Commit
•
7b32e49
1
Parent(s):
5ef7aa6
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: -
|
| 29 |
-
- Logps/chosen: -34.
|
| 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 |
-
| 0.6538 | 0.26 | 100 |
|
| 67 |
-
| 0.6015 | 0.52 | 200 |
|
| 68 |
-
| 0.5385 | 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: 0.6820
|
| 22 |
+
- Rewards/chosen: -0.2813
|
| 23 |
+
- Rewards/rejected: -0.4578
|
| 24 |
+
- Rewards/accuracies: 0.5743
|
| 25 |
+
- Rewards/margins: 0.1765
|
| 26 |
+
- Logps/rejected: -38.0889
|
| 27 |
+
- Logps/chosen: -34.3862
|
| 28 |
+
- Logits/rejected: -2.1265
|
| 29 |
+
- Logits/chosen: -2.1312
|
| 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 |
+
| 0.6538 | 0.26 | 100 | -2.2339 | -2.2290 | -34.0302 | -37.5273 | 0.6955 | 0.5108 | 0.0035 | 0.0120 | -0.0085 |
|
| 65 |
+
| 0.6015 | 0.52 | 200 | -2.2322 | -2.2274 | -34.0607 | -37.5657 | 0.6956 | 0.5249 | -0.0209 | 0.0183 | -0.0393 |
|
| 66 |
+
| 0.5385 | 0.78 | 300 | -2.2294 | -2.2246 | -34.0909 | -37.5973 | 0.6957 | 0.5399 | -0.0451 | 0.0194 | -0.0645 |
|
| 67 |
+
| 0.4017 | 1.04 | 400 | 0.6775 | -0.0241 | -0.0945 | 0.6005 | 0.0703 | -37.6347 | -34.0647 | -2.2052 | -2.2100 |
|
| 68 |
+
| 0.3216 | 1.3 | 500 | 0.6820 | -0.1165 | -0.2131 | 0.5768 | 0.0966 | -37.7830 | -34.1802 | -2.1659 | -2.1707 |
|
| 69 |
+
| 0.336 | 1.56 | 600 | 0.6618 | -0.1839 | -0.3388 | 0.6242 | 0.1549 | -37.9401 | -34.2644 | -2.1554 | -2.1602 |
|
| 70 |
+
| 0.3559 | 1.82 | 700 | 0.6947 | -0.2571 | -0.3713 | 0.5341 | 0.1141 | -37.9807 | -34.3560 | -2.1535 | -2.1583 |
|
| 71 |
+
| 0.1978 | 2.08 | 800 | 0.6838 | -0.2324 | -0.3669 | 0.5835 | 0.1345 | -37.9753 | -34.3250 | -2.1501 | -2.1549 |
|
| 72 |
+
| 0.1619 | 2.34 | 900 | 0.6788 | -0.2463 | -0.4156 | 0.5860 | 0.1693 | -38.0361 | -34.3424 | -2.1384 | -2.1431 |
|
| 73 |
+
| 0.209 | 2.6 | 1000 | 0.6777 | -0.2767 | -0.4535 | 0.5918 | 0.1767 | -38.0835 | -34.3805 | -2.1309 | -2.1357 |
|
| 74 |
+
| 0.2513 | 2.86 | 1100 | 0.6897 | -0.2986 | -0.4591 | 0.5831 | 0.1605 | -38.0905 | -34.4077 | -2.1270 | -2.1317 |
|
| 75 |
+
| 0.1713 | 3.12 | 1200 | 0.6780 | -0.2775 | -0.4614 | 0.5947 | 0.1839 | -38.0934 | -34.3814 | -2.1270 | -2.1317 |
|
| 76 |
+
| 0.1199 | 3.38 | 1300 | 0.6740 | -0.2726 | -0.4645 | 0.5980 | 0.1919 | -38.0972 | -34.3753 | -2.1269 | -2.1317 |
|
| 77 |
+
| 0.1578 | 3.64 | 1400 | 0.6839 | -0.2867 | -0.4600 | 0.5860 | 0.1734 | -38.0917 | -34.3929 | -2.1266 | -2.1314 |
|
| 78 |
+
| 0.1614 | 3.9 | 1500 | 0.6820 | -0.2813 | -0.4578 | 0.5743 | 0.1765 | -38.0889 | -34.3862 | -2.1265 | -2.1312 |
|
| 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 sha256:1e88918600a30b4ce059435ac14ae95a63b5112549f43a93d66811978ceeb9d6
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.22914457321167,
|
| 4 |
"eval_logits/rejected": -2.224289894104004,
|
| 5 |
"eval_logps/chosen": -34.094017028808594,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.358,
|
| 15 |
"eval_steps_per_second": 0.296,
|
| 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": -2.22914457321167,
|
| 4 |
"eval_logits/rejected": -2.224289894104004,
|
| 5 |
"eval_logps/chosen": -34.094017028808594,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.358,
|
| 15 |
"eval_steps_per_second": 0.296,
|
| 16 |
+
"train_loss": 0.20605388523696305,
|
| 17 |
+
"train_runtime": 10804.1609,
|
| 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": 0.20605388523696305,
|
| 4 |
+
"train_runtime": 10804.1609,
|
| 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": -1.
|
| 526 |
-
"logps/chosen": -33.
|
| 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": -33.
|
| 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": -33.
|
| 583 |
-
"logps/rejected": -29.
