Commit
•
6dd642a
1
Parent(s):
8556769
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 +1906 -90
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
| 4 |
-
- alignment-handbook
|
| 5 |
- trl
|
| 6 |
- dpo
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
| 9 |
-
datasets:
|
| 10 |
-
- hugodk-sch/aftonposten_title_prefs
|
| 11 |
model-index:
|
| 12 |
- name: aftonposten-6b-align-scan
|
| 13 |
results: []
|
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
# aftonposten-6b-align-scan
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 0.
|
| 24 |
-
- Rewards/chosen: 0.
|
| 25 |
-
- Rewards/rejected: 0.
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -33.
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,15 +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 |
-
| 0.4799 | 0.26 | 100 |
|
| 67 |
-
| 0.4548 | 0.52 | 200 |
|
| 68 |
-
| 0.41 | 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: 0.4727
|
| 22 |
+
- Rewards/chosen: 0.2480
|
| 23 |
+
- Rewards/rejected: 0.1188
|
| 24 |
+
- Rewards/accuracies: 0.6005
|
| 25 |
+
- Rewards/margins: 0.1292
|
| 26 |
+
- Logps/rejected: -37.3186
|
| 27 |
+
- Logps/chosen: -33.6212
|
| 28 |
+
- Logits/rejected: -2.1470
|
| 29 |
+
- Logits/chosen: -2.1517
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 55 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
- lr_scheduler_type: cosine
|
| 57 |
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 4
|
| 59 |
|
| 60 |
### Training results
|
| 61 |
|
| 62 |
+
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
| 63 |
+
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
| 64 |
+
| 0.4799 | 0.26 | 100 | -2.2349 | -2.2301 | -33.9029 | -37.3943 | 0.4987 | 0.5104 | 0.0790 | 0.0056 | 0.0734 |
|
| 65 |
+
| 0.4548 | 0.52 | 200 | -2.2335 | -2.2287 | -33.7696 | -37.2846 | 0.4956 | 0.5341 | 0.1590 | 0.0198 | 0.1392 |
|
| 66 |
+
| 0.41 | 0.78 | 300 | -2.2309 | -2.2261 | -33.7614 | -37.2848 | 0.4937 | 0.5361 | 0.1639 | 0.0248 | 0.1391 |
|
| 67 |
+
| 0.3497 | 1.04 | 400 | 0.4927 | 0.2171 | 0.1863 | 0.5652 | 0.0309 | -37.2062 | -33.6727 | -2.2113 | -2.2162 |
|
| 68 |
+
| 0.2906 | 1.3 | 500 | 0.4870 | 0.2484 | 0.1921 | 0.5922 | 0.0563 | -37.1964 | -33.6205 | -2.1834 | -2.1881 |
|
| 69 |
+
| 0.3014 | 1.56 | 600 | 0.4796 | 0.2630 | 0.1719 | 0.5797 | 0.0911 | -37.2301 | -33.5962 | -2.1694 | -2.1741 |
|
| 70 |
+
| 0.2776 | 1.82 | 700 | 0.4825 | 0.2341 | 0.1554 | 0.5768 | 0.0787 | -37.2576 | -33.6444 | -2.1625 | -2.1672 |
|
| 71 |
+
| 0.201 | 2.08 | 800 | 0.4766 | 0.2639 | 0.1595 | 0.5914 | 0.1043 | -37.2507 | -33.5948 | -2.1641 | -2.1689 |
|
| 72 |
+
| 0.1721 | 2.34 | 900 | 0.4749 | 0.2446 | 0.1298 | 0.5860 | 0.1148 | -37.3003 | -33.6269 | -2.1516 | -2.1563 |
|
| 73 |
+
| 0.2259 | 2.6 | 1000 | 0.4736 | 0.2483 | 0.1257 | 0.5860 | 0.1226 | -37.3072 | -33.6207 | -2.1481 | -2.1528 |
|
| 74 |
+
| 0.2405 | 2.86 | 1100 | 0.4740 | 0.2438 | 0.1229 | 0.5860 | 0.1209 | -37.3118 | -33.6283 | -2.1475 | -2.1522 |
|
| 75 |
+
| 0.1793 | 3.12 | 1200 | 0.4746 | 0.2441 | 0.1249 | 0.5685 | 0.1192 | -37.3085 | -33.6277 | -2.1469 | -2.1516 |
|
| 76 |
+
| 0.1633 | 3.38 | 1300 | 0.4744 | 0.2433 | 0.1235 | 0.6009 | 0.1198 | -37.3107 | -33.6290 | -2.1471 | -2.1518 |
|
| 77 |
+
| 0.202 | 3.64 | 1400 | 0.4748 | 0.2450 | 0.1279 | 0.5831 | 0.1170 | -37.3034 | -33.6263 | -2.1472 | -2.1519 |
|
| 78 |
+
| 0.1889 | 3.9 | 1500 | 0.4727 | 0.2480 | 0.1188 | 0.6005 | 0.1292 | -37.3186 | -33.6212 | -2.1470 | -2.1517 |
|
| 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 sha256:055ffd1fcf3c15abaeb1e658da1ca3ca33d4856bb69ba7776358c3487a3e5b13
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.230694055557251,
|
| 4 |
"eval_logits/rejected": -2.225895643234253,
|
| 5 |
"eval_logps/chosen": -33.75852966308594,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.354,
|
| 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": -2.230694055557251,
|
| 4 |
"eval_logits/rejected": -2.225895643234253,
|
| 5 |
"eval_logps/chosen": -33.75852966308594,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.354,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.19402528018146367,
|
| 17 |
+
"train_runtime": 10806.8766,
|
| 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": 0.19402528018146367,
|
| 4 |
+
"train_runtime": 10806.8766,
|
| 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.
|
| 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.
|
| 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": -
|
| 525 |
-
"logits/rejected": -
|
| 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.
|
| 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": -1.9166145324707031,
|
| 497 |
+
"logits/rejected": -1.9134905338287354,
|
| 498 |
+
"logps/chosen": -31.025157928466797,
|
| 499 |
+
"logps/rejected": -33.5601692199707,
|
| 500 |
+
"loss": 0.4452,
|
| 501 |
+
"rewards/accuracies": 0.699999988079071,
|
| 502 |
+
"rewards/chosen": 0.34691327810287476,
|
| 503 |
+
"rewards/margins": 0.23772580921649933,
|
| 504 |
+
"rewards/rejected": 0.10918750613927841,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.9646613597869873,
|
| 511 |
+
"logits/rejected": -1.952566146850586,
|
| 512 |
+
"logps/chosen": -34.02196502685547,
|
| 513 |
+
"logps/rejected": -33.40004348754883,
|
| 514 |
+
"loss": 0.4395,
|
| 515 |
+
"rewards/accuracies": 0.675000011920929,
|
| 516 |
+
"rewards/chosen": 0.34446340799331665,
|
| 517 |
+
"rewards/margins": 0.2611829340457916,
|
| 518 |
+
"rewards/rejected": 0.08328045904636383,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -1.9993693828582764,
|
| 525 |
+
"logits/rejected": -1.9980556964874268,
|
| 526 |
+
"logps/chosen": -32.787845611572266,
|
| 527 |
+
"logps/rejected": -32.25823211669922,
|
| 528 |
+
"loss": 0.4341,
|
| 529 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 530 |
+
"rewards/chosen": 0.4187692701816559,
|
| 531 |
+
"rewards/margins": 0.285956472158432,
|
| 532 |
+
"rewards/rejected": 0.1328127533197403,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.0861244201660156,
|
| 539 |
+
"logits/rejected": -2.0705418586730957,
|
| 540 |
+
"logps/chosen": -33.38449478149414,
|
| 541 |
+
"logps/rejected": -32.79410171508789,
|
| 542 |
+
"loss": 0.4348,
|
| 543 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 544 |
+
"rewards/chosen": 0.4769386351108551,
|
| 545 |
+
"rewards/margins": 0.28257787227630615,
|
| 546 |
+
"rewards/rejected": 0.19436076283454895,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.959221601486206,
|
| 553 |
+
"logits/rejected": -1.9584068059921265,
|
| 554 |
+
"logps/chosen": -32.52511215209961,
|
| 555 |
+
"logps/rejected": -32.22833251953125,
|
| 556 |
+
"loss": 0.4289,
|
| 557 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 558 |
+
"rewards/chosen": 0.45419034361839294,
|
| 559 |
+
"rewards/margins": 0.31210166215896606,
|
| 560 |
+
"rewards/rejected": 0.14208868145942688,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.9131596088409424,
|
| 567 |
+
"logits/rejected": -1.9233753681182861,
|
| 568 |
+
"logps/chosen": -31.385852813720703,
|
| 569 |
+
"logps/rejected": -35.03350067138672,
|
| 570 |
+
"loss": 0.4234,
|
| 571 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 572 |
+
"rewards/chosen": 0.49130287766456604,
|
| 573 |
+
"rewards/margins": 0.338148295879364,
|
| 574 |
+
"rewards/rejected": 0.15315461158752441,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.0467209815979004,
|
| 581 |
+
"logits/rejected": -2.0402884483337402,
|
| 582 |
+
"logps/chosen": -32.84780502319336,
|
| 583 |
+
"logps/rejected": -28.970489501953125,
|
| 584 |
+
"loss": 0.4219,
|
| 585 |
+
"rewards/accuracies": 0.800000011920929,
|
| 586 |
+
"rewards/chosen": 0.48249703645706177,
|
| 587 |
+
"rewards/margins": 0.32893475890159607,
|
| 588 |
+
"rewards/rejected": 0.15356223285198212,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.905468225479126,
|
| 595 |
+
"logits/rejected": -1.9076448678970337,
|
| 596 |
+
"logps/chosen": -33.3749885559082,
|
| 597 |
+
"logps/rejected": -30.71591567993164,
|
| 598 |
+
"loss": 0.4091,
|
| 599 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 600 |
+
"rewards/chosen": 0.5184367895126343,
|
| 601 |
+
"rewards/margins": 0.42440906167030334,
|
| 602 |
+
"rewards/rejected": 0.09402772784233093,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9335657358169556,
|
| 609 |
+
"logits/rejected": -1.9322980642318726,
|
| 610 |
+
"logps/chosen": -33.33202362060547,
|
| 611 |
+
"logps/rejected": -35.62676239013672,
|
| 612 |
+
"loss": 0.3702,
|
| 613 |
+
"rewards/accuracies": 0.7583333253860474,
|
| 614 |
+
"rewards/chosen": 0.6316612362861633,
|
| 615 |
+
"rewards/margins": 0.6141165494918823,
