,
|
| 723 |
+
"logits/rejected": -1.8966976404190063,
|
| 724 |
+
"logps/chosen": -33.858001708984375,
|
| 725 |
+
"logps/rejected": -35.19942092895508,
|
| 726 |
+
"loss": 0.2537,
|
| 727 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 728 |
+
"rewards/chosen": 1.249218225479126,
|
| 729 |
+
"rewards/margins": 1.4494789838790894,
|
| 730 |
+
"rewards/rejected": -0.20026080310344696,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.9303083419799805,
|
| 737 |
+
"logits/rejected": -1.9299156665802002,
|
| 738 |
+
"logps/chosen": -32.68412399291992,
|
| 739 |
+
"logps/rejected": -34.166542053222656,
|
| 740 |
+
"loss": 0.2809,
|
| 741 |
+
"rewards/accuracies": 0.887499988079071,
|
| 742 |
+
"rewards/chosen": 1.1873619556427002,
|
| 743 |
+
"rewards/margins": 1.1722075939178467,
|
| 744 |
+
"rewards/rejected": 0.01515440084040165,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -1.97918701171875,
|
| 751 |
+
"logits/rejected": -1.960694670677185,
|
| 752 |
+
"logps/chosen": -32.545894622802734,
|
| 753 |
+
"logps/rejected": -33.19091033935547,
|
| 754 |
+
"loss": 0.2501,
|
| 755 |
+
"rewards/accuracies": 0.9375,
|
| 756 |
+
"rewards/chosen": 1.291896104812622,
|
| 757 |
+
"rewards/margins": 1.3886420726776123,
|
| 758 |
+
"rewards/rejected": -0.09674612432718277,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.9456336498260498,
|
| 765 |
+
"logits/rejected": -1.9451408386230469,
|
| 766 |
+
"logps/chosen": -30.120983123779297,
|
| 767 |
+
"logps/rejected": -31.22079849243164,
|
| 768 |
+
"loss": 0.2773,
|
| 769 |
+
"rewards/accuracies": 0.875,
|
| 770 |
+
"rewards/chosen": 1.1922577619552612,
|
| 771 |
+
"rewards/margins": 1.1720596551895142,
|
| 772 |
+
"rewards/rejected": 0.020198041573166847,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.9087337255477905,
|
| 779 |
+
"logits/rejected": -1.9019416570663452,
|
| 780 |
+
"logps/chosen": -30.988162994384766,
|
| 781 |
+
"logps/rejected": -34.79794692993164,
|
| 782 |
+
"loss": 0.2459,
|
| 783 |
+
"rewards/accuracies": 0.875,
|
| 784 |
+
"rewards/chosen": 1.345572829246521,
|
| 785 |
+
"rewards/margins": 1.4825364351272583,
|
| 786 |
+
"rewards/rejected": -0.13696344196796417,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.1905555725097656,
|
| 792 |
+
"eval_logits/rejected": -2.1857681274414062,
|
| 793 |
+
"eval_logps/chosen": -33.6363525390625,
|
| 794 |
+
"eval_logps/rejected": -37.182701110839844,
|
| 795 |
+
"eval_loss": 0.4884692430496216,
|
| 796 |
+
"eval_rewards/accuracies": 0.5622923374176025,
|
| 797 |
+
"eval_rewards/chosen": 0.3185593783855438,
|
| 798 |
+
"eval_rewards/margins": 0.05142458155751228,
|
| 799 |
+
"eval_rewards/rejected": 0.26713478565216064,
|
| 800 |
+
"eval_runtime": 145.9464,
|
| 801 |
+
"eval_samples_per_second": 2.35,
|
| 802 |
+
"eval_steps_per_second": 0.295,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 1.32,
|
| 807 |
+
"learning_rate": 4.22929424333435e-06,
|
| 808 |
+
"logits/chosen": -1.903969168663025,
|
| 809 |
+
"logits/rejected": -1.9078088998794556,
|
| 810 |
+
"logps/chosen": -27.756145477294922,
|
| 811 |
+
"logps/rejected": -33.003501892089844,
|
| 812 |
+
"loss": 0.2721,
|
| 813 |
+
"rewards/accuracies": 0.887499988079071,
|
| 814 |
+
"rewards/chosen": 1.1349422931671143,
|
| 815 |
+
"rewards/margins": 1.2503130435943604,
|
| 816 |
+
"rewards/rejected": -0.11537061631679535,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.9105024337768555,
|
| 823 |
+
"logits/rejected": -1.9210312366485596,
|
| 824 |
+
"logps/chosen": -31.711151123046875,
|
| 825 |
+
"logps/rejected": -30.852153778076172,
|
| 826 |
+
"loss": 0.2636,
|
| 827 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 828 |
+
"rewards/chosen": 1.240944743156433,
|
| 829 |
+
"rewards/margins": 1.4013227224349976,
|
| 830 |
+
"rewards/rejected": -0.1603778600692749,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.8562660217285156,
|
| 837 |
+
"logits/rejected": -1.8539501428604126,
|
| 838 |
+
"logps/chosen": -30.094646453857422,
|
| 839 |
+
"logps/rejected": -30.31314468383789,
|
| 840 |
+
"loss": 0.2542,
|
| 841 |
+
"rewards/accuracies": 0.887499988079071,
|
| 842 |
+
"rewards/chosen": 1.2579472064971924,
|
| 843 |
+
"rewards/margins": 1.3670498132705688,
|
| 844 |
+
"rewards/rejected": -0.10910268872976303,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.9388301372528076,
|
| 851 |
+
"logits/rejected": -1.931828260421753,
|
| 852 |
+
"logps/chosen": -32.52373504638672,
|
| 853 |
+
"logps/rejected": -33.19480514526367,
|
| 854 |
+
"loss": 0.234,
|
| 855 |
+
"rewards/accuracies": 0.925000011920929,
|
| 856 |
+
"rewards/chosen": 1.3672724962234497,
|
| 857 |
+
"rewards/margins": 1.5091286897659302,
|
| 858 |
+
"rewards/rejected": -0.14185625314712524,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.9054105281829834,
|
| 865 |
+
"logits/rejected": -1.9118045568466187,
|
| 866 |
+
"logps/chosen": -30.23921775817871,
|
| 867 |
+
"logps/rejected": -32.79788589477539,
|
| 868 |
+
"loss": 0.2692,
|
| 869 |
+
"rewards/accuracies": 0.875,
|
| 870 |
+
"rewards/chosen": 1.211173176765442,
|
| 871 |
+
"rewards/margins": 1.2424962520599365,
|
| 872 |
+
"rewards/rejected": -0.03132311627268791,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -1.9779608249664307,
|
| 879 |
+
"logits/rejected": -1.9552392959594727,
|
| 880 |
+
"logps/chosen": -29.897476196289062,
|
| 881 |
+
"logps/rejected": -33.047569274902344,
|
| 882 |
+
"loss": 0.2672,
|
| 883 |
+
"rewards/accuracies": 0.887499988079071,
|
| 884 |
+
"rewards/chosen": 1.2320712804794312,
|
| 885 |
+
"rewards/margins": 1.273280143737793,
|
| 886 |
+
"rewards/rejected": -0.041208960115909576,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.9123926162719727,
|
| 893 |
+
"logits/rejected": -1.9074643850326538,
|
| 894 |
+
"logps/chosen": -30.69790267944336,
|
| 895 |
+
"logps/rejected": -32.02743911743164,
|
| 896 |
+
"loss": 0.2612,
|
| 897 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 898 |
+
"rewards/chosen": 1.3750633001327515,
|
| 899 |
+
"rewards/margins": 1.3002793788909912,
|
| 900 |
+
"rewards/rejected": 0.07478378713130951,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -1.9546325206756592,
|
| 907 |
+
"logits/rejected": -1.956498384475708,
|
| 908 |
+
"logps/chosen": -32.70453643798828,
|
| 909 |
+
"logps/rejected": -35.04417037963867,
|
| 910 |
+
"loss": 0.2395,
|
| 911 |
+
"rewards/accuracies": 0.9375,
|
| 912 |
+
"rewards/chosen": 1.337005853652954,
|
| 913 |
+
"rewards/margins": 1.5316107273101807,
|
| 914 |
+
"rewards/rejected": -0.19460472464561462,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.9385898113250732,
|
| 921 |
+
"logits/rejected": -1.9379469156265259,
|
| 922 |
+
"logps/chosen": -28.891321182250977,
|
| 923 |
+
"logps/rejected": -31.504505157470703,
|
| 924 |
+
"loss": 0.2493,
|
| 925 |
+
"rewards/accuracies": 0.887499988079071,
|
| 926 |
+
"rewards/chosen": 1.3605870008468628,
|
| 927 |
+
"rewards/margins": 1.3948794603347778,
|
| 928 |
+
"rewards/rejected": -0.034292541444301605,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -1.9496517181396484,
|
| 935 |
+
"logits/rejected": -1.9475892782211304,
|
| 936 |
+
"logps/chosen": -31.307537078857422,
|
| 937 |
+
"logps/rejected": -32.269371032714844,
|
| 938 |
+
"loss": 0.2639,
|
| 939 |
+
"rewards/accuracies": 0.949999988079071,
|
| 940 |
+
"rewards/chosen": 1.3525952100753784,
|
| 941 |
+