|
| 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": -1.9146509170532227,
|
| 497 |
+
"logits/rejected": -1.9113744497299194,
|
| 498 |
+
"logps/chosen": -31.32379150390625,
|
| 499 |
+
"logps/rejected": -33.805755615234375,
|
| 500 |
+
"loss": 0.5978,
|
| 501 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 502 |
+
"rewards/chosen": 0.2236466407775879,
|
| 503 |
+
"rewards/margins": 0.27452850341796875,
|
| 504 |
+
"rewards/rejected": -0.050881873816251755,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.9630565643310547,
|
| 511 |
+
"logits/rejected": -1.9508411884307861,
|
| 512 |
+
"logps/chosen": -34.32460403442383,
|
| 513 |
+
"logps/rejected": -33.67837905883789,
|
| 514 |
+
"loss": 0.5713,
|
| 515 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 516 |
+
"rewards/chosen": 0.217172771692276,
|
| 517 |
+
"rewards/margins": 0.3288002014160156,
|
| 518 |
+
"rewards/rejected": -0.1116274744272232,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -1.9971084594726562,
|
| 525 |
+
"logits/rejected": -1.995650053024292,
|
| 526 |
+
"logps/chosen": -33.14687728881836,
|
| 527 |
+
"logps/rejected": -32.54724884033203,
|
| 528 |
+
"loss": 0.5766,
|
| 529 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 530 |
+
"rewards/chosen": 0.2711312174797058,
|
| 531 |
+
"rewards/margins": 0.32526224851608276,
|
| 532 |
+
"rewards/rejected": -0.05413101240992546,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.0822432041168213,
|
| 539 |
+
"logits/rejected": -2.066549301147461,
|
| 540 |
+
"logps/chosen": -33.73598098754883,
|
| 541 |
+
"logps/rejected": -33.096824645996094,
|
| 542 |
+
"loss": 0.572,
|
| 543 |
+
"rewards/accuracies": 0.699999988079071,
|
| 544 |
+
"rewards/chosen": 0.35473155975341797,
|
| 545 |
+
"rewards/margins": 0.33776286244392395,
|
| 546 |
+
"rewards/rejected": 0.016968706622719765,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.9550272226333618,
|
| 553 |
+
"logits/rejected": -1.9541677236557007,
|
| 554 |
+
"logps/chosen": -32.79063415527344,
|
| 555 |
+
"logps/rejected": -32.5118293762207,
|
| 556 |
+
"loss": 0.5561,
|
| 557 |
+
"rewards/accuracies": 0.75,
|
| 558 |
+
"rewards/chosen": 0.3931701183319092,
|
| 559 |
+
"rewards/margins": 0.43051546812057495,
|
| 560 |
+
"rewards/rejected": -0.03734535723924637,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.9071505069732666,
|
| 567 |
+
"logits/rejected": -1.917394995689392,
|
| 568 |
+
"logps/chosen": -31.765522003173828,
|
| 569 |
+
"logps/rejected": -35.38612365722656,
|
| 570 |
+
"loss": 0.5497,
|
| 571 |
+
"rewards/accuracies": 0.75,
|
| 572 |
+
"rewards/chosen": 0.35133686661720276,
|
| 573 |
+
"rewards/margins": 0.4292295575141907,
|
| 574 |
+
"rewards/rejected": -0.07789268344640732,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.042468547821045,
|
| 581 |
+
"logits/rejected": -2.0360734462738037,
|
| 582 |
+
"logps/chosen": -33.295223236083984,
|
| 583 |
+
"logps/rejected": -29.314971923828125,
|
| 584 |
+
"loss": 0.5615,
|
| 585 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 586 |
+
"rewards/chosen": 0.28539514541625977,
|
| 587 |
+
"rewards/margins": 0.35623133182525635,
|
| 588 |
+
"rewards/rejected": -0.07083616405725479,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.9002101421356201,
|
| 595 |
+
"logits/rejected": -1.9024240970611572,
|
| 596 |
+
"logps/chosen": -33.61799240112305,
|
| 597 |
+
"logps/rejected": -30.948089599609375,
|
| 598 |
+
"loss": 0.5044,
|
| 599 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 600 |
+
"rewards/chosen": 0.49684303998947144,
|
| 601 |
+
"rewards/margins": 0.5572139024734497,
|
| 602 |
+
"rewards/rejected": -0.06037076562643051,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9271265268325806,
|
| 609 |
+
"logits/rejected": -1.9258983135223389,
|
| 610 |
+
"logps/chosen": -33.7630500793457,
|
| 611 |
+
"logps/rejected": -35.974002838134766,
|
| 612 |
+
"loss": 0.4531,
|
| 613 |
+
"rewards/accuracies": 0.7958333492279053,
|
| 614 |
+
"rewards/chosen": 0.4973909258842468,
|
| 615 |
+
"rewards/margins": 0.7517890930175781,
|
| 616 |
+
"rewards/rejected": -0.2543981671333313,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.8622478246688843,
|
| 623 |
+
"logits/rejected": -1.8538604974746704,
|
| 624 |
+
"logps/chosen": -31.049022674560547,
|
| 625 |
+
"logps/rejected": -36.37419509887695,
|
| 626 |
+
"loss": 0.4017,
|
| 627 |
+
"rewards/accuracies": 0.862500011920929,
|
| 628 |
+
"rewards/chosen": 0.5799586772918701,
|
| 629 |
+
"rewards/margins": 0.9268311262130737,
|
| 630 |
+
"rewards/rejected": -0.34687256813049316,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.2100114822387695,
|
| 636 |
+
"eval_logits/rejected": -2.205153226852417,
|
| 637 |
+
"eval_logps/chosen": -34.06473159790039,
|
| 638 |
+
"eval_logps/rejected": -37.634727478027344,
|
| 639 |
+
"eval_loss": 0.6775211691856384,
|
| 640 |
+
"eval_rewards/accuracies": 0.6004983186721802,
|
| 641 |
+
"eval_rewards/chosen": -0.024144915863871574,
|
| 642 |
+
"eval_rewards/margins": 0.07034175843000412,
|
| 643 |
+
"eval_rewards/rejected": -0.09448666870594025,
|
| 644 |
+
"eval_runtime": 146.4012,
|
| 645 |
+
"eval_samples_per_second": 2.343,
|
| 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": -2.0343263149261475,
|
| 653 |
+
"logits/rejected": -2.037248134613037,
|
| 654 |
+
"logps/chosen": -32.27507019042969,
|
| 655 |
+
"logps/rejected": -35.283382415771484,
|
| 656 |
+
"loss": 0.3721,
|
| 657 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 658 |
+
"rewards/chosen": 0.6230968236923218,
|
| 659 |
+
"rewards/margins": 1.0155293941497803,
|
| 660 |
+
"rewards/rejected": -0.3924326002597809,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.8720476627349854,
|
| 667 |
+
"logits/rejected": -1.8706938028335571,
|
| 668 |
+
"logps/chosen": -28.4278507232666,
|
| 669 |
+
"logps/rejected": -32.78199005126953,
|
| 670 |
+
"loss": 0.3947,
|
| 671 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 672 |
+
"rewards/chosen": 0.5117626786231995,
|
| 673 |
+
"rewards/margins": 0.8832231760025024,
|
| 674 |
+
"rewards/rejected": -0.371460497379303,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.8284943103790283,
|
| 681 |
+
"logits/rejected": -1.821554183959961,
|
| 682 |
+
"logps/chosen": -33.15713119506836,
|
| 683 |
+
"logps/rejected": -34.481746673583984,
|
| 684 |
+
"loss": 0.3744,
|
| 685 |
+
"rewards/accuracies": 0.949999988079071,
|
| 686 |
+
"rewards/chosen": 0.733561635017395,
|
| 687 |
+
"rewards/margins": 0.9227407574653625,
|
| 688 |
+
"rewards/rejected": -0.18917915225028992,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -1.998443841934204,
|
| 695 |