|
| 616 |
+
"rewards/rejected": 0.017544690519571304,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.8687912225723267,
|
| 623 |
+
"logits/rejected": -1.860594391822815,
|
| 624 |
+
"logps/chosen": -30.623647689819336,
|
| 625 |
+
"logps/rejected": -36.03453826904297,
|
| 626 |
+
"loss": 0.3497,
|
| 627 |
+
"rewards/accuracies": 0.8125,
|
| 628 |
+
"rewards/chosen": 0.6901980042457581,
|
| 629 |
+
"rewards/margins": 0.7465595006942749,
|
| 630 |
+
"rewards/rejected": -0.05636153370141983,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.2161529064178467,
|
| 636 |
+
"eval_logits/rejected": -2.2113351821899414,
|
| 637 |
+
"eval_logps/chosen": -33.672698974609375,
|
| 638 |
+
"eval_logps/rejected": -37.206199645996094,
|
| 639 |
+
"eval_loss": 0.4927479326725006,
|
| 640 |
+
"eval_rewards/accuracies": 0.5651993155479431,
|
| 641 |
+
"eval_rewards/chosen": 0.2171139419078827,
|
| 642 |
+
"eval_rewards/margins": 0.030863292515277863,
|
| 643 |
+
"eval_rewards/rejected": 0.18625062704086304,
|
| 644 |
+
"eval_runtime": 146.2598,
|
| 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": -2.0422446727752686,
|
| 653 |
+
"logits/rejected": -2.0450665950775146,
|
| 654 |
+
"logps/chosen": -31.823333740234375,
|
| 655 |
+
"logps/rejected": -34.91590118408203,
|
| 656 |
+
"loss": 0.3378,
|
| 657 |
+
"rewards/accuracies": 0.875,
|
| 658 |
+
"rewards/chosen": 0.7383630871772766,
|
| 659 |
+
"rewards/margins": 0.8121965527534485,
|
| 660 |
+
"rewards/rejected": -0.07383350282907486,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.8828691244125366,
|
| 667 |
+
"logits/rejected": -1.8815042972564697,
|
| 668 |
+
"logps/chosen": -27.965286254882812,
|
| 669 |
+
"logps/rejected": -32.40424346923828,
|
| 670 |
+
"loss": 0.3525,
|
| 671 |
+
"rewards/accuracies": 0.9375,
|
| 672 |
+
"rewards/chosen": 0.6613627672195435,
|
| 673 |
+
"rewards/margins": 0.7133124470710754,
|
| 674 |
+
"rewards/rejected": -0.05194968730211258,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.8411157131195068,
|
| 681 |
+
"logits/rejected": -1.8340963125228882,
|
| 682 |
+
"logps/chosen": -32.586795806884766,
|
| 683 |
+
"logps/rejected": -34.0372428894043,
|
| 684 |
+
"loss": 0.334,
|
| 685 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 686 |
+
"rewards/chosen": 0.8923752903938293,
|
| 687 |
+
"rewards/margins": 0.7675577402114868,
|
| 688 |
+
"rewards/rejected": 0.1248176246881485,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -2.01190185546875,
|
| 695 |
+
"logits/rejected": -2.006816864013672,
|
| 696 |
+
"logps/chosen": -30.412521362304688,
|
| 697 |
+
"logps/rejected": -32.05153274536133,
|
| 698 |
+
"loss": 0.3237,
|
| 699 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 700 |
+
"rewards/chosen": 0.8224636316299438,
|
| 701 |
+
"rewards/margins": 0.8728159666061401,
|
| 702 |
+
"rewards/rejected": -0.05035240575671196,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.8965171575546265,
|
| 709 |
+
"logits/rejected": -1.9006315469741821,
|
| 710 |
+
"logps/chosen": -32.99296569824219,
|
| 711 |
+
"logps/rejected": -32.713218688964844,
|
| 712 |
+
"loss": 0.302,
|
| 713 |
+
"rewards/accuracies": 0.9375,
|
| 714 |
+
"rewards/chosen": 0.913729190826416,
|
| 715 |
+
"rewards/margins": 0.9884780049324036,
|
| 716 |
+
"rewards/rejected": -0.07474879175424576,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.8998197317123413,
|
| 723 |
+
"logits/rejected": -1.8941733837127686,
|
| 724 |
+
"logps/chosen": -33.83064651489258,
|
| 725 |
+
"logps/rejected": -35.21913146972656,
|
| 726 |
+
"loss": 0.2931,
|
| 727 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 728 |
+
"rewards/chosen": 0.9533284902572632,
|
| 729 |
+
"rewards/margins": 1.115350365638733,
|
| 730 |
+
"rewards/rejected": -0.16202178597450256,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.9277269840240479,
|
| 737 |
+
"logits/rejected": -1.9273641109466553,
|
| 738 |
+
"logps/chosen": -32.60927963256836,
|
| 739 |
+
"logps/rejected": -34.12434005737305,
|
| 740 |
+
"loss": 0.3206,
|
| 741 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 742 |
+
"rewards/chosen": 0.9354284405708313,
|
| 743 |
+
"rewards/margins": 0.8987458348274231,
|
| 744 |
+
"rewards/rejected": 0.0366826057434082,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -1.9775463342666626,
|
| 751 |
+
"logits/rejected": -1.9590513706207275,
|
| 752 |
+
"logps/chosen": -32.508819580078125,
|
| 753 |
+
"logps/rejected": -33.2094612121582,
|
| 754 |
+
"loss": 0.2906,
|
| 755 |
+
"rewards/accuracies": 0.9375,
|
| 756 |
+
"rewards/chosen": 0.9911691546440125,
|
| 757 |
+
"rewards/margins": 1.07485830783844,
|
| 758 |
+
"rewards/rejected": -0.08368916064500809,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.9436886310577393,
|
| 765 |
+
"logits/rejected": -1.9431581497192383,
|
| 766 |
+
"logps/chosen": -30.075321197509766,
|
| 767 |
+
"logps/rejected": -31.215450286865234,
|
| 768 |
+
"loss": 0.3167,
|
| 769 |
+
"rewards/accuracies": 0.875,
|
| 770 |
+
"rewards/chosen": 0.9215895533561707,
|
| 771 |
+
"rewards/margins": 0.9032338261604309,
|
| 772 |
+
"rewards/rejected": 0.018355702981352806,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.9060132503509521,
|
| 779 |
+
"logits/rejected": -1.8992029428482056,
|
| 780 |
+
"logps/chosen": -30.9982852935791,
|
| 781 |
+
"logps/rejected": -34.8018798828125,
|
| 782 |
+
"loss": 0.2906,
|
| 783 |
+
"rewards/accuracies": 0.862500011920929,
|
| 784 |
+
"rewards/chosen": 1.0031068325042725,
|
| 785 |
+
"rewards/margins": 1.1081880331039429,
|
| 786 |
+
"rewards/rejected": -0.10508137941360474,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.1881351470947266,
|
| 792 |
+
"eval_logits/rejected": -2.183356761932373,
|
| 793 |
+
"eval_logps/chosen": -33.62051010131836,
|
| 794 |
+
"eval_logps/rejected": -37.196380615234375,
|
| 795 |
+
"eval_loss": 0.48696067929267883,
|
| 796 |
+
"eval_rewards/accuracies": 0.5921927094459534,
|
| 797 |
+
"eval_rewards/chosen": 0.24842412769794464,
|
| 798 |
+
"eval_rewards/margins": 0.0562782883644104,
|
| 799 |
+
"eval_rewards/rejected": 0.19214583933353424,
|
| 800 |
+
"eval_runtime": 145.95,
|
| 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": -1.9005699157714844,
|
| 809 |
+
"logits/rejected": -1.9044411182403564,
|
| 810 |
+
"logps/chosen": -27.744693756103516,
|
| 811 |
+
"logps/rejected": -33.00718688964844,
|
| 812 |
+
"loss": 0.3144,
|
| 813 |
+
"rewards/accuracies": 0.875,
|
| 814 |
+
"rewards/chosen": 0.8580785989761353,
|
| 815 |
+
"rewards/margins": 0.9468171000480652,
|
| 816 |
+
"rewards/rejected": -0.08873851597309113,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.907922387123108,
|
| 823 |
+
"logits/rejected": -1.9184629917144775,
|
| 824 |
+
"logps/chosen": -31.68210220336914,
|
| 825 |
+
"logps/rejected": -30.8319091796875,
|
| 826 |
+
"loss": 0.3058,
|
| 827 |
+
"rewards/accuracies": 0.887499988079071,
|
| 828 |
+
"rewards/chosen": 0.9481385946273804,
|
| 829 |
+
"rewards/margins": 1.056273102760315,
|
| 830 |
+
"rewards/rejected": -0.1081344336271286,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.8528549671173096,
|
| 837 |
+
"logits/rejected": -1.8505821228027344,
|
| 838 |
+
"logps/chosen": -30.023778915405273,
|
| 839 |
+
"logps/rejected": -30.290355682373047,
|
| 840 |
+
"loss": 0.2945,
|
| 841 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 842 |
+
"rewards/chosen": 0.9859828948974609,
|
| 843 |
+
"rewards/margins": 1.054137945175171,
|
| 844 |
+
"rewards/rejected": -0.06815493851900101,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.9326980113983154,
|
| 851 |
+
"logits/rejected": -1.925724744796753,
|
| 852 |
+
"logps/chosen": -32.48560333251953,
|
| 853 |
+
"logps/rejected": -33.21514892578125,
|
| 854 |
+
"loss": 0.2771,
|
| 855 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 856 |
+
"rewards/chosen": 1.0483348369598389,
|
| 857 |
+
"rewards/margins": 1.1669293642044067,
|
| 858 |
+
"rewards/rejected": -0.11859482526779175,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.8988561630249023,
|
| 865 |
+
"logits/rejected": -1.905217170715332,
|
| 866 |
+
"logps/chosen": -30.224956512451172,
|
| 867 |
+
"logps/rejected": -32.778907775878906,
|
| 868 |
+
"loss": 0.3108,
|
| 869 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 870 |
+
"rewards/chosen": 0.9169367551803589,
|
| 871 |
+
"rewards/margins": 0.929042637348175,
|
| 872 |
+
"rewards/rejected": -0.012105767615139484,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -1.9703216552734375,
|
| 879 |
+
"logits/rejected": -1.9476426839828491,
|
| 880 |
+
"logps/chosen": -29.846935272216797,
|
| 881 |
+