"rewards/margins": 1.3281266689300537,
|
| 942 |
+
"rewards/rejected": 0.024468522518873215,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.183182716369629,
|
| 948 |
+
"eval_logits/rejected": -2.178420066833496,
|
| 949 |
+
"eval_logps/chosen": -33.58167266845703,
|
| 950 |
+
"eval_logps/rejected": -37.20375442504883,
|
| 951 |
+
"eval_loss": 0.47496160864830017,
|
| 952 |
+
"eval_rewards/accuracies": 0.5855481624603271,
|
| 953 |
+
"eval_rewards/chosen": 0.3623029887676239,
|
| 954 |
+
"eval_rewards/margins": 0.11201038211584091,
|
| 955 |
+
"eval_rewards/rejected": 0.2502925992012024,
|
| 956 |
+
"eval_runtime": 145.9651,
|
| 957 |
+
"eval_samples_per_second": 2.35,
|
| 958 |
+
"eval_steps_per_second": 0.295,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 1.58,
|
| 963 |
+
"learning_rate": 3.7793282895240927e-06,
|
| 964 |
+
"logits/chosen": -1.9928886890411377,
|
| 965 |
+
"logits/rejected": -1.9994275569915771,
|
| 966 |
+
"logps/chosen": -30.57297706604004,
|
| 967 |
+
"logps/rejected": -32.0395622253418,
|
| 968 |
+
"loss": 0.2611,
|
| 969 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 970 |
+
"rewards/chosen": 1.3315242528915405,
|
| 971 |
+
"rewards/margins": 1.3126256465911865,
|
| 972 |
+
"rewards/rejected": 0.01889852061867714,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -1.9484421014785767,
|
| 979 |
+
"logits/rejected": -1.946126937866211,
|
| 980 |
+
"logps/chosen": -32.853248596191406,
|
| 981 |
+
"logps/rejected": -30.977664947509766,
|
| 982 |
+
"loss": 0.2377,
|
| 983 |
+
"rewards/accuracies": 0.887499988079071,
|
| 984 |
+
"rewards/chosen": 1.5221078395843506,
|
| 985 |
+
"rewards/margins": 1.5158333778381348,
|
| 986 |
+
"rewards/rejected": 0.0062743364833295345,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.8969027996063232,
|
| 993 |
+
"logits/rejected": -1.8902422189712524,
|
| 994 |
+
"logps/chosen": -33.44438171386719,
|
| 995 |
+
"logps/rejected": -32.37852478027344,
|
| 996 |
+
"loss": 0.2252,
|
| 997 |
+
"rewards/accuracies": 0.9375,
|
| 998 |
+
"rewards/chosen": 1.6158740520477295,
|
| 999 |
+
"rewards/margins": 1.6329265832901,
|
| 1000 |
+
"rewards/rejected": -0.017052406445145607,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.9018688201904297,
|
| 1007 |
+
"logits/rejected": -1.9080841541290283,
|
| 1008 |
+
"logps/chosen": -32.16619110107422,
|
| 1009 |
+
"logps/rejected": -33.13113021850586,
|
| 1010 |
+
"loss": 0.2565,
|
| 1011 |
+
"rewards/accuracies": 0.875,
|
| 1012 |
+
"rewards/chosen": 1.450488567352295,
|
| 1013 |
+
"rewards/margins": 1.3631832599639893,
|
| 1014 |
+
"rewards/rejected": 0.08730525523424149,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -1.98688542842865,
|
| 1021 |
+
"logits/rejected": -2.0009474754333496,
|
| 1022 |
+
"logps/chosen": -30.202587127685547,
|
| 1023 |
+
"logps/rejected": -31.93996238708496,
|
| 1024 |
+
"loss": 0.2613,
|
| 1025 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1026 |
+
"rewards/chosen": 1.3199015855789185,
|
| 1027 |
+
"rewards/margins": 1.2737209796905518,
|
| 1028 |
+
"rewards/rejected": 0.046180594712495804,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.861903429031372,
|
| 1035 |
+
"logits/rejected": -1.8589446544647217,
|
| 1036 |
+
"logps/chosen": -31.801767349243164,
|
| 1037 |
+
"logps/rejected": -34.892234802246094,
|
| 1038 |
+
"loss": 0.1937,
|
| 1039 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1040 |
+
"rewards/chosen": 1.6969530582427979,
|
| 1041 |
+
"rewards/margins": 1.8505403995513916,
|
| 1042 |
+
"rewards/rejected": -0.15358731150627136,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -1.9529634714126587,
|
| 1049 |
+
"logits/rejected": -1.9532558917999268,
|
| 1050 |
+
"logps/chosen": -29.594532012939453,
|
| 1051 |
+
"logps/rejected": -33.946571350097656,
|
| 1052 |
+
"loss": 0.2422,
|
| 1053 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1054 |
+
"rewards/chosen": 1.3528481721878052,
|
| 1055 |
+
"rewards/margins": 1.449000597000122,
|
| 1056 |
+
"rewards/rejected": -0.09615238755941391,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.9031528234481812,
|
| 1063 |
+
"logits/rejected": -1.9001739025115967,
|
| 1064 |
+
"logps/chosen": -29.16634178161621,
|
| 1065 |
+
"logps/rejected": -32.86220169067383,
|
| 1066 |
+
"loss": 0.2577,
|
| 1067 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1068 |
+
"rewards/chosen": 1.3648045063018799,
|
| 1069 |
+
"rewards/margins": 1.40032160282135,
|
| 1070 |
+
"rewards/rejected": -0.03551710769534111,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -1.9385061264038086,
|
| 1077 |
+
"logits/rejected": -1.9385082721710205,
|
| 1078 |
+
"logps/chosen": -27.996692657470703,
|
| 1079 |
+
"logps/rejected": -30.74009132385254,
|
| 1080 |
+
"loss": 0.2085,
|
| 1081 |
+
"rewards/accuracies": 0.9375,
|
| 1082 |
+
"rewards/chosen": 1.7331215143203735,
|
| 1083 |
+
"rewards/margins": 1.7887967824935913,
|
| 1084 |
+
"rewards/rejected": -0.05567514896392822,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.8473536968231201,
|
| 1091 |
+
"logits/rejected": -1.850669503211975,
|
| 1092 |
+
"logps/chosen": -32.4169921875,
|
| 1093 |
+
"logps/rejected": -31.4881649017334,
|
| 1094 |
+
"loss": 0.2437,
|
| 1095 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1096 |
+
"rewards/chosen": 1.6491864919662476,
|
| 1097 |
+
"rewards/margins": 1.6679694652557373,
|
| 1098 |
+
"rewards/rejected": -0.01878293789923191,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.178642988204956,
|
| 1104 |
+
"eval_logits/rejected": -2.173909902572632,
|
| 1105 |
+
"eval_logps/chosen": -33.59923553466797,
|
| 1106 |
+
"eval_logps/rejected": -37.22935485839844,
|
| 1107 |
+
"eval_loss": 0.4741733968257904,
|
| 1108 |
+
"eval_rewards/accuracies": 0.5747508406639099,
|
| 1109 |
+
"eval_rewards/chosen": 0.34825241565704346,
|
| 1110 |
+
"eval_rewards/margins": 0.1184391975402832,
|
| 1111 |
+
"eval_rewards/rejected": 0.22981324791908264,
|
| 1112 |
+
"eval_runtime": 145.9455,
|
| 1113 |
+
"eval_samples_per_second": 2.35,
|
| 1114 |
+
"eval_steps_per_second": 0.295,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 1.84,
|
| 1119 |
+
"learning_rate": 3.2639145321045933e-06,
|
| 1120 |
+
"logits/chosen": -1.9303743839263916,
|
| 1121 |
+
"logits/rejected": -1.9213898181915283,
|
| 1122 |
+
"logps/chosen": -34.50909423828125,
|
| 1123 |
+
"logps/rejected": -31.8797550201416,
|
| 1124 |
+
"loss": 0.2249,
|
| 1125 |
+
"rewards/accuracies": 0.9375,
|
| 1126 |
+
"rewards/chosen": 1.602341890335083,
|
| 1127 |
+
"rewards/margins": 1.5814943313598633,
|
| 1128 |
+
"rewards/rejected": 0.02084733545780182,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -1.9459857940673828,
|
| 1135 |
+
"logits/rejected": -1.9470113515853882,
|
| 1136 |
+
"logps/chosen": -34.359092712402344,
|
| 1137 |
+
"logps/rejected": -33.046634674072266,
|
| 1138 |
+
"loss": 0.2173,
|
| 1139 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1140 |
+
"rewards/chosen": 1.6610389947891235,
|
| 1141 |
+
"rewards/margins": 1.6375271081924438,
|
| 1142 |
+
"rewards/rejected": 0.02351185865700245,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -1.962262511253357,
|
| 1149 |
+
"logits/rejected": -1.9598932266235352,
|
| 1150 |
+
"logps/chosen": -30.564239501953125,
|
| 1151 |
+
"logps/rejected": -32.997047424316406,
|
| 1152 |
+
"loss": 0.23,
|
| 1153 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1154 |