+
"logits/rejected": -1.993316888809204,
|
| 696 |
+
"logps/chosen": -30.946176528930664,
|
| 697 |
+
"logps/rejected": -32.4330940246582,
|
| 698 |
+
"loss": 0.3658,
|
| 699 |
+
"rewards/accuracies": 0.925000011920929,
|
| 700 |
+
"rewards/chosen": 0.669696033000946,
|
| 701 |
+
"rewards/margins": 1.0420799255371094,
|
| 702 |
+
"rewards/rejected": -0.3723840117454529,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.8814995288848877,
|
| 709 |
+
"logits/rejected": -1.8857187032699585,
|
| 710 |
+
"logps/chosen": -33.60875701904297,
|
| 711 |
+
"logps/rejected": -33.16690444946289,
|
| 712 |
+
"loss": 0.3134,
|
| 713 |
+
"rewards/accuracies": 0.9375,
|
| 714 |
+
"rewards/chosen": 0.7256715297698975,
|
| 715 |
+
"rewards/margins": 1.1882824897766113,
|
| 716 |
+
"rewards/rejected": -0.4626109004020691,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.8870223760604858,
|
| 723 |
+
"logits/rejected": -1.8813492059707642,
|
| 724 |
+
"logps/chosen": -34.445613861083984,
|
| 725 |
+
"logps/rejected": -35.65070343017578,
|
| 726 |
+
"loss": 0.3011,
|
| 727 |
+
"rewards/accuracies": 0.949999988079071,
|
| 728 |
+
"rewards/chosen": 0.779128909111023,
|
| 729 |
+
"rewards/margins": 1.3404181003570557,
|
| 730 |
+
"rewards/rejected": -0.5612891316413879,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.9110435247421265,
|
| 737 |
+
"logits/rejected": -1.910775899887085,
|
| 738 |
+
"logps/chosen": -33.23371887207031,
|
| 739 |
+
"logps/rejected": -34.685794830322266,
|
| 740 |
+
"loss": 0.3634,
|
| 741 |
+
"rewards/accuracies": 0.925000011920929,
|
| 742 |
+
"rewards/chosen": 0.747688353061676,
|
| 743 |
+
"rewards/margins": 1.1479377746582031,
|
| 744 |
+
"rewards/rejected": -0.40024933218955994,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -1.9573456048965454,
|
| 751 |
+
"logits/rejected": -1.938490867614746,
|
| 752 |
+
"logps/chosen": -33.21842575073242,
|
| 753 |
+
"logps/rejected": -33.71406555175781,
|
| 754 |
+
"loss": 0.3363,
|
| 755 |
+
"rewards/accuracies": 0.925000011920929,
|
| 756 |
+
"rewards/chosen": 0.753872275352478,
|
| 757 |
+
"rewards/margins": 1.2691444158554077,
|
| 758 |
+
"rewards/rejected": -0.5152722001075745,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.9228092432022095,
|
| 765 |
+
"logits/rejected": -1.9221388101577759,
|
| 766 |
+
"logps/chosen": -30.777042388916016,
|
| 767 |
+
"logps/rejected": -31.67429542541504,
|
| 768 |
+
"loss": 0.3684,
|
| 769 |
+
"rewards/accuracies": 0.925000011920929,
|
| 770 |
+
"rewards/chosen": 0.6674107313156128,
|
| 771 |
+
"rewards/margins": 1.0100107192993164,
|
| 772 |
+
"rewards/rejected": -0.34259992837905884,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.8837568759918213,
|
| 779 |
+
"logits/rejected": -1.8767935037612915,
|
| 780 |
+
"logps/chosen": -31.731576919555664,
|
| 781 |
+
"logps/rejected": -35.3281364440918,
|
| 782 |
+
"loss": 0.3216,
|
| 783 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 784 |
+
"rewards/chosen": 0.7508381605148315,
|
| 785 |
+
"rewards/margins": 1.3119536638259888,
|
| 786 |
+
"rewards/rejected": -0.5611153841018677,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.170705556869507,
|
| 792 |
+
"eval_logits/rejected": -2.165867567062378,
|
| 793 |
+
"eval_logps/chosen": -34.18019104003906,
|
| 794 |
+
"eval_logps/rejected": -37.78302764892578,
|
| 795 |
+
"eval_loss": 0.6819708347320557,
|
| 796 |
+
"eval_rewards/accuracies": 0.5768272280693054,
|
| 797 |
+
"eval_rewards/chosen": -0.11651215702295303,
|
| 798 |
+
"eval_rewards/margins": 0.09661299735307693,
|
| 799 |
+
"eval_rewards/rejected": -0.21312516927719116,
|
| 800 |
+
"eval_runtime": 145.937,
|
| 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": -1.8785709142684937,
|
| 809 |
+
"logits/rejected": -1.8822866678237915,
|
| 810 |
+
"logps/chosen": -28.374109268188477,
|
| 811 |
+
"logps/rejected": -33.53699493408203,
|
| 812 |
+
"loss": 0.3449,
|
| 813 |
+
"rewards/accuracies": 0.875,
|
| 814 |
+
"rewards/chosen": 0.6405723094940186,
|
| 815 |
+
"rewards/margins": 1.1827360391616821,
|
| 816 |
+
"rewards/rejected": -0.5421636700630188,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.8827375173568726,
|
| 823 |
+
"logits/rejected": -1.89346444606781,
|
| 824 |
+
"logps/chosen": -32.36341857910156,
|
| 825 |
+
"logps/rejected": -31.319293975830078,
|
| 826 |
+
"loss": 0.3592,
|
| 827 |
+
"rewards/accuracies": 0.875,
|
| 828 |
+
"rewards/chosen": 0.7191308736801147,
|
| 829 |
+
"rewards/margins": 1.2532163858413696,
|
| 830 |
+
"rewards/rejected": -0.5340853929519653,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.829416036605835,
|
| 837 |
+
"logits/rejected": -1.8270241022109985,
|
| 838 |
+
"logps/chosen": -30.80521011352539,
|
| 839 |
+
"logps/rejected": -30.752178192138672,
|
| 840 |
+
"loss": 0.3535,
|
| 841 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 842 |
+
"rewards/chosen": 0.6894973516464233,
|
| 843 |
+
"rewards/margins": 1.1498301029205322,
|
| 844 |
+
"rewards/rejected": -0.46033263206481934,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.9103914499282837,
|
| 851 |
+
"logits/rejected": -1.9033727645874023,
|
| 852 |
+
"logps/chosen": -33.29057312011719,
|
| 853 |
+
"logps/rejected": -33.72592544555664,
|
| 854 |
+
"loss": 0.33,
|
| 855 |
+
"rewards/accuracies": 0.887499988079071,
|
| 856 |
+
"rewards/chosen": 0.7538042664527893,
|
| 857 |
+
"rewards/margins": 1.3205549716949463,
|
| 858 |
+
"rewards/rejected": -0.5667508244514465,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.8793199062347412,
|
| 865 |
+
"logits/rejected": -1.8857247829437256,
|
| 866 |
+
"logps/chosen": -30.955270767211914,
|
| 867 |
+
"logps/rejected": -33.297035217285156,
|
| 868 |
+
"loss": 0.3587,
|
| 869 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 870 |
+
"rewards/chosen": 0.6383341550827026,
|
| 871 |
+
"rewards/margins": 1.0689740180969238,
|
| 872 |
+
"rewards/rejected": -0.43063968420028687,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -1.949784278869629,
|
| 879 |
+
"logits/rejected": -1.9266754388809204,
|
| 880 |
+
"logps/chosen": -30.55877685546875,
|
| 881 |
+
"logps/rejected": -33.56010818481445,
|
| 882 |
+
"loss": 0.3523,
|
| 883 |
+
"rewards/accuracies": 0.925000011920929,
|
| 884 |
+
"rewards/chosen": 0.7030326128005981,
|
| 885 |
+
"rewards/margins": 1.1542710065841675,
|
| 886 |
+
"rewards/rejected": -0.45123839378356934,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.883561134338379,
|
| 893 |
+
"logits/rejected": -1.8786439895629883,
|
| 894 |
+
"logps/chosen": -31.494693756103516,
|
| 895 |
+
"logps/rejected": -32.63154220581055,
|
| 896 |
+
"loss": 0.3697,
|