"logps/rejected": -33.04885482788086,
|
| 882 |
+
"loss": 0.3058,
|
| 883 |
+
"rewards/accuracies": 0.875,
|
| 884 |
+
"rewards/chosen": 0.9543794393539429,
|
| 885 |
+
"rewards/margins": 0.9860580563545227,
|
| 886 |
+
"rewards/rejected": -0.031678564846515656,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.9047954082489014,
|
| 893 |
+
"logits/rejected": -1.899834394454956,
|
| 894 |
+
"logps/chosen": -30.691064834594727,
|
| 895 |
+
"logps/rejected": -32.055702209472656,
|
| 896 |
+
"loss": 0.305,
|
| 897 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 898 |
+
"rewards/chosen": 1.0354000329971313,
|
| 899 |
+
"rewards/margins": 0.9962730407714844,
|
| 900 |
+
"rewards/rejected": 0.039127081632614136,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -1.9474509954452515,
|
| 907 |
+
"logits/rejected": -1.9493205547332764,
|
| 908 |
+
"logps/chosen": -32.68132781982422,
|
| 909 |
+
"logps/rejected": -35.09022903442383,
|
| 910 |
+
"loss": 0.2831,
|
| 911 |
+
"rewards/accuracies": 0.9375,
|
| 912 |
+
"rewards/chosen": 1.0166804790496826,
|
| 913 |
+
"rewards/margins": 1.19026780128479,
|
| 914 |
+
"rewards/rejected": -0.17358721792697906,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.929598093032837,
|
| 921 |
+
"logits/rejected": -1.9289600849151611,
|
| 922 |
+
"logps/chosen": -28.859294891357422,
|
| 923 |
+
"logps/rejected": -31.535837173461914,
|
| 924 |
+
"loss": 0.2904,
|
| 925 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 926 |
+
"rewards/chosen": 1.0396578311920166,
|
| 927 |
+
"rewards/margins": 1.084177017211914,
|
| 928 |
+
"rewards/rejected": -0.04451901838183403,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -1.9405543804168701,
|
| 935 |
+
"logits/rejected": -1.93850576877594,
|
| 936 |
+
"logps/chosen": -31.257436752319336,
|
| 937 |
+
"logps/rejected": -32.223670959472656,
|
| 938 |
+
"loss": 0.3014,
|
| 939 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 940 |
+
"rewards/chosen": 1.0445080995559692,
|
| 941 |
+
"rewards/margins": 0.9987362623214722,
|
| 942 |
+
"rewards/rejected": 0.045771874487400055,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.174114942550659,
|
| 948 |
+
"eval_logits/rejected": -2.1693577766418457,
|
| 949 |
+
"eval_logps/chosen": -33.596221923828125,
|
| 950 |
+
"eval_logps/rejected": -37.230098724365234,
|
| 951 |
+
"eval_loss": 0.4795546531677246,
|
| 952 |
+
"eval_rewards/accuracies": 0.579734206199646,
|
| 953 |
+
"eval_rewards/chosen": 0.26299843192100525,
|
| 954 |
+
"eval_rewards/margins": 0.09108825773000717,
|
| 955 |
+
"eval_rewards/rejected": 0.17191016674041748,
|
| 956 |
+
"eval_runtime": 145.976,
|
| 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": -1.9819376468658447,
|
| 965 |
+
"logits/rejected": -1.9884719848632812,
|
| 966 |
+
"logps/chosen": -30.554279327392578,
|
| 967 |
+
"logps/rejected": -32.0571403503418,
|
| 968 |
+
"loss": 0.3038,
|
| 969 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 970 |
+
"rewards/chosen": 1.0098621845245361,
|
| 971 |
+
"rewards/margins": 1.0062358379364014,
|
| 972 |
+
"rewards/rejected": 0.0036262169014662504,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -1.9370014667510986,
|
| 979 |
+
"logits/rejected": -1.9347403049468994,
|
| 980 |
+
"logps/chosen": -32.8254508972168,
|
| 981 |
+
"logps/rejected": -31.027420043945312,
|
| 982 |
+
"loss": 0.2776,
|
| 983 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 984 |
+
"rewards/chosen": 1.1582577228546143,
|
| 985 |
+
"rewards/margins": 1.1834052801132202,
|
| 986 |
+
"rewards/rejected": -0.025147801265120506,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.8845012187957764,
|
| 993 |
+
"logits/rejected": -1.8778787851333618,
|
| 994 |
+
"logps/chosen": -33.42643737792969,
|
| 995 |
+
"logps/rejected": -32.37970733642578,
|
| 996 |
+
"loss": 0.2689,
|
| 997 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 998 |
+
"rewards/chosen": 1.2226732969284058,
|
| 999 |
+
"rewards/margins": 1.2361730337142944,
|
| 1000 |
+
"rewards/rejected": -0.013499641790986061,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.8900282382965088,
|
| 1007 |
+
"logits/rejected": -1.8962434530258179,
|
| 1008 |
+
"logps/chosen": -32.14881134033203,
|
| 1009 |
+
"logps/rejected": -33.176612854003906,
|
| 1010 |
+
"loss": 0.2981,
|
| 1011 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1012 |
+
"rewards/chosen": 1.098296880722046,
|
| 1013 |
+
"rewards/margins": 1.0601081848144531,
|
| 1014 |
+
"rewards/rejected": 0.03818872570991516,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -1.9740774631500244,
|
| 1021 |
+
"logits/rejected": -1.9881250858306885,
|
| 1022 |
+
"logps/chosen": -30.15896987915039,
|
| 1023 |
+
"logps/rejected": -31.947906494140625,
|
| 1024 |
+
"loss": 0.3021,
|
| 1025 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1026 |
+
"rewards/chosen": 1.0160971879959106,
|
| 1027 |
+
"rewards/margins": 0.9862273335456848,
|
| 1028 |
+
"rewards/rejected": 0.02987002767622471,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.8505609035491943,
|
| 1035 |
+
"logits/rejected": -1.847583532333374,
|
| 1036 |
+
"logps/chosen": -31.669322967529297,
|
| 1037 |
+
"logps/rejected": -34.97021484375,
|
| 1038 |
+
"loss": 0.2316,
|
| 1039 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1040 |
+
"rewards/chosen": 1.3521816730499268,
|
| 1041 |
+
"rewards/margins": 1.5141586065292358,
|
| 1042 |
+
"rewards/rejected": -0.16197696328163147,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -1.9402939081192017,
|
| 1049 |
+
"logits/rejected": -1.9406039714813232,
|
| 1050 |
+
"logps/chosen": -29.579147338867188,
|
| 1051 |
+
"logps/rejected": -33.99906921386719,
|
| 1052 |
+
"loss": 0.2843,
|
| 1053 |
+
"rewards/accuracies": 0.875,
|
| 1054 |
+
"rewards/chosen": 1.0238673686981201,
|
| 1055 |
+
"rewards/margins": 1.1274796724319458,
|
| 1056 |
+
"rewards/rejected": -0.10361222922801971,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.8894329071044922,
|
| 1063 |
+
"logits/rejected": -1.8864387273788452,
|
| 1064 |
+
"logps/chosen": -29.1514892578125,
|
| 1065 |
+
"logps/rejected": -32.90578079223633,
|
| 1066 |
+
"loss": 0.2983,
|
| 1067 |
+
"rewards/accuracies": 0.875,
|
| 1068 |
+
"rewards/chosen": 1.0325144529342651,
|
| 1069 |
+
"rewards/margins": 1.0853002071380615,
|
| 1070 |
+
"rewards/rejected": -0.05278577655553818,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -1.9213037490844727,
|
| 1077 |
+
"logits/rejected": -1.921303391456604,
|
| 1078 |
+
"logps/chosen": -27.972620010375977,
|
| 1079 |
+
"logps/rejected": -30.834762573242188,
|
| 1080 |
+
"loss": 0.2479,
|
| 1081 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1082 |
+
"rewards/chosen": 1.314284086227417,
|
| 1083 |
+
"rewards/margins": 1.4128423929214478,
|
| 1084 |
+
"rewards/rejected": -0.09855831414461136,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.8346879482269287,
|
| 1091 |
+
"logits/rejected": -1.838030219078064,
|
| 1092 |
+
"logps/chosen": -32.32703399658203,
|
| 1093 |
+
"logps/rejected": -31.50747299194336,
|
| 1094 |
+
"loss": 0.2776,
|
| 1095 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1096 |
+
"rewards/chosen": 1.2908637523651123,
|
| 1097 |
+
"rewards/margins": 1.316535234451294,
|
| 1098 |
+
"rewards/rejected": -0.025671422481536865,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.1672425270080566,
|
| 1104 |
+
"eval_logits/rejected": -2.162508487701416,
|
| 1105 |
+
"eval_logps/chosen": -33.64436340332031,
|
| 1106 |
+
"eval_logps/rejected": -37.25758743286133,
|
| 1107 |
+
"eval_loss": 0.4824546277523041,
|
| 1108 |
+
"eval_rewards/accuracies": 0.5768272280693054,
|
| 1109 |
+
"eval_rewards/chosen": 0.23411329090595245,
|
| 1110 |
+
"eval_rewards/margins": 0.07869596779346466,
|
| 1111 |
+
"eval_rewards/rejected": 0.155417338013649,
|
| 1112 |
+
"eval_runtime": 145.94,
|
| 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": -1.9178850650787354,
|
| 1121 |
+
"logits/rejected": -1.9089438915252686,
|
| 1122 |
+
"logps/chosen": -34.47451400756836,
|
| 1123 |
+
"logps/rejected": -31.956844329833984,
|
| 1124 |
+
"loss": 0.2673,
|
| 1125 |
+
"rewards/accuracies": 0.9375,
|
| 1126 |
+
"rewards/chosen": 1.2225029468536377,
|
| 1127 |
+
"rewards/margins": 1.2531225681304932,
|
| 1128 |
+
"rewards/rejected": -0.030619746074080467,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -1.9318691492080688,
|
| 1135 |
+
"logits/rejected": -1.9329137802124023,
|
| 1136 |
+
"logps/chosen": -34.28239440917969,
|
| 1137 |
+
"logps/rejected": -33.12510681152344,
|
| 1138 |
+
"loss": 0.2565,
|
| 1139 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1140 |
+