+
"rewards/chosen": 1.5992707014083862,
|
| 1155 |
+
"rewards/margins": 1.5993889570236206,
|
| 1156 |
+
"rewards/rejected": -0.00011831596202682704,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -1.9474010467529297,
|
| 1163 |
+
"logits/rejected": -1.952619194984436,
|
| 1164 |
+
"logps/chosen": -29.694049835205078,
|
| 1165 |
+
"logps/rejected": -33.282554626464844,
|
| 1166 |
+
"loss": 0.2069,
|
| 1167 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1168 |
+
"rewards/chosen": 1.6733427047729492,
|
| 1169 |
+
"rewards/margins": 1.7755340337753296,
|
| 1170 |
+
"rewards/rejected": -0.10219136625528336,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.892179250717163,
|
| 1177 |
+
"logits/rejected": -1.8910681009292603,
|
| 1178 |
+
"logps/chosen": -31.141448974609375,
|
| 1179 |
+
"logps/rejected": -34.69000244140625,
|
| 1180 |
+
"loss": 0.24,
|
| 1181 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1182 |
+
"rewards/chosen": 1.5868273973464966,
|
| 1183 |
+
"rewards/margins": 1.5900630950927734,
|
| 1184 |
+
"rewards/rejected": -0.0032357454765588045,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.7668031454086304,
|
| 1191 |
+
"logits/rejected": -1.7617915868759155,
|
| 1192 |
+
"logps/chosen": -33.82783889770508,
|
| 1193 |
+
"logps/rejected": -35.33191680908203,
|
| 1194 |
+
"loss": 0.2193,
|
| 1195 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1196 |
+
"rewards/chosen": 1.7364822626113892,
|
| 1197 |
+
"rewards/margins": 1.8352861404418945,
|
| 1198 |
+
"rewards/rejected": -0.0988037958741188,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.9073164463043213,
|
| 1205 |
+
"logits/rejected": -1.9110107421875,
|
| 1206 |
+
"logps/chosen": -33.157711029052734,
|
| 1207 |
+
"logps/rejected": -33.576969146728516,
|
| 1208 |
+
"loss": 0.2221,
|
| 1209 |
+
"rewards/accuracies": 0.8708333969116211,
|
| 1210 |
+
"rewards/chosen": 1.6447594165802002,
|
| 1211 |
+
"rewards/margins": 1.6304632425308228,
|
| 1212 |
+
"rewards/rejected": 0.014296052046120167,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.8515024185180664,
|
| 1219 |
+
"logits/rejected": -1.8497060537338257,
|
| 1220 |
+
"logps/chosen": -31.297021865844727,
|
| 1221 |
+
"logps/rejected": -35.01959991455078,
|
| 1222 |
+
"loss": 0.1312,
|
| 1223 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1224 |
+
"rewards/chosen": 2.0836594104766846,
|
| 1225 |
+
"rewards/margins": 2.575618028640747,
|
| 1226 |
+
"rewards/rejected": -0.4919588565826416,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.899592638015747,
|
| 1233 |
+
"logits/rejected": -1.8983027935028076,
|
| 1234 |
+
"logps/chosen": -30.751636505126953,
|
| 1235 |
+
"logps/rejected": -33.40520477294922,
|
| 1236 |
+
"loss": 0.1431,
|
| 1237 |
+
"rewards/accuracies": 0.9375,
|
| 1238 |
+
"rewards/chosen": 2.0992822647094727,
|
| 1239 |
+
"rewards/margins": 2.354191541671753,
|
| 1240 |
+
"rewards/rejected": -0.25490933656692505,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.83371901512146,
|
| 1247 |
+
"logits/rejected": -1.8279550075531006,
|
| 1248 |
+
"logps/chosen": -31.871252059936523,
|
| 1249 |
+
"logps/rejected": -31.73343849182129,
|
| 1250 |
+
"loss": 0.1567,
|
| 1251 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1252 |
+
"rewards/chosen": 2.0094635486602783,
|
| 1253 |
+
"rewards/margins": 2.224806547164917,
|
| 1254 |
+
"rewards/rejected": -0.21534299850463867,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.1802878379821777,
|
| 1260 |
+
"eval_logits/rejected": -2.1755456924438477,
|
| 1261 |
+
"eval_logps/chosen": -33.54963302612305,
|
| 1262 |
+
"eval_logps/rejected": -37.20660400390625,
|
| 1263 |
+
"eval_loss": 0.4695200026035309,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5826411843299866,
|
| 1265 |
+
"eval_rewards/chosen": 0.3879339396953583,
|
| 1266 |
+
"eval_rewards/margins": 0.1399237960577011,
|
| 1267 |
+
"eval_rewards/rejected": 0.24801018834114075,
|
| 1268 |
+
"eval_runtime": 145.8879,
|
| 1269 |
+
"eval_samples_per_second": 2.351,
|
| 1270 |
+
"eval_steps_per_second": 0.295,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.1,
|
| 1275 |
+
"learning_rate": 2.7094204786572254e-06,
|
| 1276 |
+
"logits/chosen": -1.9300823211669922,
|
| 1277 |
+
"logits/rejected": -1.9375168085098267,
|
| 1278 |
+
"logps/chosen": -29.62691307067871,
|
| 1279 |
+
"logps/rejected": -34.63849639892578,
|
| 1280 |
+
"loss": 0.1613,
|
| 1281 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1282 |
+
"rewards/chosen": 2.0010743141174316,
|
| 1283 |
+
"rewards/margins": 2.3072001934051514,
|
| 1284 |
+
"rewards/rejected": -0.3061259686946869,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.9036821126937866,
|
| 1291 |
+
"logits/rejected": -1.9064573049545288,
|
| 1292 |
+
"logps/chosen": -30.34005355834961,
|
| 1293 |
+
"logps/rejected": -33.30119705200195,
|
| 1294 |
+
"loss": 0.1999,
|
| 1295 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1296 |
+
"rewards/chosen": 1.784820556640625,
|
| 1297 |
+
"rewards/margins": 1.8419349193572998,
|
| 1298 |
+
"rewards/rejected": -0.05711423233151436,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.805069923400879,
|
| 1305 |
+
"logits/rejected": -1.8091821670532227,
|
| 1306 |
+
"logps/chosen": -30.127094268798828,
|
| 1307 |
+
"logps/rejected": -35.30133056640625,
|
| 1308 |
+
"loss": 0.1557,
|
| 1309 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1310 |
+
"rewards/chosen": 1.9714915752410889,
|
| 1311 |
+
"rewards/margins": 2.2844126224517822,
|
| 1312 |
+
"rewards/rejected": -0.31292054057121277,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.8699356317520142,
|
| 1319 |
+
"logits/rejected": -1.865962028503418,
|
| 1320 |
+
"logps/chosen": -30.581180572509766,
|
| 1321 |
+
"logps/rejected": -33.02351379394531,
|
| 1322 |
+
"loss": 0.1821,
|
| 1323 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1324 |
+
"rewards/chosen": 1.9099775552749634,
|
| 1325 |
+
"rewards/margins": 1.9330689907073975,
|
| 1326 |
+
"rewards/rejected": -0.02309143915772438,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -1.9513721466064453,
|
| 1333 |
+
"logits/rejected": -1.9533140659332275,
|
| 1334 |
+
"logps/chosen": -30.263479232788086,
|
| 1335 |
+
"logps/rejected": -36.080474853515625,
|
| 1336 |
+
"loss": 0.1402,
|
| 1337 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1338 |
+
"rewards/chosen": 2.0094857215881348,
|
| 1339 |
+
"rewards/margins": 2.344318389892578,
|
| 1340 |
+
"rewards/rejected": -0.3348326086997986,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.7577623128890991,
|
| 1347 |
+
"logits/rejected": -1.7520900964736938,
|
| 1348 |
+
"logps/chosen": -33.64806365966797,
|
| 1349 |
+
"logps/rejected": -33.127098083496094,
|
| 1350 |
+
"loss": 0.157,
|
| 1351 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1352 |
+
"rewards/chosen": 2.106886625289917,
|
| 1353 |
+
"rewards/margins": 2.2950429916381836,
|
| 1354 |
+
"rewards/rejected": -0.18815645575523376,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.8190982341766357,
|
| 1361 |
+
"logits/rejected": -1.8191255331039429,
|
| 1362 |
+
"logps/chosen": -33.16907501220703,
|
| 1363 |
+
"logps/rejected": -35.244049072265625,
|
| 1364 |
+
"loss": 0.1428,
|
| 1365 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1366 |
+
"rewards/chosen": 2.138986110687256,
|
| 1367 |
+
"rewards/margins": 2.457213878631592,