| 897 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 898 |
+
"rewards/chosen": 0.7376305460929871,
|
| 899 |
+
"rewards/margins": 1.1461317539215088,
|
| 900 |
+
"rewards/rejected": -0.4085012972354889,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -1.9235492944717407,
|
| 907 |
+
"logits/rejected": -1.9254448413848877,
|
| 908 |
+
"logps/chosen": -33.56511688232422,
|
| 909 |
+
"logps/rejected": -35.665016174316406,
|
| 910 |
+
"loss": 0.3133,
|
| 911 |
+
"rewards/accuracies": 0.9375,
|
| 912 |
+
"rewards/chosen": 0.6485405564308167,
|
| 913 |
+
"rewards/margins": 1.339817762374878,
|
| 914 |
+
"rewards/rejected": -0.6912771463394165,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.9060910940170288,
|
| 921 |
+
"logits/rejected": -1.9054912328720093,
|
| 922 |
+
"logps/chosen": -29.636911392211914,
|
| 923 |
+
"logps/rejected": -32.14686965942383,
|
| 924 |
+
"loss": 0.332,
|
| 925 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 926 |
+
"rewards/chosen": 0.7641160488128662,
|
| 927 |
+
"rewards/margins": 1.312302827835083,
|
| 928 |
+
"rewards/rejected": -0.5481868386268616,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -1.919725775718689,
|
| 935 |
+
"logits/rejected": -1.9174749851226807,
|
| 936 |
+
"logps/chosen": -31.991186141967773,
|
| 937 |
+
"logps/rejected": -32.90464782714844,
|
| 938 |
+
"loss": 0.336,
|
| 939 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 940 |
+
"rewards/chosen": 0.8056806325912476,
|
| 941 |
+
"rewards/margins": 1.2894303798675537,
|
| 942 |
+
"rewards/rejected": -0.4837496280670166,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.160193920135498,
|
| 948 |
+
"eval_logits/rejected": -2.1554007530212402,
|
| 949 |
+
"eval_logps/chosen": -34.264404296875,
|
| 950 |
+
"eval_logps/rejected": -37.94013214111328,
|
| 951 |
+
"eval_loss": 0.6617584228515625,
|
| 952 |
+
"eval_rewards/accuracies": 0.6241694688796997,
|
| 953 |
+
"eval_rewards/chosen": -0.18387840688228607,
|
| 954 |
+
"eval_rewards/margins": 0.15492965281009674,
|
| 955 |
+
"eval_rewards/rejected": -0.3388080298900604,
|
| 956 |
+
"eval_runtime": 145.9206,
|
| 957 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.9637705087661743,
|
| 965 |
+
"logits/rejected": -1.9703562259674072,
|
| 966 |
+
"logps/chosen": -31.264089584350586,
|
| 967 |
+
"logps/rejected": -32.738319396972656,
|
| 968 |
+
"loss": 0.3311,
|
| 969 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 970 |
+
"rewards/chosen": 0.7786344289779663,
|
| 971 |
+
"rewards/margins": 1.3187401294708252,
|
| 972 |
+
"rewards/rejected": -0.5401058197021484,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -1.9196746349334717,
|
| 979 |
+
"logits/rejected": -1.9172741174697876,
|
| 980 |
+
"logps/chosen": -33.669281005859375,
|
| 981 |
+
"logps/rejected": -31.596511840820312,
|
| 982 |
+
"loss": 0.3157,
|
| 983 |
+
"rewards/accuracies": 0.887499988079071,
|
| 984 |
+
"rewards/chosen": 0.869280993938446,
|
| 985 |
+
"rewards/margins": 1.3580834865570068,
|
| 986 |
+
"rewards/rejected": -0.48880258202552795,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.8648065328598022,
|
| 993 |
+
"logits/rejected": -1.8580557107925415,
|
| 994 |
+
"logps/chosen": -34.281700134277344,
|
| 995 |
+
"logps/rejected": -33.09290313720703,
|
| 996 |
+
"loss": 0.3058,
|
| 997 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 998 |
+
"rewards/chosen": 0.9460186958312988,
|
| 999 |
+
"rewards/margins": 1.5345759391784668,
|
| 1000 |
+
"rewards/rejected": -0.5885573029518127,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.8759320974349976,
|
| 1007 |
+
"logits/rejected": -1.8822145462036133,
|
| 1008 |
+
"logps/chosen": -32.987449645996094,
|
| 1009 |
+
"logps/rejected": -33.8790283203125,
|
| 1010 |
+
"loss": 0.3408,
|
| 1011 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1012 |
+
"rewards/chosen": 0.7934818863868713,
|
| 1013 |
+
"rewards/margins": 1.3044978380203247,
|
| 1014 |
+
"rewards/rejected": -0.5110160708427429,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -1.9594967365264893,
|
| 1021 |
+
"logits/rejected": -1.973659873008728,
|
| 1022 |
+
"logps/chosen": -31.017492294311523,
|
| 1023 |
+
"logps/rejected": -32.631839752197266,
|
| 1024 |
+
"loss": 0.3573,
|
| 1025 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1026 |
+
"rewards/chosen": 0.6679788827896118,
|
| 1027 |
+
"rewards/margins": 1.1753003597259521,
|
| 1028 |
+
"rewards/rejected": -0.5073216557502747,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.8336044549942017,
|
| 1035 |
+
"logits/rejected": -1.8305423259735107,
|
| 1036 |
+
"logps/chosen": -32.8193473815918,
|
| 1037 |
+
"logps/rejected": -35.608089447021484,
|
| 1038 |
+
"loss": 0.2779,
|
| 1039 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1040 |
+
"rewards/chosen": 0.8828850984573364,
|
| 1041 |
+
"rewards/margins": 1.6091547012329102,
|
| 1042 |
+
"rewards/rejected": -0.726269543170929,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -1.9243332147598267,
|
| 1049 |
+
"logits/rejected": -1.9244455099105835,
|
| 1050 |
+
"logps/chosen": -30.38960838317871,
|
| 1051 |
+
"logps/rejected": -34.67184066772461,
|
| 1052 |
+
"loss": 0.3114,
|
| 1053 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1054 |
+
"rewards/chosen": 0.7167855501174927,
|
| 1055 |
+
"rewards/margins": 1.3931505680084229,
|
| 1056 |
+
"rewards/rejected": -0.676365077495575,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.8737255334854126,
|
| 1063 |
+
"logits/rejected": -1.8705682754516602,
|
| 1064 |
+
"logps/chosen": -29.965240478515625,
|
| 1065 |
+
"logps/rejected": -33.58513259887695,
|
| 1066 |
+
"loss": 0.3412,
|
| 1067 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1068 |
+
"rewards/chosen": 0.7256845235824585,
|
| 1069 |
+
"rewards/margins": 1.339548945426941,
|
| 1070 |
+
"rewards/rejected": -0.613864541053772,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -1.908151626586914,
|
| 1077 |
+
"logits/rejected": -1.908117651939392,
|
| 1078 |
+
"logps/chosen": -28.949966430664062,
|
| 1079 |
+
"logps/rejected": -31.339040756225586,
|
| 1080 |
+
"loss": 0.3023,
|
| 1081 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1082 |
+
"rewards/chosen": 0.9705016016960144,
|
| 1083 |
+
"rewards/margins": 1.5053366422653198,
|
| 1084 |
+
"rewards/rejected": -0.5348349809646606,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.8213777542114258,
|
| 1091 |
+
"logits/rejected": -1.824627161026001,
|
| 1092 |
+
"logps/chosen": -33.35689163208008,
|
| 1093 |
+
"logps/rejected": -32.15576934814453,
|
| 1094 |
+
"loss": 0.3559,
|
| 1095 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1096 |
+