"rewards/chosen": 1.291794776916504,
|
| 1141 |
+
"rewards/margins": 1.3212449550628662,
|
| 1142 |
+
"rewards/rejected": -0.02945004776120186,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -1.9492158889770508,
|
| 1149 |
+
"logits/rejected": -1.9468295574188232,
|
| 1150 |
+
"logps/chosen": -30.499465942382812,
|
| 1151 |
+
"logps/rejected": -33.04973220825195,
|
| 1152 |
+
"loss": 0.2684,
|
| 1153 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1154 |
+
"rewards/chosen": 1.2383159399032593,
|
| 1155 |
+
"rewards/margins": 1.270018219947815,
|
| 1156 |
+
"rewards/rejected": -0.03170236200094223,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -1.933258056640625,
|
| 1163 |
+
"logits/rejected": -1.938474416732788,
|
| 1164 |
+
"logps/chosen": -29.6435546875,
|
| 1165 |
+
"logps/rejected": -33.32258224487305,
|
| 1166 |
+
"loss": 0.2485,
|
| 1167 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1168 |
+
"rewards/chosen": 1.2853031158447266,
|
| 1169 |
+
"rewards/margins": 1.38596510887146,
|
| 1170 |
+
"rewards/rejected": -0.10066213458776474,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.8792998790740967,
|
| 1177 |
+
"logits/rejected": -1.8781516551971436,
|
| 1178 |
+
"logps/chosen": -31.056386947631836,
|
| 1179 |
+
"logps/rejected": -34.74039840698242,
|
| 1180 |
+
"loss": 0.274,
|
| 1181 |
+
"rewards/accuracies": 0.875,
|
| 1182 |
+
"rewards/chosen": 1.2411561012268066,
|
| 1183 |
+
"rewards/margins": 1.2738196849822998,
|
| 1184 |
+
"rewards/rejected": -0.032663825899362564,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.754612684249878,
|
| 1191 |
+
"logits/rejected": -1.7496144771575928,
|
| 1192 |
+
"logps/chosen": -33.715721130371094,
|
| 1193 |
+
"logps/rejected": -35.355167388916016,
|
| 1194 |
+
"loss": 0.253,
|
| 1195 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1196 |
+
"rewards/chosen": 1.3696316480636597,
|
| 1197 |
+
"rewards/margins": 1.4576833248138428,
|
| 1198 |
+
"rewards/rejected": -0.08805163949728012,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.8957417011260986,
|
| 1205 |
+
"logits/rejected": -1.8994983434677124,
|
| 1206 |
+
"logps/chosen": -33.10436248779297,
|
| 1207 |
+
"logps/rejected": -33.5843505859375,
|
| 1208 |
+
"loss": 0.2679,
|
| 1209 |
+
"rewards/accuracies": 0.8833333849906921,
|
| 1210 |
+
"rewards/chosen": 1.2655742168426514,
|
| 1211 |
+
"rewards/margins": 1.2592796087265015,
|
| 1212 |
+
"rewards/rejected": 0.006294435355812311,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.8399032354354858,
|
| 1219 |
+
"logits/rejected": -1.8381221294403076,
|
| 1220 |
+
"logps/chosen": -31.176462173461914,
|
| 1221 |
+
"logps/rejected": -35.09248733520508,
|
| 1222 |
+
"loss": 0.1728,
|
| 1223 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1224 |
+
"rewards/chosen": 1.6350797414779663,
|
| 1225 |
+
"rewards/margins": 2.047778606414795,
|
| 1226 |
+
"rewards/rejected": -0.412698894739151,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.8872671127319336,
|
| 1233 |
+
"logits/rejected": -1.8859704732894897,
|
| 1234 |
+
"logps/chosen": -30.6798095703125,
|
| 1235 |
+
"logps/rejected": -33.45615005493164,
|
| 1236 |
+
"loss": 0.1882,
|
| 1237 |
+
"rewards/accuracies": 0.9375,
|
| 1238 |
+
"rewards/chosen": 1.6175569295883179,
|
| 1239 |
+
"rewards/margins": 1.8393020629882812,
|
| 1240 |
+
"rewards/rejected": -0.2217450588941574,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.8201385736465454,
|
| 1247 |
+
"logits/rejected": -1.8143587112426758,
|
| 1248 |
+
"logps/chosen": -31.758071899414062,
|
| 1249 |
+
"logps/rejected": -31.750629425048828,
|
| 1250 |
+
"loss": 0.201,
|
| 1251 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1252 |
+
"rewards/chosen": 1.5750051736831665,
|
| 1253 |
+
"rewards/margins": 1.7468246221542358,
|
| 1254 |
+
"rewards/rejected": -0.17181925475597382,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.168851375579834,
|
| 1260 |
+
"eval_logits/rejected": -2.164125919342041,
|
| 1261 |
+
"eval_logps/chosen": -33.594791412353516,
|
| 1262 |
+
"eval_logps/rejected": -37.25070571899414,
|
| 1263 |
+
"eval_loss": 0.4766196310520172,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5913621187210083,
|
| 1265 |
+
"eval_rewards/chosen": 0.2638549208641052,
|
| 1266 |
+
"eval_rewards/margins": 0.10430653393268585,
|
| 1267 |
+
"eval_rewards/rejected": 0.15954840183258057,
|
| 1268 |
+
"eval_runtime": 145.9506,
|
| 1269 |
+
"eval_samples_per_second": 2.35,
|
| 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": -1.916377067565918,
|
| 1277 |
+
"logits/rejected": -1.923803687095642,
|
| 1278 |
+
"logps/chosen": -29.560104370117188,
|
| 1279 |
+
"logps/rejected": -34.730369567871094,
|
| 1280 |
+
"loss": 0.2043,
|
| 1281 |
+
"rewards/accuracies": 0.9375,
|
| 1282 |
+
"rewards/chosen": 1.5408909320831299,
|
| 1283 |
+
"rewards/margins": 1.8256099224090576,
|
| 1284 |
+
"rewards/rejected": -0.2847188711166382,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.8899390697479248,
|
| 1291 |
+
"logits/rejected": -1.8926613330841064,
|
| 1292 |
+
"logps/chosen": -30.270843505859375,
|
| 1293 |
+
"logps/rejected": -33.329734802246094,
|
| 1294 |
+
"loss": 0.2432,
|
| 1295 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1296 |
+
"rewards/chosen": 1.3801411390304565,
|
| 1297 |
+
"rewards/margins": 1.4400979280471802,
|
| 1298 |
+
"rewards/rejected": -0.05995672196149826,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.7909173965454102,
|
| 1305 |
+
"logits/rejected": -1.7950553894042969,
|
| 1306 |
+
"logps/chosen": -30.025737762451172,
|
| 1307 |
+
"logps/rejected": -35.363929748535156,
|
| 1308 |
+
"loss": 0.1984,
|
| 1309 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1310 |
+
"rewards/chosen": 1.5394319295883179,
|
| 1311 |
+
"rewards/margins": 1.8116801977157593,
|
| 1312 |
+
"rewards/rejected": -0.2722483277320862,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.8557007312774658,
|
| 1319 |
+
"logits/rejected": -1.851709008216858,
|
| 1320 |
+
"logps/chosen": -30.498058319091797,
|
| 1321 |
+
"logps/rejected": -33.070526123046875,
|
| 1322 |
+
"loss": 0.2262,
|
| 1323 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1324 |
+
"rewards/chosen": 1.482353925704956,
|
| 1325 |
+
"rewards/margins": 1.527881145477295,
|
| 1326 |
+
"rewards/rejected": -0.04552716761827469,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -1.9369919300079346,
|
| 1333 |
+
"logits/rejected": -1.9389537572860718,
|
| 1334 |
+
"logps/chosen": -30.1755428314209,
|
| 1335 |
+
"logps/rejected": -36.147056579589844,
|
| 1336 |
+
"loss": 0.1878,
|
| 1337 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1338 |
+
"rewards/chosen": 1.5598775148391724,
|
| 1339 |
+
"rewards/margins": 1.850951910018921,
|
| 1340 |
+
"rewards/rejected": -0.29107466340065,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.7436918020248413,
|
| 1347 |
+
"logits/rejected": -1.738072156906128,
|
| 1348 |
+
"logps/chosen": -33.61563491821289,
|
| 1349 |
+
"logps/rejected": -33.156700134277344,
|
| 1350 |
+
"loss": 0.2065,
|
| 1351 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1352 |
+
"rewards/chosen": 1.5996261835098267,
|
| 1353 |
+
"rewards/margins": 1.7585023641586304,
|
| 1354 |
+
"rewards/rejected": -0.15887616574764252,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.8033043146133423,
|
| 1361 |
+
"logits/rejected": -1.803333044052124,
|
| 1362 |
+
"logps/chosen": -33.05567169189453,
|
| 1363 |
+
"logps/rejected": -35.33915328979492,
|
| 1364 |
+
"loss": 0.1857,
|
| 1365 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1366 |
+
"rewards/chosen": 1.6722819805145264,
|
| 1367 |
+
"rewards/margins": 1.9680156707763672,
|
| 1368 |
+
"rewards/rejected": -0.29573339223861694,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.850695013999939,
|
| 1375 |
+
"logits/rejected": -1.839464783668518,
|
| 1376 |
+
"logps/chosen": -31.547618865966797,
|
| 1377 |
+
"logps/rejected": -33.39630126953125,
|
| 1378 |
+
"loss": 0.1969,
|
| 1379 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1380 |
+
"rewards/chosen": 1.535375714302063,
|
| 1381 |
+
"rewards/margins": 1.8199365139007568,
|
| 1382 |
+
"rewards/rejected": -0.2845609188079834,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -1.9542624950408936,
|
| 1389 |
+
"logits/rejected": -1.9587194919586182,
|
| 1390 |
+
"logps/chosen": -30.739770889282227,
|
| 1391 |
+
"logps/rejected": -33.01618194580078,
|
| 1392 |
+
"loss": 0.1947,
|
| 1393 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1394 |
+
"rewards/chosen": 1.5540999174118042,
|
| 1395 |
+
"rewards/margins": 1.9019180536270142,
|
| 1396 |
+