|
| 1368 |
+
"rewards/rejected": -0.31822749972343445,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.867693543434143,
|
| 1375 |
+
"logits/rejected": -1.8564653396606445,
|
| 1376 |
+
"logps/chosen": -31.62196922302246,
|
| 1377 |
+
"logps/rejected": -33.34123992919922,
|
| 1378 |
+
"loss": 0.1484,
|
| 1379 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1380 |
+
"rewards/chosen": 1.9876899719238281,
|
| 1381 |
+
"rewards/margins": 2.3230578899383545,
|
| 1382 |
+
"rewards/rejected": -0.3353681266307831,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -1.9699180126190186,
|
| 1389 |
+
"logits/rejected": -1.9743999242782593,
|
| 1390 |
+
"logps/chosen": -30.770462036132812,
|
| 1391 |
+
"logps/rejected": -32.943992614746094,
|
| 1392 |
+
"loss": 0.1449,
|
| 1393 |
+
"rewards/accuracies": 0.9375,
|
| 1394 |
+
"rewards/chosen": 2.0475802421569824,
|
| 1395 |
+
"rewards/margins": 2.453592300415039,
|
| 1396 |
+
"rewards/rejected": -0.40601205825805664,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.8890403509140015,
|
| 1403 |
+
"logits/rejected": -1.881980538368225,
|
| 1404 |
+
"logps/chosen": -32.13774490356445,
|
| 1405 |
+
"logps/rejected": -32.89883041381836,
|
| 1406 |
+
"loss": 0.131,
|
| 1407 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1408 |
+
"rewards/chosen": 2.173067808151245,
|
| 1409 |
+
"rewards/margins": 2.5704257488250732,
|
| 1410 |
+
"rewards/rejected": -0.397358238697052,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.1704888343811035,
|
| 1416 |
+
"eval_logits/rejected": -2.165771961212158,
|
| 1417 |
+
"eval_logps/chosen": -33.59296798706055,
|
| 1418 |
+
"eval_logps/rejected": -37.2408447265625,
|
| 1419 |
+
"eval_loss": 0.4716451168060303,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1421 |
+
"eval_rewards/chosen": 0.3532681465148926,
|
| 1422 |
+
"eval_rewards/margins": 0.1326470971107483,
|
| 1423 |
+
"eval_rewards/rejected": 0.22062106430530548,
|
| 1424 |
+
"eval_runtime": 145.952,
|
| 1425 |
+
"eval_samples_per_second": 2.35,
|
| 1426 |
+
"eval_steps_per_second": 0.295,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 2.36,
|
| 1431 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 1432 |
+
"logits/chosen": -1.8803346157073975,
|
| 1433 |
+
"logits/rejected": -1.8807852268218994,
|
| 1434 |
+
"logps/chosen": -28.69057846069336,
|
| 1435 |
+
"logps/rejected": -35.19638442993164,
|
| 1436 |
+
"loss": 0.1531,
|
| 1437 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1438 |
+
"rewards/chosen": 1.9812660217285156,
|
| 1439 |
+
"rewards/margins": 2.3866193294525146,
|
| 1440 |
+
"rewards/rejected": -0.4053533673286438,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.8385648727416992,
|
| 1447 |
+
"logits/rejected": -1.8306169509887695,
|
| 1448 |
+
"logps/chosen": -29.56528091430664,
|
| 1449 |
+
"logps/rejected": -34.04883575439453,
|
| 1450 |
+
"loss": 0.1693,
|
| 1451 |
+
"rewards/accuracies": 0.9375,
|
| 1452 |
+
"rewards/chosen": 1.8189325332641602,
|
| 1453 |
+
"rewards/margins": 2.2009072303771973,
|
| 1454 |
+
"rewards/rejected": -0.3819747567176819,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -2.004300117492676,
|
| 1461 |
+
"logits/rejected": -2.004765033721924,
|
| 1462 |
+
"logps/chosen": -27.645904541015625,
|
| 1463 |
+
"logps/rejected": -33.17176818847656,
|
| 1464 |
+
"loss": 0.1901,
|
| 1465 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1466 |
+
"rewards/chosen": 1.8809070587158203,
|
| 1467 |
+
"rewards/margins": 2.126433849334717,
|
| 1468 |
+
"rewards/rejected": -0.24552695453166962,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.883927583694458,
|
| 1475 |
+
"logits/rejected": -1.893393874168396,
|
| 1476 |
+
"logps/chosen": -31.66170883178711,
|
| 1477 |
+
"logps/rejected": -33.47935104370117,
|
| 1478 |
+
"loss": 0.1308,
|
| 1479 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1480 |
+
"rewards/chosen": 2.2106852531433105,
|
| 1481 |
+
"rewards/margins": 2.602447509765625,
|
| 1482 |
+
"rewards/rejected": -0.3917620778083801,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -1.9267995357513428,
|
| 1489 |
+
"logits/rejected": -1.9306179285049438,
|
| 1490 |
+
"logps/chosen": -28.294620513916016,
|
| 1491 |
+
"logps/rejected": -32.00605010986328,
|
| 1492 |
+
"loss": 0.1628,
|
| 1493 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1494 |
+
"rewards/chosen": 1.882483720779419,
|
| 1495 |
+
"rewards/margins": 2.0984606742858887,
|
| 1496 |
+
"rewards/rejected": -0.21597695350646973,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -1.9250777959823608,
|
| 1503 |
+
"logits/rejected": -1.921263337135315,
|
| 1504 |
+
"logps/chosen": -31.393346786499023,
|
| 1505 |
+
"logps/rejected": -34.266326904296875,
|
| 1506 |
+
"loss": 0.1531,
|
| 1507 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1508 |
+
"rewards/chosen": 2.145495891571045,
|
| 1509 |
+
"rewards/margins": 2.4334449768066406,
|
| 1510 |
+
"rewards/rejected": -0.28794899582862854,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.8407471179962158,
|
| 1517 |
+
"logits/rejected": -1.84335458278656,
|
| 1518 |
+
"logps/chosen": -30.177082061767578,
|
| 1519 |
+
"logps/rejected": -35.5606803894043,
|
| 1520 |
+
"loss": 0.15,
|
| 1521 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1522 |
+
"rewards/chosen": 2.1085331439971924,
|
| 1523 |
+
"rewards/margins": 2.411801815032959,
|
| 1524 |
+
"rewards/rejected": -0.3032683730125427,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.8759149312973022,
|
| 1531 |
+
"logits/rejected": -1.8695703744888306,
|
| 1532 |
+
"logps/chosen": -32.879905700683594,
|
| 1533 |
+
"logps/rejected": -36.688812255859375,
|
| 1534 |
+
"loss": 0.1591,
|
| 1535 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1536 |
+
"rewards/chosen": 2.020131826400757,
|
| 1537 |
+
"rewards/margins": 2.3616156578063965,
|
| 1538 |
+
"rewards/rejected": -0.34148353338241577,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.8502181768417358,
|
| 1545 |
+
"logits/rejected": -1.845171332359314,
|
| 1546 |
+
"logps/chosen": -28.556514739990234,
|
| 1547 |
+
"logps/rejected": -36.38850784301758,
|
| 1548 |
+
"loss": 0.167,
|
| 1549 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1550 |
+
"rewards/chosen": 1.9571492671966553,
|
| 1551 |
+
"rewards/margins": 2.2760462760925293,
|
| 1552 |
+
"rewards/rejected": -0.31889697909355164,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.8279926776885986,
|
| 1559 |
+
"logits/rejected": -1.8174269199371338,
|
| 1560 |
+
"logps/chosen": -29.500579833984375,
|
| 1561 |
+
"logps/rejected": -29.74460220336914,
|
| 1562 |
+
"loss": 0.1784,
|
| 1563 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1564 |
+
"rewards/chosen": 1.798485517501831,
|
| 1565 |
+
"rewards/margins": 1.9143791198730469,
|
| 1566 |
+
"rewards/rejected": -0.11589355766773224,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.169346570968628,
|
| 1572 |
+
"eval_logits/rejected": -2.1646294593811035,
|
| 1573 |
+
"eval_logps/chosen": -33.57487106323242,
|
| 1574 |
+
"eval_logps/rejected": -37.250404357910156,
|
| 1575 |
+
"eval_loss": 0.4672938883304596,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1577 |
+
"eval_rewards/chosen": 0.3677443265914917,
|
| 1578 |
+
"eval_rewards/margins": 0.15477292239665985,
|
| 1579 |
+
"eval_rewards/rejected": 0.21297141909599304,
|
| 1580 |
+
"eval_runtime": 145.9529,
|
| 1581 |