"rewards/chosen": 0.8972657322883606,
|
| 1097 |
+
"rewards/margins": 1.4501286745071411,
|
| 1098 |
+
"rewards/rejected": -0.5528628826141357,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.158259153366089,
|
| 1104 |
+
"eval_logits/rejected": -2.1534817218780518,
|
| 1105 |
+
"eval_logps/chosen": -34.35598373413086,
|
| 1106 |
+
"eval_logps/rejected": -37.98072814941406,
|
| 1107 |
+
"eval_loss": 0.6946965456008911,
|
| 1108 |
+
"eval_rewards/accuracies": 0.5340532064437866,
|
| 1109 |
+
"eval_rewards/chosen": -0.2571484446525574,
|
| 1110 |
+
"eval_rewards/margins": 0.1141389012336731,
|
| 1111 |
+
"eval_rewards/rejected": -0.37128737568855286,
|
| 1112 |
+
"eval_runtime": 145.9197,
|
| 1113 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.9044933319091797,
|
| 1121 |
+
"logits/rejected": -1.8955074548721313,
|
| 1122 |
+
"logps/chosen": -35.51199722290039,
|
| 1123 |
+
"logps/rejected": -32.60811233520508,
|
| 1124 |
+
"loss": 0.3093,
|
| 1125 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1126 |
+
"rewards/chosen": 0.8000162839889526,
|
| 1127 |
+
"rewards/margins": 1.3618541955947876,
|
| 1128 |
+
"rewards/rejected": -0.5618377923965454,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -1.9182393550872803,
|
| 1135 |
+
"logits/rejected": -1.9193944931030273,
|
| 1136 |
+
"logps/chosen": -35.33127212524414,
|
| 1137 |
+
"logps/rejected": -33.76591110229492,
|
| 1138 |
+
"loss": 0.3149,
|
| 1139 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1140 |
+
"rewards/chosen": 0.8832913637161255,
|
| 1141 |
+
"rewards/margins": 1.435203194618225,
|
| 1142 |
+
"rewards/rejected": -0.5519118309020996,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -1.9362226724624634,
|
| 1149 |
+
"logits/rejected": -1.9337679147720337,
|
| 1150 |
+
"logps/chosen": -31.4454345703125,
|
| 1151 |
+
"logps/rejected": -33.67931365966797,
|
| 1152 |
+
"loss": 0.3081,
|
| 1153 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1154 |
+
"rewards/chosen": 0.8943157196044922,
|
| 1155 |
+
"rewards/margins": 1.4402475357055664,
|
| 1156 |
+
"rewards/rejected": -0.5459317564964294,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -1.9176380634307861,
|
| 1163 |
+
"logits/rejected": -1.922844648361206,
|
| 1164 |
+
"logps/chosen": -30.585041046142578,
|
| 1165 |
+
"logps/rejected": -33.932350158691406,
|
| 1166 |
+
"loss": 0.2741,
|
| 1167 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1168 |
+
"rewards/chosen": 0.9605496525764465,
|
| 1169 |
+
"rewards/margins": 1.5825778245925903,
|
| 1170 |
+
"rewards/rejected": -0.6220281720161438,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.864728569984436,
|
| 1177 |
+
"logits/rejected": -1.8635470867156982,
|
| 1178 |
+
"logps/chosen": -32.103065490722656,
|
| 1179 |
+
"logps/rejected": -35.512298583984375,
|
| 1180 |
+
"loss": 0.3075,
|
| 1181 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1182 |
+
"rewards/chosen": 0.8175305128097534,
|
| 1183 |
+
"rewards/margins": 1.478601098060608,
|
| 1184 |
+
"rewards/rejected": -0.6610704660415649,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.7420361042022705,
|
| 1191 |
+
"logits/rejected": -1.7370550632476807,
|
| 1192 |
+
"logps/chosen": -34.88098907470703,
|
| 1193 |
+
"logps/rejected": -35.97365951538086,
|
| 1194 |
+
"loss": 0.3032,
|
| 1195 |
+
"rewards/accuracies": 0.875,
|
| 1196 |
+
"rewards/chosen": 0.8939619064331055,
|
| 1197 |
+
"rewards/margins": 1.506162166595459,
|
| 1198 |
+
"rewards/rejected": -0.6122000813484192,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.8814598321914673,
|
| 1205 |
+
"logits/rejected": -1.8852901458740234,
|
| 1206 |
+
"logps/chosen": -34.113304138183594,
|
| 1207 |
+
"logps/rejected": -34.245601654052734,
|
| 1208 |
+
"loss": 0.3271,
|
| 1209 |
+
"rewards/accuracies": 0.85833340883255,
|
| 1210 |
+
"rewards/chosen": 0.8802832365036011,
|
| 1211 |
+
"rewards/margins": 1.4008936882019043,
|
| 1212 |
+
"rewards/rejected": -0.5206103324890137,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.8257869482040405,
|
| 1219 |
+
"logits/rejected": -1.8239891529083252,
|
| 1220 |
+
"logps/chosen": -32.38594436645508,
|
| 1221 |
+
"logps/rejected": -35.74359893798828,
|
| 1222 |
+
"loss": 0.154,
|
| 1223 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1224 |
+
"rewards/chosen": 1.2125210762023926,
|
| 1225 |
+
"rewards/margins": 2.2836754322052,
|
| 1226 |
+
"rewards/rejected": -1.071154236793518,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.8724597692489624,
|
| 1233 |
+
"logits/rejected": -1.8708446025848389,
|
| 1234 |
+
"logps/chosen": -31.859329223632812,
|
| 1235 |
+
"logps/rejected": -34.107154846191406,
|
| 1236 |
+
"loss": 0.1817,
|
| 1237 |
+
"rewards/accuracies": 0.9375,
|
| 1238 |
+
"rewards/chosen": 1.2131247520446777,
|
| 1239 |
+
"rewards/margins": 2.029592752456665,
|
| 1240 |
+
"rewards/rejected": -0.8164681196212769,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.8005281686782837,
|
| 1247 |
+
"logits/rejected": -1.7946115732192993,
|
| 1248 |
+
"logps/chosen": -32.934303283691406,
|
| 1249 |
+
"logps/rejected": -32.45271301269531,
|
| 1250 |
+
"loss": 0.1978,
|
| 1251 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1252 |
+
"rewards/chosen": 1.1590179204940796,
|
| 1253 |
+
"rewards/margins": 1.9497814178466797,
|
| 1254 |
+
"rewards/rejected": -0.7907635569572449,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.154881000518799,
|
| 1260 |
+
"eval_logits/rejected": -2.1501035690307617,
|
| 1261 |
+
"eval_logps/chosen": -34.32504653930664,
|
| 1262 |
+
"eval_logps/rejected": -37.97528076171875,
|
| 1263 |
+
"eval_loss": 0.6837867498397827,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5834717750549316,
|
| 1265 |
+
"eval_rewards/chosen": -0.2323925942182541,
|
| 1266 |
+
"eval_rewards/margins": 0.1345335692167282,
|
| 1267 |
+
"eval_rewards/rejected": -0.3669261634349823,
|
| 1268 |
+
"eval_runtime": 145.9433,
|
| 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": -1.898033857345581,
|
| 1277 |
+
"logits/rejected": -1.905503511428833,
|
| 1278 |
+
"logps/chosen": -30.634164810180664,
|
| 1279 |
+
"logps/rejected": -35.38495635986328,
|
| 1280 |
+
"loss": 0.189,
|
| 1281 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1282 |
+
"rewards/chosen": 1.1952745914459229,
|
| 1283 |
+
"rewards/margins": 2.0985639095306396,
|
| 1284 |
+
"rewards/rejected": -0.9032894372940063,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.872579574584961,
|
| 1291 |
+
"logits/rejected": -1.875286340713501,
|
| 1292 |
+
"logps/chosen": -31.190942764282227,
|
| 1293 |
+
"logps/rejected": -34.01335906982422,
|
| 1294 |
+
"loss": 0.2433,
|