"rewards/rejected": -0.34781819581985474,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.8733972311019897,
|
| 1403 |
+
"logits/rejected": -1.8663629293441772,
|
| 1404 |
+
"logps/chosen": -31.99909019470215,
|
| 1405 |
+
"logps/rejected": -33.01835632324219,
|
| 1406 |
+
"loss": 0.1721,
|
| 1407 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1408 |
+
"rewards/chosen": 1.7129936218261719,
|
| 1409 |
+
"rewards/margins": 2.0827279090881348,
|
| 1410 |
+
"rewards/rejected": -0.3697340786457062,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.156324625015259,
|
| 1416 |
+
"eval_logits/rejected": -2.151609420776367,
|
| 1417 |
+
"eval_logps/chosen": -33.62693786621094,
|
| 1418 |
+
"eval_logps/rejected": -37.30033493041992,
|
| 1419 |
+
"eval_loss": 0.4748935401439667,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1421 |
+
"eval_rewards/chosen": 0.2445676475763321,
|
| 1422 |
+
"eval_rewards/margins": 0.11479608714580536,
|
| 1423 |
+
"eval_rewards/rejected": 0.12977157533168793,
|
| 1424 |
+
"eval_runtime": 145.9673,
|
| 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": -1.864915132522583,
|
| 1433 |
+
"logits/rejected": -1.8653583526611328,
|
| 1434 |
+
"logps/chosen": -28.593013763427734,
|
| 1435 |
+
"logps/rejected": -35.23602294921875,
|
| 1436 |
+
"loss": 0.1986,
|
| 1437 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1438 |
+
"rewards/chosen": 1.5444881916046143,
|
| 1439 |
+
"rewards/margins": 1.8722877502441406,
|
| 1440 |
+
"rewards/rejected": -0.3277994692325592,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.8232961893081665,
|
| 1447 |
+
"logits/rejected": -1.815349817276001,
|
| 1448 |
+
"logps/chosen": -29.518482208251953,
|
| 1449 |
+
"logps/rejected": -34.129737854003906,
|
| 1450 |
+
"loss": 0.2142,
|
| 1451 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1452 |
+
"rewards/chosen": 1.3922770023345947,
|
| 1453 |
+
"rewards/margins": 1.7272971868515015,
|
| 1454 |
+
"rewards/rejected": -0.3350200951099396,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -1.9893302917480469,
|
| 1461 |
+
"logits/rejected": -1.9897470474243164,
|
| 1462 |
+
"logps/chosen": -27.58512306213379,
|
| 1463 |
+
"logps/rejected": -33.1852912902832,
|
| 1464 |
+
"loss": 0.233,
|
| 1465 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1466 |
+
"rewards/chosen": 1.4471487998962402,
|
| 1467 |
+
"rewards/margins": 1.6394065618515015,
|
| 1468 |
+
"rewards/rejected": -0.1922578513622284,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.8695605993270874,
|
| 1475 |
+
"logits/rejected": -1.8790569305419922,
|
| 1476 |
+
"logps/chosen": -31.543869018554688,
|
| 1477 |
+
"logps/rejected": -33.519187927246094,
|
| 1478 |
+
"loss": 0.1757,
|
| 1479 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1480 |
+
"rewards/chosen": 1.7287185192108154,
|
| 1481 |
+
"rewards/margins": 2.046445369720459,
|
| 1482 |
+
"rewards/rejected": -0.31772682070732117,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -1.9113248586654663,
|
| 1489 |
+
"logits/rejected": -1.9151248931884766,
|
| 1490 |
+
"logps/chosen": -28.23110580444336,
|
| 1491 |
+
"logps/rejected": -32.094093322753906,
|
| 1492 |
+
"loss": 0.2093,
|
| 1493 |
+
"rewards/accuracies": 0.9375,
|
| 1494 |
+
"rewards/chosen": 1.4499690532684326,
|
| 1495 |
+
"rewards/margins": 1.664777398109436,
|
| 1496 |
+
"rewards/rejected": -0.21480807662010193,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -1.9074852466583252,
|
| 1503 |
+
"logits/rejected": -1.9036372900009155,
|
| 1504 |
+
"logps/chosen": -31.310367584228516,
|
| 1505 |
+
"logps/rejected": -34.39436721801758,
|
| 1506 |
+
"loss": 0.1936,
|
| 1507 |
+
"rewards/accuracies": 0.9375,
|
| 1508 |
+
"rewards/chosen": 1.6589105129241943,
|
| 1509 |
+
"rewards/margins": 1.951696753501892,
|
| 1510 |
+
"rewards/rejected": -0.2927861213684082,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.8242709636688232,
|
| 1517 |
+
"logits/rejected": -1.8268591165542603,
|
| 1518 |
+
"logps/chosen": -30.07489585876465,
|
| 1519 |
+
"logps/rejected": -35.65007400512695,
|
| 1520 |
+
"loss": 0.1939,
|
| 1521 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1522 |
+
"rewards/chosen": 1.642709493637085,
|
| 1523 |
+
"rewards/margins": 1.9237987995147705,
|
| 1524 |
+
"rewards/rejected": -0.28108924627304077,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.8587324619293213,
|
| 1531 |
+
"logits/rejected": -1.8523610830307007,
|
| 1532 |
+
"logps/chosen": -32.797035217285156,
|
| 1533 |
+
"logps/rejected": -36.79326248168945,
|
| 1534 |
+
"loss": 0.2006,
|
| 1535 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1536 |
+
"rewards/chosen": 1.564818263053894,
|
| 1537 |
+
"rewards/margins": 1.8836040496826172,
|
| 1538 |
+
"rewards/rejected": -0.3187856078147888,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.8334388732910156,
|
| 1545 |
+
"logits/rejected": -1.8283811807632446,
|
| 1546 |
+
"logps/chosen": -28.48984718322754,
|
| 1547 |
+
"logps/rejected": -36.4734001159668,
|
| 1548 |
+
"loss": 0.2097,
|
| 1549 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1550 |
+
"rewards/chosen": 1.5078611373901367,
|
| 1551 |
+
"rewards/margins": 1.7979707717895508,
|
| 1552 |
+
"rewards/rejected": -0.2901095449924469,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.8117061853408813,
|
| 1559 |
+
"logits/rejected": -1.8011939525604248,
|
| 1560 |
+
"logps/chosen": -29.4431095123291,
|
| 1561 |
+
"logps/rejected": -29.787906646728516,
|
| 1562 |
+
"loss": 0.2259,
|
| 1563 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1564 |
+
"rewards/chosen": 1.3833496570587158,
|
| 1565 |
+
"rewards/margins": 1.4962539672851562,
|
| 1566 |
+
"rewards/rejected": -0.11290428787469864,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.1527934074401855,
|
| 1572 |
+
"eval_logits/rejected": -2.1480958461761475,
|
| 1573 |
+
"eval_logps/chosen": -33.62074661254883,
|
| 1574 |
+
"eval_logps/rejected": -37.30718994140625,
|
| 1575 |
+
"eval_loss": 0.4736105799674988,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1577 |
+
"eval_rewards/chosen": 0.24828366935253143,
|
| 1578 |
+
"eval_rewards/margins": 0.12262485176324844,
|
| 1579 |
+
"eval_rewards/rejected": 0.12565883994102478,
|
| 1580 |
+
"eval_runtime": 145.9544,
|
| 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": -1.9208866357803345,
|
| 1589 |
+
"logits/rejected": -1.9203205108642578,
|
| 1590 |
+
"logps/chosen": -29.327899932861328,
|
| 1591 |
+
"logps/rejected": -30.58652114868164,
|
| 1592 |
+
"loss": 0.2156,
|
| 1593 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1594 |
+
"rewards/chosen": 1.4977772235870361,
|
| 1595 |
+
"rewards/margins": 1.6946022510528564,
|
| 1596 |
+
"rewards/rejected": -0.1968248337507248,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -1.9610830545425415,
|
| 1603 |
+
"logits/rejected": -1.9618806838989258,
|
| 1604 |
+
"logps/chosen": -29.420618057250977,
|
| 1605 |
+
"logps/rejected": -31.693790435791016,
|
| 1606 |
+
"loss": 0.1936,
|
| 1607 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1608 |
+
"rewards/chosen": 1.5137760639190674,
|
| 1609 |
+
"rewards/margins": 1.830562949180603,
|
| 1610 |
+
"rewards/rejected": -0.31678706407546997,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -1.9309966564178467,
|
| 1617 |
+
"logits/rejected": -1.934212327003479,
|
| 1618 |
+
"logps/chosen": -29.75461769104004,
|
| 1619 |
+
"logps/rejected": -28.163843154907227,
|
| 1620 |
+
"loss": 0.2199,
|
| 1621 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1622 |
+
"rewards/chosen": 1.5152724981307983,
|
| 1623 |
+
"rewards/margins": 1.6233495473861694,
|
| 1624 |
+
"rewards/rejected": -0.10807719081640244,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -1.9402587413787842,
|
| 1631 |
+
"logits/rejected": -1.9325793981552124,
|
| 1632 |
+
"logps/chosen": -28.6353702545166,
|
| 1633 |
+
"logps/rejected": -34.016441345214844,
|
| 1634 |
+
"loss": 0.1958,
|
| 1635 |
+
"rewards/accuracies": 0.9375,
|
| 1636 |
+
"rewards/chosen": 1.5255171060562134,
|
| 1637 |
+
"rewards/margins": 1.8352447748184204,
|
| 1638 |
+
"rewards/rejected": -0.30972760915756226,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.8774486780166626,
|
| 1645 |
+
"logits/rejected": -1.888318419456482,
|
| 1646 |
+
"logps/chosen": -30.820209503173828,
|
| 1647 |
+
"logps/rejected": -32.5506706237793,
|
| 1648 |
+
"loss": 0.1925,
|
| 1649 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1650 |
+
"rewards/chosen": 1.6397380828857422,
|
| 1651 |
+
"rewards/margins": 1.8231319189071655,
|
| 1652 |
+
"rewards/rejected": -0.18339397013187408,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -1.9557583332061768,