+
"eval_samples_per_second": 2.35,
|
| 1582 |
+
"eval_steps_per_second": 0.295,
|
| 1583 |
+
"step": 1000
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 2.62,
|
| 1587 |
+
"learning_rate": 1.5972066659083796e-06,
|
| 1588 |
+
"logits/chosen": -1.938071846961975,
|
| 1589 |
+
"logits/rejected": -1.9374967813491821,
|
| 1590 |
+
"logps/chosen": -29.38279151916504,
|
| 1591 |
+
"logps/rejected": -30.51906394958496,
|
| 1592 |
+
"loss": 0.1659,
|
| 1593 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1594 |
+
"rewards/chosen": 1.953122854232788,
|
| 1595 |
+
"rewards/margins": 2.1615920066833496,
|
| 1596 |
+
"rewards/rejected": -0.20846888422966003,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -1.9777361154556274,
|
| 1603 |
+
"logits/rejected": -1.9785385131835938,
|
| 1604 |
+
"logps/chosen": -29.455398559570312,
|
| 1605 |
+
"logps/rejected": -31.60211753845215,
|
| 1606 |
+
"loss": 0.1457,
|
| 1607 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1608 |
+
"rewards/chosen": 1.9905421733856201,
|
| 1609 |
+
"rewards/margins": 2.3395895957946777,
|
| 1610 |
+
"rewards/rejected": -0.3490474820137024,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -1.948656439781189,
|
| 1617 |
+
"logits/rejected": -1.951843500137329,
|
| 1618 |
+
"logps/chosen": -29.8292236328125,
|
| 1619 |
+
"logps/rejected": -28.163440704345703,
|
| 1620 |
+
"loss": 0.1712,
|
| 1621 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1622 |
+
"rewards/chosen": 1.960677146911621,
|
| 1623 |
+
"rewards/margins": 2.1044540405273438,
|
| 1624 |
+
"rewards/rejected": -0.14377683401107788,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -1.9582103490829468,
|
| 1631 |
+
"logits/rejected": -1.9505630731582642,
|
| 1632 |
+
"logps/chosen": -28.690776824951172,
|
| 1633 |
+
"logps/rejected": -33.91931915283203,
|
| 1634 |
+
"loss": 0.1486,
|
| 1635 |
+
"rewards/accuracies": 0.9375,
|
| 1636 |
+
"rewards/chosen": 1.9896939992904663,
|
| 1637 |
+
"rewards/margins": 2.3249640464782715,
|
| 1638 |
+
"rewards/rejected": -0.33527034521102905,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.8943002223968506,
|
| 1645 |
+
"logits/rejected": -1.905225157737732,
|
| 1646 |
+
"logps/chosen": -30.873615264892578,
|
| 1647 |
+
"logps/rejected": -32.47500228881836,
|
| 1648 |
+
"loss": 0.1427,
|
| 1649 |
+
"rewards/accuracies": 0.9375,
|
| 1650 |
+
"rewards/chosen": 2.143594264984131,
|
| 1651 |
+
"rewards/margins": 2.3275885581970215,
|
| 1652 |
+
"rewards/rejected": -0.18399406969547272,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -1.9732818603515625,
|
| 1659 |
+
"logits/rejected": -1.9640705585479736,
|
| 1660 |
+
"logps/chosen": -30.988611221313477,
|
| 1661 |
+
"logps/rejected": -36.728416442871094,
|
| 1662 |
+
"loss": 0.171,
|
| 1663 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1664 |
+
"rewards/chosen": 2.009733200073242,
|
| 1665 |
+
"rewards/margins": 2.327359199523926,
|
| 1666 |
+
"rewards/rejected": -0.31762614846229553,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.8804638385772705,
|
| 1673 |
+
"logits/rejected": -1.8824098110198975,
|
| 1674 |
+
"logps/chosen": -33.24834442138672,
|
| 1675 |
+
"logps/rejected": -33.089820861816406,
|
| 1676 |
+
"loss": 0.1391,
|
| 1677 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1678 |
+
"rewards/chosen": 2.1927542686462402,
|
| 1679 |
+
"rewards/margins": 2.506317615509033,
|
| 1680 |
+
"rewards/rejected": -0.31356340646743774,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -2.0114994049072266,
|
| 1687 |
+
"logits/rejected": -2.012619972229004,
|
| 1688 |
+
"logps/chosen": -30.447467803955078,
|
| 1689 |
+
"logps/rejected": -34.64774703979492,
|
| 1690 |
+
"loss": 0.1376,
|
| 1691 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1692 |
+
"rewards/chosen": 2.088136911392212,
|
| 1693 |
+
"rewards/margins": 2.513786792755127,
|
| 1694 |
+
"rewards/rejected": -0.42564988136291504,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -1.9806941747665405,
|
| 1701 |
+
"logits/rejected": -1.9849655628204346,
|
| 1702 |
+
"logps/chosen": -28.488088607788086,
|
| 1703 |
+
"logps/rejected": -33.84292984008789,
|
| 1704 |
+
"loss": 0.1691,
|
| 1705 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1706 |
+
"rewards/chosen": 1.9241825342178345,
|
| 1707 |
+
"rewards/margins": 2.160259485244751,
|
| 1708 |
+
"rewards/rejected": -0.23607663810253143,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -1.9700733423233032,
|
| 1715 |
+
"logits/rejected": -1.9752953052520752,
|
| 1716 |
+
"logps/chosen": -31.235626220703125,
|
| 1717 |
+
"logps/rejected": -30.409204483032227,
|
| 1718 |
+
"loss": 0.1956,
|
| 1719 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1720 |
+
"rewards/chosen": 1.8667011260986328,
|
| 1721 |
+
"rewards/margins": 1.8727245330810547,
|
| 1722 |
+
"rewards/rejected": -0.0060233683325350285,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.1669461727142334,
|
| 1728 |
+
"eval_logits/rejected": -2.162247896194458,
|
| 1729 |
+
"eval_logps/chosen": -33.58707809448242,
|
| 1730 |
+
"eval_logps/rejected": -37.24416732788086,
|
| 1731 |
+
"eval_loss": 0.47056397795677185,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1733 |
+
"eval_rewards/chosen": 0.3579804301261902,
|
| 1734 |
+
"eval_rewards/margins": 0.14001740515232086,
|
| 1735 |
+
"eval_rewards/rejected": 0.21796302497386932,
|
| 1736 |
+
"eval_runtime": 145.9563,
|
| 1737 |
+
"eval_samples_per_second": 2.35,
|
| 1738 |
+
"eval_steps_per_second": 0.295,
|
| 1739 |
+
"step": 1100
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 2.88,
|
| 1743 |
+
"learning_rate": 1.0963854773524548e-06,
|
| 1744 |
+
"logits/chosen": -1.9610811471939087,
|
| 1745 |
+
"logits/rejected": -1.9610563516616821,
|
| 1746 |
+
"logps/chosen": -30.302623748779297,
|
| 1747 |
+
"logps/rejected": -31.24424171447754,
|
| 1748 |
+
"loss": 0.1495,
|
| 1749 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1750 |
+
"rewards/chosen": 2.1100215911865234,
|
| 1751 |
+
"rewards/margins": 2.2926793098449707,
|
| 1752 |
+
"rewards/rejected": -0.18265783786773682,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.8055435419082642,
|
| 1759 |
+
"logits/rejected": -1.8071558475494385,
|
| 1760 |
+
"logps/chosen": -31.964313507080078,
|
| 1761 |
+
"logps/rejected": -33.263916015625,
|
| 1762 |
+
"loss": 0.1722,
|
| 1763 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1764 |
+
"rewards/chosen": 2.08052921295166,
|
| 1765 |
+
"rewards/margins": 2.258647918701172,
|
| 1766 |
+
"rewards/rejected": -0.17811879515647888,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.8441641330718994,
|
| 1773 |
+
"logits/rejected": -1.8414533138275146,
|
| 1774 |
+
"logps/chosen": -28.680904388427734,
|
| 1775 |
+
"logps/rejected": -31.291534423828125,
|
| 1776 |
+
"loss": 0.1972,
|
| 1777 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1778 |
+
"rewards/chosen": 1.8646504878997803,
|
| 1779 |
+
"rewards/margins": 1.9652849435806274,
|
| 1780 |
+
"rewards/rejected": -0.10063423216342926,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -1.979723334312439,
|
| 1787 |
+
"logits/rejected": -1.9741865396499634,
|
| 1788 |
+
"logps/chosen": -30.334341049194336,
|
| 1789 |
+
"logps/rejected": -33.13202667236328,
|
| 1790 |
+
"loss": 0.1522,
|
| 1791 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1792 |
+
"rewards/chosen": 2.1643624305725098,