| 1295 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1296 |
+
"rewards/chosen": 1.104109525680542,
|
| 1297 |
+
"rewards/margins": 1.7309566736221313,
|
| 1298 |
+
"rewards/rejected": -0.6268469095230103,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.7723095417022705,
|
| 1305 |
+
"logits/rejected": -1.7764465808868408,
|
| 1306 |
+
"logps/chosen": -31.214824676513672,
|
| 1307 |
+
"logps/rejected": -36.142818450927734,
|
| 1308 |
+
"loss": 0.1808,
|
| 1309 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1310 |
+
"rewards/chosen": 1.1013058423995972,
|
| 1311 |
+
"rewards/margins": 2.0874154567718506,
|
| 1312 |
+
"rewards/rejected": -0.9861096143722534,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.8381569385528564,
|
| 1319 |
+
"logits/rejected": -1.8343050479888916,
|
| 1320 |
+
"logps/chosen": -31.591388702392578,
|
| 1321 |
+
"logps/rejected": -33.85132598876953,
|
| 1322 |
+
"loss": 0.2508,
|
| 1323 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1324 |
+
"rewards/chosen": 1.1018110513687134,
|
| 1325 |
+
"rewards/margins": 1.7871520519256592,
|
| 1326 |
+
"rewards/rejected": -0.6853411793708801,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -1.9189765453338623,
|
| 1333 |
+
"logits/rejected": -1.9210529327392578,
|
| 1334 |
+
"logps/chosen": -31.255529403686523,
|
| 1335 |
+
"logps/rejected": -36.854835510253906,
|
| 1336 |
+
"loss": 0.1574,
|
| 1337 |
+
"rewards/accuracies": 1.0,
|
| 1338 |
+
"rewards/chosen": 1.2158453464508057,
|
| 1339 |
+
"rewards/margins": 2.1701653003692627,
|
| 1340 |
+
"rewards/rejected": -0.9543198347091675,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.7254867553710938,
|
| 1347 |
+
"logits/rejected": -1.7197097539901733,
|
| 1348 |
+
"logps/chosen": -34.86113357543945,
|
| 1349 |
+
"logps/rejected": -33.97612762451172,
|
| 1350 |
+
"loss": 0.2044,
|
| 1351 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1352 |
+
"rewards/chosen": 1.1364357471466064,
|
| 1353 |
+
"rewards/margins": 2.003814697265625,
|
| 1354 |
+
"rewards/rejected": -0.8673788905143738,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.7829675674438477,
|
| 1361 |
+
"logits/rejected": -1.7830712795257568,
|
| 1362 |
+
"logps/chosen": -34.304718017578125,
|
| 1363 |
+
"logps/rejected": -35.99662780761719,
|
| 1364 |
+
"loss": 0.181,
|
| 1365 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1366 |
+
"rewards/chosen": 1.2304695844650269,
|
| 1367 |
+
"rewards/margins": 2.150757312774658,
|
| 1368 |
+
"rewards/rejected": -0.9202877283096313,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.8317804336547852,
|
| 1375 |
+
"logits/rejected": -1.8203016519546509,
|
| 1376 |
+
"logps/chosen": -32.700950622558594,
|
| 1377 |
+
"logps/rejected": -34.08301544189453,
|
| 1378 |
+
"loss": 0.1692,
|
| 1379 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1380 |
+
"rewards/chosen": 1.1245036125183105,
|
| 1381 |
+
"rewards/margins": 2.0532915592193604,
|
| 1382 |
+
"rewards/rejected": -0.9287878274917603,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -1.9331659078598022,
|
| 1389 |
+
"logits/rejected": -1.9377342462539673,
|
| 1390 |
+
"logps/chosen": -31.932796478271484,
|
| 1391 |
+
"logps/rejected": -33.6646728515625,
|
| 1392 |
+
"loss": 0.1905,
|
| 1393 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1394 |
+
"rewards/chosen": 1.1177093982696533,
|
| 1395 |
+
"rewards/margins": 2.100264310836792,
|
| 1396 |
+
"rewards/rejected": -0.9825549125671387,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.8520333766937256,
|
| 1403 |
+
"logits/rejected": -1.844834327697754,
|
| 1404 |
+
"logps/chosen": -33.29985427856445,
|
| 1405 |
+
"logps/rejected": -33.635623931884766,
|
| 1406 |
+
"loss": 0.1619,
|
| 1407 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1408 |
+
"rewards/chosen": 1.243377923965454,
|
| 1409 |
+
"rewards/margins": 2.2301688194274902,
|
| 1410 |
+
"rewards/rejected": -0.9867912530899048,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.1431386470794678,
|
| 1416 |
+
"eval_logits/rejected": -2.138375997543335,
|
| 1417 |
+
"eval_logps/chosen": -34.34238052368164,
|
| 1418 |
+
"eval_logps/rejected": -38.03608703613281,
|
| 1419 |
+
"eval_loss": 0.6788281202316284,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1421 |
+
"eval_rewards/chosen": -0.24626301229000092,
|
| 1422 |
+
"eval_rewards/margins": 0.16930998861789703,
|
| 1423 |
+
"eval_rewards/rejected": -0.41557300090789795,
|
| 1424 |
+
"eval_runtime": 145.9104,
|
| 1425 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.8457978963851929,
|
| 1433 |
+
"logits/rejected": -1.8462024927139282,
|
| 1434 |
+
"logps/chosen": -29.82485008239746,
|
| 1435 |
+
"logps/rejected": -36.06700897216797,
|
| 1436 |
+
"loss": 0.1786,
|
| 1437 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1438 |
+
"rewards/chosen": 1.0738495588302612,
|
| 1439 |
+
"rewards/margins": 2.1757030487060547,
|
| 1440 |
+
"rewards/rejected": -1.1018531322479248,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.8021042346954346,
|
| 1447 |
+
"logits/rejected": -1.7940893173217773,
|
| 1448 |
+
"logps/chosen": -30.521142959594727,
|
| 1449 |
+
"logps/rejected": -34.798255920410156,
|
| 1450 |
+
"loss": 0.1883,
|
| 1451 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1452 |
+
"rewards/chosen": 1.0542433261871338,
|
| 1453 |
+
"rewards/margins": 2.0357508659362793,
|
| 1454 |
+
"rewards/rejected": -0.9815075993537903,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -1.9680095911026,
|
| 1461 |
+
"logits/rejected": -1.968182921409607,
|
| 1462 |
+
"logps/chosen": -28.589441299438477,
|
| 1463 |
+
"logps/rejected": -33.940185546875,
|
| 1464 |
+
"loss": 0.2295,
|
| 1465 |
+
"rewards/accuracies": 0.9375,
|
| 1466 |
+
"rewards/chosen": 1.1260759830474854,
|
| 1467 |
+
"rewards/margins": 1.9863367080688477,
|
| 1468 |
+
"rewards/rejected": -0.8602607846260071,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.847324013710022,
|
| 1475 |
+
"logits/rejected": -1.857102394104004,
|
| 1476 |
+
"logps/chosen": -32.901668548583984,
|
| 1477 |
+
"logps/rejected": -34.3242073059082,
|
| 1478 |
+
"loss": 0.1675,
|
| 1479 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1480 |
+
"rewards/chosen": 1.2187172174453735,
|
| 1481 |
+
"rewards/margins": 2.2863640785217285,
|
| 1482 |
+
"rewards/rejected": -1.0676472187042236,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -1.8901779651641846,
|
| 1489 |
+
"logits/rejected": -1.8939307928085327,
|
| 1490 |
+
"logps/chosen": -29.412405014038086,
|
| 1491 |
+
"logps/rejected": -32.94640350341797,
|
| 1492 |
+
"loss": 0.2048,
|
| 1493 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1494 |