|
| 1659 |
+
"logits/rejected": -1.9465582370758057,
|
| 1660 |
+
"logps/chosen": -30.896408081054688,
|
| 1661 |
+
"logps/rejected": -36.84226608276367,
|
| 1662 |
+
"loss": 0.2111,
|
| 1663 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1664 |
+
"rewards/chosen": 1.5626221895217896,
|
| 1665 |
+
"rewards/margins": 1.8691476583480835,
|
| 1666 |
+
"rewards/rejected": -0.3065256178379059,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.8635246753692627,
|
| 1673 |
+
"logits/rejected": -1.8655227422714233,
|
| 1674 |
+
"logps/chosen": -33.10884475708008,
|
| 1675 |
+
"logps/rejected": -33.1629753112793,
|
| 1676 |
+
"loss": 0.1786,
|
| 1677 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1678 |
+
"rewards/chosen": 1.7282660007476807,
|
| 1679 |
+
"rewards/margins": 2.0073320865631104,
|
| 1680 |
+
"rewards/rejected": -0.2790663540363312,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -1.993586540222168,
|
| 1687 |
+
"logits/rejected": -1.9946495294570923,
|
| 1688 |
+
"logps/chosen": -30.365489959716797,
|
| 1689 |
+
"logps/rejected": -34.76932907104492,
|
| 1690 |
+
"loss": 0.1816,
|
| 1691 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1692 |
+
"rewards/chosen": 1.6152902841567993,
|
| 1693 |
+
"rewards/margins": 2.0074784755706787,
|
| 1694 |
+
"rewards/rejected": -0.39218825101852417,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -1.9641876220703125,
|
| 1701 |
+
"logits/rejected": -1.9684619903564453,
|
| 1702 |
+
"logps/chosen": -28.4173583984375,
|
| 1703 |
+
"logps/rejected": -33.92015838623047,
|
| 1704 |
+
"loss": 0.2133,
|
| 1705 |
+
"rewards/accuracies": 0.9375,
|
| 1706 |
+
"rewards/chosen": 1.4855743646621704,
|
| 1707 |
+
"rewards/margins": 1.7089710235595703,
|
| 1708 |
+
"rewards/rejected": -0.22339674830436707,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -1.9531924724578857,
|
| 1715 |
+
"logits/rejected": -1.958428144454956,
|
| 1716 |
+
"logps/chosen": -31.164684295654297,
|
| 1717 |
+
"logps/rejected": -30.430471420288086,
|
| 1718 |
+
"loss": 0.2405,
|
| 1719 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1720 |
+
"rewards/chosen": 1.4425894021987915,
|
| 1721 |
+
"rewards/margins": 1.459869623184204,
|
| 1722 |
+
"rewards/rejected": -0.01728021167218685,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.1522130966186523,
|
| 1728 |
+
"eval_logits/rejected": -2.147509813308716,
|
| 1729 |
+
"eval_logps/chosen": -33.62826156616211,
|
| 1730 |
+
"eval_logps/rejected": -37.31184005737305,
|
| 1731 |
+
"eval_loss": 0.4740113317966461,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1733 |
+
"eval_rewards/chosen": 0.24377261102199554,
|
| 1734 |
+
"eval_rewards/margins": 0.1209048256278038,
|
| 1735 |
+
"eval_rewards/rejected": 0.12286783009767532,
|
| 1736 |
+
"eval_runtime": 145.9558,
|
| 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": -1.9431841373443604,
|
| 1745 |
+
"logits/rejected": -1.9432331323623657,
|
| 1746 |
+
"logps/chosen": -30.208242416381836,
|
| 1747 |
+
"logps/rejected": -31.27047348022461,
|
| 1748 |
+
"loss": 0.1964,
|
| 1749 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1750 |
+
"rewards/chosen": 1.6391448974609375,
|
| 1751 |
+
"rewards/margins": 1.7918792963027954,
|
| 1752 |
+
"rewards/rejected": -0.15273427963256836,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.788575530052185,
|
| 1759 |
+
"logits/rejected": -1.7902095317840576,
|
| 1760 |
+
"logps/chosen": -31.8612117767334,
|
| 1761 |
+
"logps/rejected": -33.35196304321289,
|
| 1762 |
+
"loss": 0.2123,
|
| 1763 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1764 |
+
"rewards/chosen": 1.6222608089447021,
|
| 1765 |
+
"rewards/margins": 1.8086793422698975,
|
| 1766 |
+
"rewards/rejected": -0.18641847372055054,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.8251508474349976,
|
| 1773 |
+
"logits/rejected": -1.8224670886993408,
|
| 1774 |
+
"logps/chosen": -28.645374298095703,
|
| 1775 |
+
"logps/rejected": -31.369842529296875,
|
| 1776 |
+
"loss": 0.2402,
|
| 1777 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1778 |
+
"rewards/chosen": 1.4198086261749268,
|
| 1779 |
+
"rewards/margins": 1.542269229888916,
|
| 1780 |
+
"rewards/rejected": -0.12246046215295792,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -1.9626710414886475,
|
| 1787 |
+
"logits/rejected": -1.957177758216858,
|
| 1788 |
+
"logps/chosen": -30.216720581054688,
|
| 1789 |
+
"logps/rejected": -33.21830368041992,
|
| 1790 |
+
"loss": 0.1933,
|
| 1791 |
+
"rewards/accuracies": 0.9375,
|
| 1792 |
+
"rewards/chosen": 1.6938445568084717,
|
| 1793 |
+
"rewards/margins": 1.8524214029312134,
|
| 1794 |
+
"rewards/rejected": -0.15857680141925812,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.8174095153808594,
|
| 1801 |
+
"logits/rejected": -1.807636022567749,
|
| 1802 |
+
"logps/chosen": -30.805782318115234,
|
| 1803 |
+
"logps/rejected": -31.54193687438965,
|
| 1804 |
+
"loss": 0.2166,
|
| 1805 |
+
"rewards/accuracies": 0.9375,
|
| 1806 |
+
"rewards/chosen": 1.5875717401504517,
|
| 1807 |
+
"rewards/margins": 1.737291693687439,
|
| 1808 |
+
"rewards/rejected": -0.14972007274627686,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.8910198211669922,
|
| 1815 |
+
"logits/rejected": -1.8909051418304443,
|
| 1816 |
+
"logps/chosen": -30.962215423583984,
|
| 1817 |
+
"logps/rejected": -31.599491119384766,
|
| 1818 |
+
"loss": 0.1905,
|
| 1819 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1820 |
+
"rewards/chosen": 1.6317226886749268,
|
| 1821 |
+
"rewards/margins": 1.8705470561981201,
|
| 1822 |
+
"rewards/rejected": -0.23882456123828888,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -1.96527099609375,
|
| 1829 |
+
"logits/rejected": -1.9599376916885376,
|
| 1830 |
+
"logps/chosen": -30.861770629882812,
|
| 1831 |
+
"logps/rejected": -33.239105224609375,
|
| 1832 |
+
"loss": 0.1977,
|
| 1833 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1834 |
+
"rewards/chosen": 1.5985205173492432,
|
| 1835 |
+
"rewards/margins": 1.7772538661956787,
|
| 1836 |
+
"rewards/rejected": -0.17873358726501465,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.8232942819595337,
|
| 1843 |
+
"logits/rejected": -1.8309834003448486,
|
| 1844 |
+
"logps/chosen": -30.7927188873291,
|
| 1845 |
+
"logps/rejected": -32.76128387451172,
|
| 1846 |
+
"loss": 0.2011,
|
| 1847 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1848 |
+
"rewards/chosen": 1.5241663455963135,
|
| 1849 |
+
"rewards/margins": 1.7640721797943115,
|
| 1850 |
+
"rewards/rejected": -0.23990590870380402,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -1.8941972255706787,
|
| 1857 |
+
"logits/rejected": -1.8883616924285889,
|
| 1858 |
+
"logps/chosen": -29.16994857788086,
|
| 1859 |
+
"logps/rejected": -33.72978973388672,
|
| 1860 |
+
"loss": 0.1689,
|
| 1861 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1862 |
+
"rewards/chosen": 1.8021999597549438,
|
| 1863 |
+
"rewards/margins": 2.0700056552886963,
|
| 1864 |
+
"rewards/rejected": -0.26780563592910767,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.8379895687103271,
|
| 1871 |
+
"logits/rejected": -1.8400685787200928,
|
| 1872 |
+
"logps/chosen": -30.718734741210938,
|
| 1873 |
+
"logps/rejected": -33.35673141479492,
|
| 1874 |
+
"loss": 0.1793,
|
| 1875 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1876 |
+
"rewards/chosen": 1.6630252599716187,
|
| 1877 |
+
"rewards/margins": 2.0210342407226562,
|
| 1878 |
+
"rewards/rejected": -0.35800907015800476,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.151599884033203,
|
| 1884 |
+
"eval_logits/rejected": -2.1469039916992188,
|
| 1885 |
+
"eval_logps/chosen": -33.627723693847656,
|
| 1886 |
+
"eval_logps/rejected": -37.30845642089844,
|
| 1887 |
+
"eval_loss": 0.47458338737487793,
|
| 1888 |
+
"eval_rewards/accuracies": 0.5685215592384338,
|
| 1889 |
+
"eval_rewards/chosen": 0.24409595131874084,
|
| 1890 |
+
"eval_rewards/margins": 0.11919664591550827,
|
| 1891 |
+
"eval_rewards/rejected": 0.12489929050207138,
|
| 1892 |
+
"eval_runtime": 145.9719,
|
| 1893 |
+
"eval_samples_per_second": 2.35,
|
| 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": -1.7688223123550415,
|
| 1901 |
+
"logits/rejected": -1.7678744792938232,
|
| 1902 |
+
"logps/chosen": -33.03181076049805,
|
| 1903 |
+
"logps/rejected": -33.192596435546875,
|
| 1904 |
+
"loss": 0.1724,
|
| 1905 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1906 |
+
"rewards/chosen": 1.8124043941497803,
|
| 1907 |
+
"rewards/margins": 2.102344036102295,
|
| 1908 |
+
"rewards/rejected": -0.2899395823478699,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -1.9436089992523193,