|
| 1793 |
+
"rewards/margins": 2.3067753314971924,
|
| 1794 |
+
"rewards/rejected": -0.14241299033164978,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.8336684703826904,
|
| 1801 |
+
"logits/rejected": -1.823862075805664,
|
| 1802 |
+
"logps/chosen": -30.90520668029785,
|
| 1803 |
+
"logps/rejected": -31.456531524658203,
|
| 1804 |
+
"loss": 0.1763,
|
| 1805 |
+
"rewards/accuracies": 0.9375,
|
| 1806 |
+
"rewards/chosen": 2.037224292755127,
|
| 1807 |
+
"rewards/margins": 2.1685280799865723,
|
| 1808 |
+
"rewards/rejected": -0.13130377233028412,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.9085521697998047,
|
| 1815 |
+
"logits/rejected": -1.908399224281311,
|
| 1816 |
+
"logps/chosen": -31.05959701538086,
|
| 1817 |
+
"logps/rejected": -31.56801414489746,
|
| 1818 |
+
"loss": 0.1469,
|
| 1819 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1820 |
+
"rewards/chosen": 2.09771990776062,
|
| 1821 |
+
"rewards/margins": 2.3909716606140137,
|
| 1822 |
+
"rewards/rejected": -0.2932516634464264,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -1.9817787408828735,
|
| 1829 |
+
"logits/rejected": -1.9764562845230103,
|
| 1830 |
+
"logps/chosen": -30.92386817932129,
|
| 1831 |
+
"logps/rejected": -33.19076919555664,
|
| 1832 |
+
"loss": 0.1496,
|
| 1833 |
+
"rewards/accuracies": 0.9375,
|
| 1834 |
+
"rewards/chosen": 2.0816802978515625,
|
| 1835 |
+
"rewards/margins": 2.2813220024108887,
|
| 1836 |
+
"rewards/rejected": -0.19964155554771423,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.8389742374420166,
|
| 1843 |
+
"logits/rejected": -1.8466594219207764,
|
| 1844 |
+
"logps/chosen": -30.842212677001953,
|
| 1845 |
+
"logps/rejected": -32.71555709838867,
|
| 1846 |
+
"loss": 0.1498,
|
| 1847 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1848 |
+
"rewards/chosen": 1.9926265478134155,
|
| 1849 |
+
"rewards/margins": 2.2759203910827637,
|
| 1850 |
+
"rewards/rejected": -0.28329357504844666,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -1.9125378131866455,
|
| 1857 |
+
"logits/rejected": -1.9067026376724243,
|
| 1858 |
+
"logps/chosen": -29.305688858032227,
|
| 1859 |
+
"logps/rejected": -33.628910064697266,
|
| 1860 |
+
"loss": 0.1274,
|
| 1861 |
+
"rewards/accuracies": 0.9375,
|
| 1862 |
+
"rewards/chosen": 2.294342279434204,
|
| 1863 |
+
"rewards/margins": 2.570711612701416,
|
| 1864 |
+
"rewards/rejected": -0.27636927366256714,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.8545643091201782,
|
| 1871 |
+
"logits/rejected": -1.8566617965698242,
|
| 1872 |
+
"logps/chosen": -30.835662841796875,
|
| 1873 |
+
"logps/rejected": -33.26459503173828,
|
| 1874 |
+
"loss": 0.137,
|
| 1875 |
+
"rewards/accuracies": 0.9375,
|
| 1876 |
+
"rewards/chosen": 2.1238210201263428,
|
| 1877 |
+
"rewards/margins": 2.5274574756622314,
|
| 1878 |
+
"rewards/rejected": -0.403636634349823,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.167179584503174,
|
| 1884 |
+
"eval_logits/rejected": -2.1624755859375,
|
| 1885 |
+
"eval_logps/chosen": -33.572837829589844,
|
| 1886 |
+
"eval_logps/rejected": -37.24382400512695,
|
| 1887 |
+
"eval_loss": 0.4679512083530426,
|
| 1888 |
+
"eval_rewards/accuracies": 0.6063122749328613,
|
| 1889 |
+
"eval_rewards/chosen": 0.36937177181243896,
|
| 1890 |
+
"eval_rewards/margins": 0.15113388001918793,
|
| 1891 |
+
"eval_rewards/rejected": 0.21823787689208984,
|
| 1892 |
+
"eval_runtime": 145.8851,
|
| 1893 |
+
"eval_samples_per_second": 2.351,
|
| 1894 |
+
"eval_steps_per_second": 0.295,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 3.14,
|
| 1899 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1900 |
+
"logits/chosen": -1.7856515645980835,
|
| 1901 |
+
"logits/rejected": -1.784690260887146,
|
| 1902 |
+
"logps/chosen": -33.143531799316406,
|
| 1903 |
+
"logps/rejected": -33.10202407836914,
|
| 1904 |
+
"loss": 0.1273,
|
| 1905 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1906 |
+
"rewards/chosen": 2.3271608352661133,
|
| 1907 |
+
"rewards/margins": 2.641291618347168,
|
| 1908 |
+
"rewards/rejected": -0.3141304850578308,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -1.9604772329330444,
|
| 1915 |
+
"logits/rejected": -1.9546642303466797,
|
| 1916 |
+
"logps/chosen": -28.979385375976562,
|
| 1917 |
+
"logps/rejected": -33.37938690185547,
|
| 1918 |
+
"loss": 0.1352,
|
| 1919 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1920 |
+
"rewards/chosen": 2.1794400215148926,
|
| 1921 |
+
"rewards/margins": 2.5801169872283936,
|
| 1922 |
+
"rewards/rejected": -0.40067729353904724,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.8098831176757812,
|
| 1929 |
+
"logits/rejected": -1.798370361328125,
|
| 1930 |
+
"logps/chosen": -31.50105857849121,
|
| 1931 |
+
"logps/rejected": -33.25225067138672,
|
| 1932 |
+
"loss": 0.1311,
|
| 1933 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1934 |
+
"rewards/chosen": 2.3076980113983154,
|
| 1935 |
+
"rewards/margins": 2.5944409370422363,
|
| 1936 |
+
"rewards/rejected": -0.2867427468299866,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.9010969400405884,
|
| 1943 |
+
"logits/rejected": -1.8942989110946655,
|
| 1944 |
+
"logps/chosen": -28.258697509765625,
|
| 1945 |
+
"logps/rejected": -29.34195327758789,
|
| 1946 |
+
"loss": 0.1603,
|
| 1947 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1948 |
+
"rewards/chosen": 1.9532296657562256,
|
| 1949 |
+
"rewards/margins": 2.33274507522583,
|
| 1950 |
+
"rewards/rejected": -0.3795151114463806,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -1.9308841228485107,
|
| 1957 |
+
"logits/rejected": -1.9373613595962524,
|
| 1958 |
+
"logps/chosen": -32.524845123291016,
|
| 1959 |
+
"logps/rejected": -31.774621963500977,
|
| 1960 |
+
"loss": 0.1315,
|
| 1961 |
+
"rewards/accuracies": 0.9375,
|
| 1962 |
+
"rewards/chosen": 2.2964370250701904,
|
| 1963 |
+
"rewards/margins": 2.5014843940734863,
|
| 1964 |
+
"rewards/rejected": -0.20504704117774963,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.9048112630844116,
|
| 1971 |
+
"logits/rejected": -1.9054410457611084,
|
| 1972 |
+
"logps/chosen": -30.393224716186523,
|
| 1973 |
+
"logps/rejected": -33.286048889160156,
|
| 1974 |
+
"loss": 0.1317,
|
| 1975 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1976 |
+
"rewards/chosen": 2.1082680225372314,
|
| 1977 |
+
"rewards/margins": 2.485210418701172,
|
| 1978 |
+
"rewards/rejected": -0.37694239616394043,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.846727728843689,
|
| 1985 |
+
"logits/rejected": -1.8463557958602905,
|
| 1986 |
+
"logps/chosen": -30.23922348022461,
|
| 1987 |
+
"logps/rejected": -33.92576217651367,
|
| 1988 |
+
"loss": 0.1224,
|
| 1989 |
+
"rewards/accuracies": 0.9375,
|
| 1990 |
+
"rewards/chosen": 2.2812910079956055,
|
| 1991 |
+
"rewards/margins": 2.6054136753082275,
|
| 1992 |
+
"rewards/rejected": -0.3241223692893982,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.730064034461975,
|
| 1999 |
+
"logits/rejected": -1.7253376245498657,
|
| 2000 |
+
"logps/chosen": -32.30054473876953,
|
| 2001 |
+
"logps/rejected": -31.246471405029297,
|
| 2002 |
+
"loss": 0.1154,
|
| 2003 |
+
"rewards/accuracies": 0.9375,
|
| 2004 |
+
"rewards/chosen": 2.349520206451416,
|
| 2005 |
+
"rewards/margins": 2.605508327484131,
|
| 2006 |
+
"rewards/rejected": -0.2559877932071686,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -1.9358218908309937,
|
| 2013 |
+