+
"rewards/chosen": 0.9882534146308899,
|
| 1495 |
+
"rewards/margins": 1.9565131664276123,
|
| 1496 |
+
"rewards/rejected": -0.9682596921920776,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -1.8851401805877686,
|
| 1503 |
+
"logits/rejected": -1.8811264038085938,
|
| 1504 |
+
"logps/chosen": -32.44501495361328,
|
| 1505 |
+
"logps/rejected": -34.933692932128906,
|
| 1506 |
+
"loss": 0.2061,
|
| 1507 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1508 |
+
"rewards/chosen": 1.3041638135910034,
|
| 1509 |
+
"rewards/margins": 2.126009225845337,
|
| 1510 |
+
"rewards/rejected": -0.8218452334403992,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.8027584552764893,
|
| 1517 |
+
"logits/rejected": -1.805301308631897,
|
| 1518 |
+
"logps/chosen": -31.277942657470703,
|
| 1519 |
+
"logps/rejected": -36.393043518066406,
|
| 1520 |
+
"loss": 0.1663,
|
| 1521 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1522 |
+
"rewards/chosen": 1.2278459072113037,
|
| 1523 |
+
"rewards/margins": 2.197005033493042,
|
| 1524 |
+
"rewards/rejected": -0.9691593050956726,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.835371971130371,
|
| 1531 |
+
"logits/rejected": -1.8288919925689697,
|
| 1532 |
+
"logps/chosen": -34.11837387084961,
|
| 1533 |
+
"logps/rejected": -37.60203552246094,
|
| 1534 |
+
"loss": 0.2066,
|
| 1535 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1536 |
+
"rewards/chosen": 1.0293549299240112,
|
| 1537 |
+
"rewards/margins": 2.101419687271118,
|
| 1538 |
+
"rewards/rejected": -1.0720646381378174,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.813190221786499,
|
| 1545 |
+
"logits/rejected": -1.807956337928772,
|
| 1546 |
+
"logps/chosen": -29.670574188232422,
|
| 1547 |
+
"logps/rejected": -37.203330993652344,
|
| 1548 |
+
"loss": 0.2099,
|
| 1549 |
+
"rewards/accuracies": 0.9375,
|
| 1550 |
+
"rewards/chosen": 1.06589937210083,
|
| 1551 |
+
"rewards/margins": 2.0366604328155518,
|
| 1552 |
+
"rewards/rejected": -0.970760703086853,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.7904030084609985,
|
| 1559 |
+
"logits/rejected": -1.779764175415039,
|
| 1560 |
+
"logps/chosen": -30.418493270874023,
|
| 1561 |
+
"logps/rejected": -30.5950984954834,
|
| 1562 |
+
"loss": 0.209,
|
| 1563 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1564 |
+
"rewards/chosen": 1.0641578435897827,
|
| 1565 |
+
"rewards/margins": 1.8604481220245361,
|
| 1566 |
+
"rewards/rejected": -0.796290397644043,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.1356685161590576,
|
| 1572 |
+
"eval_logits/rejected": -2.130932331085205,
|
| 1573 |
+
"eval_logps/chosen": -34.380455017089844,
|
| 1574 |
+
"eval_logps/rejected": -38.083457946777344,
|
| 1575 |
+
"eval_loss": 0.6776629090309143,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5917773842811584,
|
| 1577 |
+
"eval_rewards/chosen": -0.2767248749732971,
|
| 1578 |
+
"eval_rewards/margins": 0.17674396932125092,
|
| 1579 |
+
"eval_rewards/rejected": -0.45346885919570923,
|
| 1580 |
+
"eval_runtime": 145.9185,
|
| 1581 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.896877646446228,
|
| 1589 |
+
"logits/rejected": -1.896228551864624,
|
| 1590 |
+
"logps/chosen": -30.421009063720703,
|
| 1591 |
+
"logps/rejected": -31.305410385131836,
|
| 1592 |
+
"loss": 0.22,
|
| 1593 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1594 |
+
"rewards/chosen": 1.1225507259368896,
|
| 1595 |
+
"rewards/margins": 1.9600980281829834,
|
| 1596 |
+
"rewards/rejected": -0.8375473022460938,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -1.9368255138397217,
|
| 1603 |
+
"logits/rejected": -1.9375327825546265,
|
| 1604 |
+
"logps/chosen": -30.484455108642578,
|
| 1605 |
+
"logps/rejected": -32.34059524536133,
|
| 1606 |
+
"loss": 0.177,
|
| 1607 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1608 |
+
"rewards/chosen": 1.167297601699829,
|
| 1609 |
+
"rewards/margins": 2.1071255207061768,
|
| 1610 |
+
"rewards/rejected": -0.9398279190063477,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -1.9036754369735718,
|
| 1617 |
+
"logits/rejected": -1.9070818424224854,
|
| 1618 |
+
"logps/chosen": -30.951061248779297,
|
| 1619 |
+
"logps/rejected": -28.893712997436523,
|
| 1620 |
+
"loss": 0.227,
|
| 1621 |
+
"rewards/accuracies": 0.9375,
|
| 1622 |
+
"rewards/chosen": 1.0632069110870361,
|
| 1623 |
+
"rewards/margins": 1.7912060022354126,
|
| 1624 |
+
"rewards/rejected": -0.7279989719390869,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -1.91388738155365,
|
| 1631 |
+
"logits/rejected": -1.906005859375,
|
| 1632 |
+
"logps/chosen": -29.750720977783203,
|
| 1633 |
+
"logps/rejected": -34.767372131347656,
|
| 1634 |
+
"loss": 0.1882,
|
| 1635 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1636 |
+
"rewards/chosen": 1.1417419910430908,
|
| 1637 |
+
"rewards/margins": 2.1554553508758545,
|
| 1638 |
+
"rewards/rejected": -1.0137132406234741,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.8500654697418213,
|
| 1645 |
+
"logits/rejected": -1.8609867095947266,
|
| 1646 |
+
"logps/chosen": -32.1203727722168,
|
| 1647 |
+
"logps/rejected": -33.35521697998047,
|
| 1648 |
+
"loss": 0.1865,
|
| 1649 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1650 |
+
"rewards/chosen": 1.1461875438690186,
|
| 1651 |
+
"rewards/margins": 2.034356117248535,
|
| 1652 |
+
"rewards/rejected": -0.888168454170227,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -1.9280004501342773,
|
| 1659 |
+
"logits/rejected": -1.9186464548110962,
|
| 1660 |
+
"logps/chosen": -32.13129425048828,
|
| 1661 |
+
"logps/rejected": -37.68938446044922,
|
| 1662 |
+
"loss": 0.2127,
|
| 1663 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1664 |
+
"rewards/chosen": 1.0955873727798462,
|
| 1665 |
+
"rewards/margins": 2.181983470916748,
|
| 1666 |
+
"rewards/rejected": -1.0863959789276123,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.835701584815979,
|
| 1673 |
+
"logits/rejected": -1.837551474571228,
|
| 1674 |
+
"logps/chosen": -34.348480224609375,
|
| 1675 |
+
"logps/rejected": -33.834922790527344,
|
| 1676 |
+
"loss": 0.1815,
|
| 1677 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1678 |
+
"rewards/chosen": 1.3126420974731445,
|
| 1679 |
+
"rewards/margins": 2.2222840785980225,
|
| 1680 |
+
"rewards/rejected": -0.9096420407295227,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -1.9658441543579102,
|
| 1687 |
+
"logits/rejected": -1.9667274951934814,
|
| 1688 |
+
"logps/chosen": -31.538421630859375,
|
| 1689 |
+
"logps/rejected": -35.54855728149414,
|
| 1690 |
+
"loss": 0.1617,
|
| 1691 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1692 |