|
| 1915 |
+
"logits/rejected": -1.9377737045288086,
|
| 1916 |
+
"logps/chosen": -28.891845703125,
|
| 1917 |
+
"logps/rejected": -33.52208709716797,
|
| 1918 |
+
"loss": 0.1785,
|
| 1919 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1920 |
+
"rewards/chosen": 1.6871049404144287,
|
| 1921 |
+
"rewards/margins": 2.0732340812683105,
|
| 1922 |
+
"rewards/rejected": -0.38612908124923706,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.79269278049469,
|
| 1929 |
+
"logits/rejected": -1.7812188863754272,
|
| 1930 |
+
"logps/chosen": -31.352039337158203,
|
| 1931 |
+
"logps/rejected": -33.33822250366211,
|
| 1932 |
+
"loss": 0.1708,
|
| 1933 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1934 |
+
"rewards/chosen": 1.820186972618103,
|
| 1935 |
+
"rewards/margins": 2.086826801300049,
|
| 1936 |
+
"rewards/rejected": -0.2666396200656891,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.8849557638168335,
|
| 1943 |
+
"logits/rejected": -1.878165602684021,
|
| 1944 |
+
"logps/chosen": -28.257898330688477,
|
| 1945 |
+
"logps/rejected": -29.390756607055664,
|
| 1946 |
+
"loss": 0.2132,
|
| 1947 |
+
"rewards/accuracies": 0.9375,
|
| 1948 |
+
"rewards/chosen": 1.4654014110565186,
|
| 1949 |
+
"rewards/margins": 1.7793182134628296,
|
| 1950 |
+
"rewards/rejected": -0.31391677260398865,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -1.9139467477798462,
|
| 1957 |
+
"logits/rejected": -1.9205329418182373,
|
| 1958 |
+
"logps/chosen": -32.439273834228516,
|
| 1959 |
+
"logps/rejected": -31.840866088867188,
|
| 1960 |
+
"loss": 0.1773,
|
| 1961 |
+
"rewards/accuracies": 0.9375,
|
| 1962 |
+
"rewards/chosen": 1.7736711502075195,
|
| 1963 |
+
"rewards/margins": 1.9672033786773682,
|
| 1964 |
+
"rewards/rejected": -0.19353218376636505,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.8897371292114258,
|
| 1971 |
+
"logits/rejected": -1.890363097190857,
|
| 1972 |
+
"logps/chosen": -30.296092987060547,
|
| 1973 |
+
"logps/rejected": -33.34063720703125,
|
| 1974 |
+
"loss": 0.1781,
|
| 1975 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1976 |
+
"rewards/chosen": 1.639479637145996,
|
| 1977 |
+
"rewards/margins": 1.9549386501312256,
|
| 1978 |
+
"rewards/rejected": -0.3154590129852295,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.8287382125854492,
|
| 1985 |
+
"logits/rejected": -1.8283493518829346,
|
| 1986 |
+
"logps/chosen": -30.095571517944336,
|
| 1987 |
+
"logps/rejected": -34.024879455566406,
|
| 1988 |
+
"loss": 0.1635,
|
| 1989 |
+
"rewards/accuracies": 0.9375,
|
| 1990 |
+
"rewards/chosen": 1.7971582412719727,
|
| 1991 |
+
"rewards/margins": 2.0997211933135986,
|
| 1992 |
+
"rewards/rejected": -0.30256301164627075,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.7129100561141968,
|
| 1999 |
+
"logits/rejected": -1.7082204818725586,
|
| 2000 |
+
"logps/chosen": -32.2062873840332,
|
| 2001 |
+
"logps/rejected": -31.325271606445312,
|
| 2002 |
+
"loss": 0.1593,
|
| 2003 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2004 |
+
"rewards/chosen": 1.818695068359375,
|
| 2005 |
+
"rewards/margins": 2.0579652786254883,
|
| 2006 |
+
"rewards/rejected": -0.2392701357603073,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -1.9185450077056885,
|
| 2013 |
+
"logits/rejected": -1.903119444847107,
|
| 2014 |
+
"logps/chosen": -31.7047119140625,
|
| 2015 |
+
"logps/rejected": -35.49399948120117,
|
| 2016 |
+
"loss": 0.1718,
|
| 2017 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2018 |
+
"rewards/chosen": 1.5857194662094116,
|
| 2019 |
+
"rewards/margins": 2.022928237915039,
|
| 2020 |
+
"rewards/rejected": -0.43720903992652893,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.8624118566513062,
|
| 2027 |
+
"logits/rejected": -1.8672168254852295,
|
| 2028 |
+
"logps/chosen": -30.548635482788086,
|
| 2029 |
+
"logps/rejected": -36.78308868408203,
|
| 2030 |
+
"loss": 0.1633,
|
| 2031 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2032 |
+
"rewards/chosen": 1.781390905380249,
|
| 2033 |
+
"rewards/margins": 2.1893134117126465,
|
| 2034 |
+
"rewards/rejected": -0.4079226553440094,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.151759624481201,
|
| 2040 |
+
"eval_logits/rejected": -2.14705753326416,
|
| 2041 |
+
"eval_logps/chosen": -33.628997802734375,
|
| 2042 |
+
"eval_logps/rejected": -37.31074523925781,
|
| 2043 |
+
"eval_loss": 0.4743698537349701,
|
| 2044 |
+
"eval_rewards/accuracies": 0.6009136438369751,
|
| 2045 |
+
"eval_rewards/chosen": 0.24333173036575317,
|
| 2046 |
+
"eval_rewards/margins": 0.11980798840522766,
|
| 2047 |
+
"eval_rewards/rejected": 0.12352374196052551,
|
| 2048 |
+
"eval_runtime": 145.9607,
|
| 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": -1.8351796865463257,
|
| 2057 |
+
"logits/rejected": -1.8353471755981445,
|
| 2058 |
+
"logps/chosen": -33.059471130371094,
|
| 2059 |
+
"logps/rejected": -37.527793884277344,
|
| 2060 |
+
"loss": 0.1742,
|
| 2061 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2062 |
+
"rewards/chosen": 1.9107275009155273,
|
| 2063 |
+
"rewards/margins": 2.176933765411377,
|
| 2064 |
+
"rewards/rejected": -0.26620611548423767,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -1.9216110706329346,
|
| 2071 |
+
"logits/rejected": -1.9202417135238647,
|
| 2072 |
+
"logps/chosen": -29.405132293701172,
|
| 2073 |
+
"logps/rejected": -31.430126190185547,
|
| 2074 |
+
"loss": 0.2013,
|
| 2075 |
+
"rewards/accuracies": 0.9375,
|
| 2076 |
+
"rewards/chosen": 1.6132328510284424,
|
| 2077 |
+
"rewards/margins": 1.8992938995361328,
|
| 2078 |
+
"rewards/rejected": -0.2860611379146576,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.8282123804092407,
|
| 2085 |
+
"logits/rejected": -1.829431176185608,
|
| 2086 |
+
"logps/chosen": -31.4184627532959,
|
| 2087 |
+
"logps/rejected": -34.05662155151367,
|
| 2088 |
+
"loss": 0.181,
|
| 2089 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2090 |
+
"rewards/chosen": 1.773472785949707,
|
| 2091 |
+
"rewards/margins": 2.0182414054870605,
|
| 2092 |
+
"rewards/rejected": -0.24476829171180725,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.8872970342636108,
|
| 2099 |
+
"logits/rejected": -1.8860971927642822,
|
| 2100 |
+
"logps/chosen": -28.597339630126953,
|
| 2101 |
+
"logps/rejected": -32.5714225769043,
|
| 2102 |
+
"loss": 0.1828,
|
| 2103 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2104 |
+
"rewards/chosen": 1.5607964992523193,
|
| 2105 |
+
"rewards/margins": 1.908465027809143,
|
| 2106 |
+
"rewards/rejected": -0.34766843914985657,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -1.9047062397003174,
|
| 2113 |
+
"logits/rejected": -1.890319585800171,
|
| 2114 |
+
"logps/chosen": -31.581689834594727,
|
| 2115 |
+
"logps/rejected": -36.782928466796875,
|
| 2116 |
+
"loss": 0.1839,
|
| 2117 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2118 |
+
"rewards/chosen": 1.6409885883331299,
|
| 2119 |
+
"rewards/margins": 1.9520866870880127,
|
| 2120 |
+
"rewards/rejected": -0.31109803915023804,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -1.9034063816070557,
|
| 2127 |
+
"logits/rejected": -1.9158471822738647,
|
| 2128 |
+
"logps/chosen": -29.19135093688965,
|
| 2129 |
+
"logps/rejected": -32.5649299621582,
|
| 2130 |
+
"loss": 0.1788,
|
| 2131 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2132 |
+
"rewards/chosen": 1.726271390914917,
|
| 2133 |
+
"rewards/margins": 1.921381950378418,
|
| 2134 |
+
"rewards/rejected": -0.19511063396930695,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -1.9984302520751953,
|
| 2141 |
+
"logits/rejected": -1.9977290630340576,
|
| 2142 |
+
"logps/chosen": -29.15370750427246,
|
| 2143 |
+
"logps/rejected": -36.60697555541992,
|
| 2144 |
+
"loss": 0.1649,
|
| 2145 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2146 |
+
"rewards/chosen": 1.7260326147079468,
|
| 2147 |
+
"rewards/margins": 2.191688060760498,
|
| 2148 |
+
"rewards/rejected": -0.46565547585487366,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.863462209701538,
|
| 2155 |
+
"logits/rejected": -1.8711704015731812,
|
| 2156 |
+
"logps/chosen": -31.19002914428711,
|
| 2157 |
+
"logps/rejected": -37.98977279663086,
|
| 2158 |
+
"loss": 0.183,
|
| 2159 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2160 |
+
"rewards/chosen": 1.7942688465118408,
|
| 2161 |
+
"rewards/margins": 2.118346929550171,
|
| 2162 |
+
"rewards/rejected": -0.3240783214569092,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.8185231685638428,
|
| 2169 |
+
"logits/rejected": -1.8144798278808594,
|
| 2170 |
+
"logps/chosen": -31.748987197875977,
|
| 2171 |