"logits/rejected": -1.9203758239746094,
|
| 2014 |
+
"logps/chosen": -31.7402400970459,
|
| 2015 |
+
"logps/rejected": -35.388179779052734,
|
| 2016 |
+
"loss": 0.1234,
|
| 2017 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2018 |
+
"rewards/chosen": 2.085869550704956,
|
| 2019 |
+
"rewards/margins": 2.5841598510742188,
|
| 2020 |
+
"rewards/rejected": -0.4982902407646179,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.879227638244629,
|
| 2027 |
+
"logits/rejected": -1.8840656280517578,
|
| 2028 |
+
"logps/chosen": -30.69500732421875,
|
| 2029 |
+
"logps/rejected": -36.70278549194336,
|
| 2030 |
+
"loss": 0.1211,
|
| 2031 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2032 |
+
"rewards/chosen": 2.2580931186676025,
|
| 2033 |
+
"rewards/margins": 2.7377448081970215,
|
| 2034 |
+
"rewards/rejected": -0.4796522259712219,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.1668777465820312,
|
| 2040 |
+
"eval_logits/rejected": -2.1621785163879395,
|
| 2041 |
+
"eval_logps/chosen": -33.58045959472656,
|
| 2042 |
+
"eval_logps/rejected": -37.23927688598633,
|
| 2043 |
+
"eval_loss": 0.47049835324287415,
|
| 2044 |
+
"eval_rewards/accuracies": 0.5917773842811584,
|
| 2045 |
+
"eval_rewards/chosen": 0.3632733225822449,
|
| 2046 |
+
"eval_rewards/margins": 0.14139726758003235,
|
| 2047 |
+
"eval_rewards/rejected": 0.22187604010105133,
|
| 2048 |
+
"eval_runtime": 145.9507,
|
| 2049 |
+
"eval_samples_per_second": 2.35,
|
| 2050 |
+
"eval_steps_per_second": 0.295,
|
| 2051 |
+
"step": 1300
|
| 2052 |
+
},
|
| 2053 |
+
{
|
| 2054 |
+
"epoch": 3.4,
|
| 2055 |
+
"learning_rate": 3.321087280364757e-07,
|
| 2056 |
+
"logits/chosen": -1.8520374298095703,
|
| 2057 |
+
"logits/rejected": -1.8522069454193115,
|
| 2058 |
+
"logps/chosen": -33.18568801879883,
|
| 2059 |
+
"logps/rejected": -37.46868896484375,
|
| 2060 |
+
"loss": 0.134,
|
| 2061 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2062 |
+
"rewards/chosen": 2.446664333343506,
|
| 2063 |
+
"rewards/margins": 2.7543227672576904,
|
| 2064 |
+
"rewards/rejected": -0.30765873193740845,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -1.9394071102142334,
|
| 2071 |
+
"logits/rejected": -1.9380232095718384,
|
| 2072 |
+
"logps/chosen": -29.4769344329834,
|
| 2073 |
+
"logps/rejected": -31.30813980102539,
|
| 2074 |
+
"loss": 0.159,
|
| 2075 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2076 |
+
"rewards/chosen": 2.093536853790283,
|
| 2077 |
+
"rewards/margins": 2.3773601055145264,
|
| 2078 |
+
"rewards/rejected": -0.2838229537010193,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.844186782836914,
|
| 2085 |
+
"logits/rejected": -1.8454091548919678,
|
| 2086 |
+
"logps/chosen": -31.52813720703125,
|
| 2087 |
+
"logps/rejected": -33.989559173583984,
|
| 2088 |
+
"loss": 0.1366,
|
| 2089 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2090 |
+
"rewards/chosen": 2.276893138885498,
|
| 2091 |
+
"rewards/margins": 2.549605131149292,
|
| 2092 |
+
"rewards/rejected": -0.27271172404289246,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.9051072597503662,
|
| 2099 |
+
"logits/rejected": -1.9038896560668945,
|
| 2100 |
+
"logps/chosen": -28.674198150634766,
|
| 2101 |
+
"logps/rejected": -32.49821090698242,
|
| 2102 |
+
"loss": 0.1373,
|
| 2103 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2104 |
+
"rewards/chosen": 2.01957631111145,
|
| 2105 |
+
"rewards/margins": 2.424564838409424,
|
| 2106 |
+
"rewards/rejected": -0.4049888551235199,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -1.9206342697143555,
|
| 2113 |
+
"logits/rejected": -1.9062645435333252,
|
| 2114 |
+
"logps/chosen": -31.68826675415039,
|
| 2115 |
+
"logps/rejected": -36.69408416748047,
|
| 2116 |
+
"loss": 0.1396,
|
| 2117 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2118 |
+
"rewards/chosen": 2.102721691131592,
|
| 2119 |
+
"rewards/margins": 2.4464478492736816,
|
| 2120 |
+
"rewards/rejected": -0.34372615814208984,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -1.921002984046936,
|
| 2127 |
+
"logits/rejected": -1.9334627389907837,
|
| 2128 |
+
"logps/chosen": -29.323871612548828,
|
| 2129 |
+
"logps/rejected": -32.485260009765625,
|
| 2130 |
+
"loss": 0.1368,
|
| 2131 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2132 |
+
"rewards/chosen": 2.195680856704712,
|
| 2133 |
+
"rewards/margins": 2.3920950889587402,
|
| 2134 |
+
"rewards/rejected": -0.1964142769575119,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -2.0142505168914795,
|
| 2141 |
+
"logits/rejected": -2.0135538578033447,
|
| 2142 |
+
"logps/chosen": -29.253097534179688,
|
| 2143 |
+
"logps/rejected": -36.559532165527344,
|
| 2144 |
+
"loss": 0.1185,
|
| 2145 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2146 |
+
"rewards/chosen": 2.2218658924102783,
|
| 2147 |
+
"rewards/margins": 2.8047871589660645,
|
| 2148 |
+
"rewards/rejected": -0.5829211473464966,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.8797496557235718,
|
| 2155 |
+
"logits/rejected": -1.8874984979629517,
|
| 2156 |
+
"logps/chosen": -31.32064437866211,
|
| 2157 |
+
"logps/rejected": -37.89208221435547,
|
| 2158 |
+
"loss": 0.144,
|
| 2159 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2160 |
+
"rewards/chosen": 2.287868022918701,
|
| 2161 |
+
"rewards/margins": 2.641819477081299,
|
| 2162 |
+
"rewards/rejected": -0.3539513051509857,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.8338594436645508,
|
| 2169 |
+
"logits/rejected": -1.8298187255859375,
|
| 2170 |
+
"logps/chosen": -31.842514038085938,
|
| 2171 |
+
"logps/rejected": -31.77047348022461,
|
| 2172 |
+
"loss": 0.1419,
|
| 2173 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2174 |
+
"rewards/chosen": 2.3397603034973145,
|
| 2175 |
+
"rewards/margins": 2.4782257080078125,
|
| 2176 |
+
"rewards/rejected": -0.13846543431282043,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -1.9958465099334717,
|
| 2183 |
+
"logits/rejected": -1.9933058023452759,
|
| 2184 |
+
"logps/chosen": -28.800048828125,
|
| 2185 |
+
"logps/rejected": -32.510841369628906,
|
| 2186 |
+
"loss": 0.1553,
|
| 2187 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2188 |
+
"rewards/chosen": 1.9972022771835327,
|
| 2189 |
+
"rewards/margins": 2.2903003692626953,
|
| 2190 |
+
"rewards/rejected": -0.29309794306755066,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.1668412685394287,
|
| 2196 |
+
"eval_logits/rejected": -2.162137985229492,
|
| 2197 |
+
"eval_logps/chosen": -33.57229232788086,
|
| 2198 |
+
"eval_logps/rejected": -37.258155822753906,
|
| 2199 |
+
"eval_loss": 0.46543803811073303,
|
| 2200 |
+
"eval_rewards/accuracies": 0.6034052968025208,
|
| 2201 |
+
"eval_rewards/chosen": 0.36980658769607544,
|
| 2202 |
+
"eval_rewards/margins": 0.16303294897079468,
|
| 2203 |
+
"eval_rewards/rejected": 0.20677369832992554,
|
| 2204 |
+
"eval_runtime": 145.9296,
|
| 2205 |
+
"eval_samples_per_second": 2.35,
|
| 2206 |
+
"eval_steps_per_second": 0.295,
|
| 2207 |
+
"step": 1400
|
| 2208 |
+
},
|
| 2209 |
+
{
|
| 2210 |
+
"epoch": 3.66,
|
| 2211 |
+
"learning_rate": 1.0775199359171346e-07,
|
| 2212 |
+
"logits/chosen": -1.9372856616973877,
|
| 2213 |
+
"logits/rejected": -1.9331865310668945,
|
| 2214 |
+
"logps/chosen": -30.8883113861084,
|
| 2215 |
+
"logps/rejected": -29.84084129333496,
|
| 2216 |
+
"loss": 0.1363,
|
| 2217 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2218 |
+
"rewards/chosen": 2.2033774852752686,
|
| 2219 |
+
"rewards/margins": 2.4324822425842285,
|
| 2220 |
+
"rewards/rejected": -0.22910502552986145,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.8948686122894287,
|
| 2227 |
+
"logits/rejected": -1.8960151672363281,
|
| 2228 |
+
"logps/chosen": -30.679889678955078,
|
| 2229 |
+
"logps/rejected": -35.14348602294922,
|
| 2230 |
+
"loss": 0.152,
|
| 2231 |
+
"rewards/accuracies": 0.9375,
|
| 2232 |
+
"rewards/chosen": 2.1479902267456055,
|
| 2233 |
+
"rewards/margins": 2.323575735092163,
|
| 2234 |
+
"rewards/rejected": -0.17558541893959045,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.8465667963027954,
|
| 2241 |
+
"logits/rejected": -1.8398901224136353,
|
| 2242 |
+
"logps/chosen": -31.448232650756836,
|
| 2243 |
+
"logps/rejected": -37.158573150634766,
|
| 2244 |
+
"loss": 0.1367,
|
| 2245 |
+
"rewards/accuracies": 0.9375,
|
| 2246 |
+
"rewards/chosen": 2.1548283100128174,
|
| 2247 |
+
"rewards/margins": 2.530158519744873,
|
| 2248 |
+
"rewards/rejected": -0.3753301799297333,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.859572172164917,
|
| 2255 |
+
"logits/rejected": -1.8541902303695679,
|
| 2256 |
+
"logps/chosen": -26.713485717773438,
|
| 2257 |
+
"logps/rejected": -32.027252197265625,
|
| 2258 |
+
"loss": 0.1488,
|
| 2259 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2260 |
+
"rewards/chosen": 2.0798051357269287,
|
| 2261 |
+
"rewards/margins": 2.317080020904541,
|
| 2262 |
+
"rewards/rejected": -0.2372746467590332,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.8826467990875244,
|
| 2269 |
+
"logits/rejected": -1.8725192546844482,
|
| 2270 |
+
"logps/chosen": -30.08587074279785,
|
| 2271 |
+
"logps/rejected": -36.018455505371094,
|
| 2272 |
+
"loss": 0.1431,
|
| 2273 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2274 |
+
"rewards/chosen": 1.9165277481079102,
|
| 2275 |
+
"rewards/margins": 2.503002405166626,
|
| 2276 |
+
"rewards/rejected": -0.5864745378494263,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.8568382263183594,
|
| 2283 |
+
"logits/rejected": -1.8540499210357666,
|
| 2284 |
+
"logps/chosen": -30.78998374938965,
|
| 2285 |
+
"logps/rejected": -31.02651596069336,
|
| 2286 |
+
"loss": 0.1648,
|
| 2287 |
+
"rewards/accuracies": 0.9375,
|
| 2288 |
+
"rewards/chosen": 2.0158348083496094,
|
| 2289 |
+
"rewards/margins": 2.124389886856079,
|
| 2290 |
+
"rewards/rejected": -0.10855510085821152,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -1.9178192615509033,
|
| 2297 |
+
"logits/rejected": -1.924325704574585,
|
| 2298 |
+
"logps/chosen": -29.135822296142578,
|
| 2299 |
+
"logps/rejected": -33.322288513183594,
|
| 2300 |
+
"loss": 0.1271,
|
| 2301 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2302 |
+
"rewards/chosen": 2.3466856479644775,
|
| 2303 |
+
"rewards/margins": 2.634577989578247,
|
| 2304 |
+
"rewards/rejected": -0.28789228200912476,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -1.9184119701385498,
|
| 2311 |
+
"logits/rejected": -1.9287493228912354,
|
| 2312 |
+
"logps/chosen": -29.0635986328125,
|
| 2313 |
+
"logps/rejected": -29.924457550048828,
|
| 2314 |
+
"loss": 0.1386,
|
| 2315 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2316 |
+
"rewards/chosen": 2.1053242683410645,
|
| 2317 |
+
"rewards/margins": 2.342036008834839,
|
| 2318 |
+
"rewards/rejected": -0.2367115318775177,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.8501653671264648,
|
| 2325 |
+
"logits/rejected": -1.8432512283325195,
|
| 2326 |
+
"logps/chosen": -31.104766845703125,
|
| 2327 |
+
"logps/rejected": -32.60163879394531,
|
| 2328 |
+
"loss": 0.144,
|
| 2329 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2330 |
+
"rewards/chosen": 2.0413756370544434,
|
| 2331 |
+
"rewards/margins": 2.3543033599853516,
|
| 2332 |
+
"rewards/rejected": -0.31292763352394104,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.9025371074676514,
|
| 2339 |
+
"logits/rejected": -1.8992702960968018,
|
| 2340 |
+
"logps/chosen": -31.016992568969727,
|
| 2341 |
+
"logps/rejected": -33.745033264160156,
|
| 2342 |
+
"loss": 0.1447,
|
| 2343 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2344 |
+
"rewards/chosen": 2.086864709854126,
|
| 2345 |
+
"rewards/margins": 2.474714517593384,
|
| 2346 |
+
"rewards/rejected": -0.3878494203090668,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.1669082641601562,
|
| 2352 |
+
"eval_logits/rejected": -2.1621992588043213,
|
| 2353 |
+
"eval_logps/chosen": -33.575931549072266,
|
| 2354 |
+
"eval_logps/rejected": -37.24650192260742,
|
| 2355 |
+
"eval_loss": 0.46842658519744873,
|
| 2356 |
+
"eval_rewards/accuracies": 0.574335515499115,
|
| 2357 |
+
"eval_rewards/chosen": 0.3668961524963379,
|
| 2358 |
+
"eval_rewards/margins": 0.15079911053180695,
|
| 2359 |
+
"eval_rewards/rejected": 0.21609705686569214,
|
| 2360 |
+
"eval_runtime": 145.9178,
|
| 2361 |
+
"eval_samples_per_second": 2.351,
|
| 2362 |
+
"eval_steps_per_second": 0.295,
|
| 2363 |
+
"step": 1500
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 3.92,
|
| 2367 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2368 |
+
"logits/chosen": -1.9817174673080444,
|
| 2369 |
+
"logits/rejected": -1.9823877811431885,
|
| 2370 |
+
"logps/chosen": -27.261791229248047,
|
| 2371 |
+
"logps/rejected": -33.12643814086914,
|
| 2372 |
+
"loss": 0.1654,
|
| 2373 |
+
"rewards/accuracies": 0.875,
|
| 2374 |
+
"rewards/chosen": 2.0435855388641357,
|
| 2375 |
+
"rewards/margins": 2.35237717628479,
|
| 2376 |
+
"rewards/rejected": -0.30879148840904236,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.834775686264038,
|
| 2383 |
+
"logits/rejected": -1.8347667455673218,
|
| 2384 |
+
"logps/chosen": -30.5068416595459,
|
| 2385 |
+
"logps/rejected": -34.2733154296875,
|
| 2386 |
+
"loss": 0.1565,
|
| 2387 |
+
"rewards/accuracies": 0.9375,
|
| 2388 |
+
"rewards/chosen": 2.0280237197875977,
|
| 2389 |
+
"rewards/margins": 2.347689151763916,
|
| 2390 |
+
"rewards/rejected": -0.31966525316238403,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -1.986581563949585,
|
| 2397 |
+
"logits/rejected": -1.9839556217193604,
|
| 2398 |
+
"logps/chosen": -29.06634521484375,
|
| 2399 |
+
"logps/rejected": -31.9202938079834,
|
| 2400 |
+
"loss": 0.1594,
|
| 2401 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2402 |
+
"rewards/chosen": 1.926171898841858,
|
| 2403 |
+
"rewards/margins": 2.2754859924316406,
|
| 2404 |
+
"rewards/rejected": -0.34931421279907227,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -1.9706960916519165,
|
| 2411 |
+
"logits/rejected": -1.9718106985092163,
|
| 2412 |
+
"logps/chosen": -28.249774932861328,
|
| 2413 |
+
"logps/rejected": -30.143047332763672,
|
| 2414 |
+
"loss": 0.1594,
|
| 2415 |
+
"rewards/accuracies": 0.9583333730697632,
|
| 2416 |
+
"rewards/chosen": 1.8961349725723267,
|
| 2417 |
+
"rewards/margins": 2.178429126739502,
|
| 2418 |
+
"rewards/rejected": -0.2822941541671753,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 0.16076272354497537,
|
| 2426 |
+
"train_runtime": 10802.3033,
|
| 2427 |
+
"train_samples_per_second": 1.14,
|
| 2428 |
+
"train_steps_per_second": 0.143
|
| 2429 |
}
|
| 2430 |
],
|
| 2431 |
"logging_steps": 10,
|
| 2432 |
+
"max_steps": 1540,
|
| 2433 |
"num_input_tokens_seen": 0,
|
| 2434 |
+
"num_train_epochs": 4,
|
| 2435 |
"save_steps": 100,
|
| 2436 |
"total_flos": 0.0,
|
| 2437 |
"train_batch_size": 4,
|