+
"logps/rejected": -31.824060440063477,
|
| 2172 |
+
"loss": 0.1855,
|
| 2173 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2174 |
+
"rewards/chosen": 1.8109363317489624,
|
| 2175 |
+
"rewards/margins": 1.9469363689422607,
|
| 2176 |
+
"rewards/rejected": -0.1359998881816864,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -1.9791545867919922,
|
| 2183 |
+
"logits/rejected": -1.9766286611557007,
|
| 2184 |
+
"logps/chosen": -28.732391357421875,
|
| 2185 |
+
"logps/rejected": -32.56377410888672,
|
| 2186 |
+
"loss": 0.202,
|
| 2187 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2188 |
+
"rewards/chosen": 1.5384950637817383,
|
| 2189 |
+
"rewards/margins": 1.790077567100525,
|
| 2190 |
+
"rewards/rejected": -0.25158247351646423,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.1518642902374268,
|
| 2196 |
+
"eval_logits/rejected": -2.147162437438965,
|
| 2197 |
+
"eval_logps/chosen": -33.62628173828125,
|
| 2198 |
+
"eval_logps/rejected": -37.30340576171875,
|
| 2199 |
+
"eval_loss": 0.4747585952281952,
|
| 2200 |
+
"eval_rewards/accuracies": 0.5830564498901367,
|
| 2201 |
+
"eval_rewards/chosen": 0.24496087431907654,
|
| 2202 |
+
"eval_rewards/margins": 0.11703355610370636,
|
| 2203 |
+
"eval_rewards/rejected": 0.1279272884130478,
|
| 2204 |
+
"eval_runtime": 145.9829,
|
| 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": -1.9209346771240234,
|
| 2213 |
+
"logits/rejected": -1.916852355003357,
|
| 2214 |
+
"logps/chosen": -30.77335548400879,
|
| 2215 |
+
"logps/rejected": -29.9306697845459,
|
| 2216 |
+
"loss": 0.1787,
|
| 2217 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2218 |
+
"rewards/chosen": 1.7215063571929932,
|
| 2219 |
+
"rewards/margins": 1.947231650352478,
|
| 2220 |
+
"rewards/rejected": -0.22572548687458038,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.878213882446289,
|
| 2227 |
+
"logits/rejected": -1.8793401718139648,
|
| 2228 |
+
"logps/chosen": -30.57177734375,
|
| 2229 |
+
"logps/rejected": -35.1507682800293,
|
| 2230 |
+
"loss": 0.1981,
|
| 2231 |
+
"rewards/accuracies": 0.9375,
|
| 2232 |
+
"rewards/chosen": 1.6758613586425781,
|
| 2233 |
+
"rewards/margins": 1.8119198083877563,
|
| 2234 |
+
"rewards/rejected": -0.13605818152427673,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.8281654119491577,
|
| 2241 |
+
"logits/rejected": -1.8215233087539673,
|
| 2242 |
+
"logps/chosen": -31.386241912841797,
|
| 2243 |
+
"logps/rejected": -37.23295211791992,
|
| 2244 |
+
"loss": 0.1849,
|
| 2245 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2246 |
+
"rewards/chosen": 1.653316855430603,
|
| 2247 |
+
"rewards/margins": 1.9794448614120483,
|
| 2248 |
+
"rewards/rejected": -0.3261282444000244,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.8422043323516846,
|
| 2255 |
+
"logits/rejected": -1.836806297302246,
|
| 2256 |
+
"logps/chosen": -26.592382431030273,
|
| 2257 |
+
"logps/rejected": -32.13856887817383,
|
| 2258 |
+
"loss": 0.1895,
|
| 2259 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2260 |
+
"rewards/chosen": 1.6325156688690186,
|
| 2261 |
+
"rewards/margins": 1.8772624731063843,
|
| 2262 |
+
"rewards/rejected": -0.2447468340396881,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.8666261434555054,
|
| 2269 |
+
"logits/rejected": -1.8564945459365845,
|
| 2270 |
+
"logps/chosen": -30.06821632385254,
|
| 2271 |
+
"logps/rejected": -36.11262130737305,
|
| 2272 |
+
"loss": 0.1911,
|
| 2273 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2274 |
+
"rewards/chosen": 1.4479871988296509,
|
| 2275 |
+
"rewards/margins": 1.9443438053131104,
|
| 2276 |
+
"rewards/rejected": -0.49635666608810425,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.8395572900772095,
|
| 2283 |
+
"logits/rejected": -1.836773157119751,
|
| 2284 |
+
"logps/chosen": -30.69950294494629,
|
| 2285 |
+
"logps/rejected": -31.030874252319336,
|
| 2286 |
+
"loss": 0.21,
|
| 2287 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2288 |
+
"rewards/chosen": 1.5661640167236328,
|
| 2289 |
+
"rewards/margins": 1.6501954793930054,
|
| 2290 |
+
"rewards/rejected": -0.08403147757053375,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -1.9007151126861572,
|
| 2297 |
+
"logits/rejected": -1.9071681499481201,
|
| 2298 |
+
"logps/chosen": -29.030803680419922,
|
| 2299 |
+
"logps/rejected": -33.39077377319336,
|
| 2300 |
+
"loss": 0.1701,
|
| 2301 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2302 |
+
"rewards/chosen": 1.8230218887329102,
|
| 2303 |
+
"rewards/margins": 2.0800347328186035,
|
| 2304 |
+
"rewards/rejected": -0.25701290369033813,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -1.9018220901489258,
|
| 2311 |
+
"logits/rejected": -1.9121367931365967,
|
| 2312 |
+
"logps/chosen": -28.98052406311035,
|
| 2313 |
+
"logps/rejected": -30.033344268798828,
|
| 2314 |
+
"loss": 0.184,
|
| 2315 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2316 |
+
"rewards/chosen": 1.628839135169983,
|
| 2317 |
+
"rewards/margins": 1.8717044591903687,
|
| 2318 |
+
"rewards/rejected": -0.24286529421806335,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.8325939178466797,
|
| 2325 |
+
"logits/rejected": -1.8257277011871338,
|
| 2326 |
+
"logps/chosen": -31.01336669921875,
|
| 2327 |
+
"logps/rejected": -32.70115280151367,
|
| 2328 |
+
"loss": 0.188,
|
| 2329 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2330 |
+
"rewards/chosen": 1.5858715772628784,
|
| 2331 |
+
"rewards/margins": 1.8802779912948608,
|
| 2332 |
+
"rewards/rejected": -0.29440635442733765,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.8861255645751953,
|
| 2339 |
+
"logits/rejected": -1.8828990459442139,
|
| 2340 |
+
"logps/chosen": -30.904102325439453,
|
| 2341 |
+
"logps/rejected": -33.84788513183594,
|
| 2342 |
+
"loss": 0.1889,
|
| 2343 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2344 |
+
"rewards/chosen": 1.6328811645507812,
|
| 2345 |
+
"rewards/margins": 1.9854776859283447,
|
| 2346 |
+
"rewards/rejected": -0.3525964617729187,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.1516811847686768,
|
| 2352 |
+
"eval_logits/rejected": -2.146984577178955,
|
| 2353 |
+
"eval_logps/chosen": -33.62123107910156,
|
| 2354 |
+
"eval_logps/rejected": -37.318565368652344,
|
| 2355 |
+
"eval_loss": 0.4726825952529907,
|
| 2356 |
+
"eval_rewards/accuracies": 0.6004983186721802,
|
| 2357 |
+
"eval_rewards/chosen": 0.24799373745918274,
|
| 2358 |
+
"eval_rewards/margins": 0.12916025519371033,
|
| 2359 |
+
"eval_rewards/rejected": 0.11883345246315002,
|
| 2360 |
+
"eval_runtime": 145.8433,
|
| 2361 |
+
"eval_samples_per_second": 2.352,
|
| 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": -1.9646100997924805,
|
| 2369 |
+
"logits/rejected": -1.9652503728866577,
|
| 2370 |
+
"logps/chosen": -27.146465301513672,
|
| 2371 |
+
"logps/rejected": -33.218292236328125,
|
| 2372 |
+
"loss": 0.2041,
|
| 2373 |
+
"rewards/accuracies": 0.875,
|
| 2374 |
+
"rewards/chosen": 1.6018869876861572,
|
| 2375 |
+
"rewards/margins": 1.8885917663574219,
|
| 2376 |
+
"rewards/rejected": -0.2867048680782318,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.8171708583831787,
|
| 2383 |
+
"logits/rejected": -1.8171415328979492,
|
| 2384 |
+
"logps/chosen": -30.371408462524414,
|
| 2385 |
+
"logps/rejected": -34.34622573852539,
|
| 2386 |
+
"loss": 0.1949,
|
| 2387 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2388 |
+
"rewards/chosen": 1.6022796630859375,
|
| 2389 |
+
"rewards/margins": 1.885776162147522,
|
| 2390 |
+
"rewards/rejected": -0.2834966778755188,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -1.9699510335922241,
|
| 2397 |
+
"logits/rejected": -1.9673391580581665,
|
| 2398 |
+
"logps/chosen": -28.992298126220703,
|
| 2399 |
+
"logps/rejected": -32.012474060058594,
|
| 2400 |
+
"loss": 0.2043,
|
| 2401 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2402 |
+
"rewards/chosen": 1.489058256149292,
|
| 2403 |
+
"rewards/margins": 1.806349754333496,
|
| 2404 |
+
"rewards/rejected": -0.3172914385795593,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -1.95291268825531,
|
| 2411 |
+
"logits/rejected": -1.9540326595306396,
|
| 2412 |
+
"logps/chosen": -28.15656089782715,
|
| 2413 |
+
"logps/rejected": -30.210216522216797,
|
| 2414 |
+
"loss": 0.204,
|
| 2415 |
+
"rewards/accuracies": 0.970833420753479,
|
| 2416 |
+
"rewards/chosen": 1.4780277013778687,
|
| 2417 |
+
"rewards/margins": 1.730049729347229,
|
| 2418 |
+
"rewards/rejected": -0.2520219683647156,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 0.19402528018146367,
|
| 2426 |
+
"train_runtime": 10806.8766,
|
| 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,
|