+
"rewards/chosen": 1.2153749465942383,
|
| 1693 |
+
"rewards/margins": 2.3616721630096436,
|
| 1694 |
+
"rewards/rejected": -1.1462972164154053,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -1.9372714757919312,
|
| 1701 |
+
"logits/rejected": -1.9413518905639648,
|
| 1702 |
+
"logps/chosen": -29.44711685180664,
|
| 1703 |
+
"logps/rejected": -34.69606399536133,
|
| 1704 |
+
"loss": 0.2085,
|
| 1705 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1706 |
+
"rewards/chosen": 1.156960129737854,
|
| 1707 |
+
"rewards/margins": 2.0755460262298584,
|
| 1708 |
+
"rewards/rejected": -0.9185860753059387,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -1.9265114068984985,
|
| 1715 |
+
"logits/rejected": -1.9317893981933594,
|
| 1716 |
+
"logps/chosen": -32.267982482910156,
|
| 1717 |
+
"logps/rejected": -31.293163299560547,
|
| 1718 |
+
"loss": 0.2513,
|
| 1719 |
+
"rewards/accuracies": 0.9375,
|
| 1720 |
+
"rewards/chosen": 1.0408172607421875,
|
| 1721 |
+
"rewards/margins": 1.7540076971054077,
|
| 1722 |
+
"rewards/rejected": -0.7131900191307068,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.1317365169525146,
|
| 1728 |
+
"eval_logits/rejected": -2.126985549926758,
|
| 1729 |
+
"eval_logps/chosen": -34.40774917602539,
|
| 1730 |
+
"eval_logps/rejected": -38.09048080444336,
|
| 1731 |
+
"eval_loss": 0.6897340416908264,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5830564498901367,
|
| 1733 |
+
"eval_rewards/chosen": -0.2985561192035675,
|
| 1734 |
+
"eval_rewards/margins": 0.1605302393436432,
|
| 1735 |
+
"eval_rewards/rejected": -0.4590863585472107,
|
| 1736 |
+
"eval_runtime": 145.9092,
|
| 1737 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.918264389038086,
|
| 1745 |
+
"logits/rejected": -1.9183082580566406,
|
| 1746 |
+
"logps/chosen": -31.37972068786621,
|
| 1747 |
+
"logps/rejected": -31.99541664123535,
|
| 1748 |
+
"loss": 0.1933,
|
| 1749 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1750 |
+
"rewards/chosen": 1.248342514038086,
|
| 1751 |
+
"rewards/margins": 2.031942844390869,
|
| 1752 |
+
"rewards/rejected": -0.783600389957428,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.7625865936279297,
|
| 1759 |
+
"logits/rejected": -1.7643556594848633,
|
| 1760 |
+
"logps/chosen": -33.15352249145508,
|
| 1761 |
+
"logps/rejected": -34.140689849853516,
|
| 1762 |
+
"loss": 0.2375,
|
| 1763 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1764 |
+
"rewards/chosen": 1.1291639804840088,
|
| 1765 |
+
"rewards/margins": 2.008704900741577,
|
| 1766 |
+
"rewards/rejected": -0.8795410394668579,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.7994060516357422,
|
| 1773 |
+
"logits/rejected": -1.7968616485595703,
|
| 1774 |
+
"logps/chosen": -29.835479736328125,
|
| 1775 |
+
"logps/rejected": -32.159576416015625,
|
| 1776 |
+
"loss": 0.2885,
|
| 1777 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1778 |
+
"rewards/chosen": 0.940991222858429,
|
| 1779 |
+
"rewards/margins": 1.736061692237854,
|
| 1780 |
+
"rewards/rejected": -0.7950705289840698,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -1.9358510971069336,
|
| 1787 |
+
"logits/rejected": -1.9303522109985352,
|
| 1788 |
+
"logps/chosen": -31.38555335998535,
|
| 1789 |
+
"logps/rejected": -34.0118408203125,
|
| 1790 |
+
"loss": 0.1779,
|
| 1791 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1792 |
+
"rewards/chosen": 1.3233940601348877,
|
| 1793 |
+
"rewards/margins": 2.1696605682373047,
|
| 1794 |
+
"rewards/rejected": -0.8462664484977722,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.7935917377471924,
|
| 1801 |
+
"logits/rejected": -1.783705472946167,
|
| 1802 |
+
"logps/chosen": -32.11836624145508,
|
| 1803 |
+
"logps/rejected": -32.378910064697266,
|
| 1804 |
+
"loss": 0.2083,
|
| 1805 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1806 |
+
"rewards/chosen": 1.066697120666504,
|
| 1807 |
+
"rewards/margins": 1.9359056949615479,
|
| 1808 |
+
"rewards/rejected": -0.8692083358764648,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.8640037775039673,
|
| 1815 |
+
"logits/rejected": -1.8639949560165405,
|
| 1816 |
+
"logps/chosen": -32.08525848388672,
|
| 1817 |
+
"logps/rejected": -32.31430435180664,
|
| 1818 |
+
"loss": 0.1695,
|
| 1819 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1820 |
+
"rewards/chosen": 1.2771949768066406,
|
| 1821 |
+
"rewards/margins": 2.167479991912842,
|
| 1822 |
+
"rewards/rejected": -0.8902848958969116,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -1.9403139352798462,
|
| 1829 |
+
"logits/rejected": -1.9347970485687256,
|
| 1830 |
+
"logps/chosen": -32.035804748535156,
|
| 1831 |
+
"logps/rejected": -34.015647888183594,
|
| 1832 |
+
"loss": 0.186,
|
| 1833 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1834 |
+
"rewards/chosen": 1.1921296119689941,
|
| 1835 |
+
"rewards/margins": 2.0516738891601562,
|
| 1836 |
+
"rewards/rejected": -0.8595443964004517,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.7985414266586304,
|
| 1843 |
+
"logits/rejected": -1.806396484375,
|
| 1844 |
+
"logps/chosen": -31.918773651123047,
|
| 1845 |
+
"logps/rejected": -33.53784942626953,
|
| 1846 |
+
"loss": 0.175,
|
| 1847 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1848 |
+
"rewards/chosen": 1.1313796043395996,
|
| 1849 |
+
"rewards/margins": 2.0725083351135254,
|
| 1850 |
+
"rewards/rejected": -0.9411287307739258,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -1.86725652217865,
|
| 1857 |
+
"logits/rejected": -1.8614327907562256,
|
| 1858 |
+
"logps/chosen": -30.524410247802734,
|
| 1859 |
+
"logps/rejected": -34.46692657470703,
|
| 1860 |
+
"loss": 0.1668,
|
| 1861 |
+
"rewards/accuracies": 0.9375,
|
| 1862 |
+
"rewards/chosen": 1.319364309310913,
|
| 1863 |
+
"rewards/margins": 2.266146659851074,
|
| 1864 |
+
"rewards/rejected": -0.9467819929122925,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.8097419738769531,
|
| 1871 |
+
"logits/rejected": -1.8118005990982056,
|
| 1872 |
+
"logps/chosen": -32.011409759521484,
|
| 1873 |
+
"logps/rejected": -34.247711181640625,
|
| 1874 |
+
"loss": 0.1713,
|
| 1875 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1876 |
+
"rewards/chosen": 1.1832258701324463,
|
| 1877 |
+
"rewards/margins": 2.373357057571411,
|
| 1878 |
+
"rewards/rejected": -1.1901309490203857,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.131728410720825,
|
| 1884 |
+
"eval_logits/rejected": -2.1269805431365967,
|
| 1885 |
+
"eval_logps/chosen": -34.38140869140625,
|
| 1886 |
+
"eval_logps/rejected": -38.09336471557617,
|
| 1887 |
+
"eval_loss": 0.6780257821083069,
|
| 1888 |
+
"eval_rewards/accuracies": 0.594684362411499,
|
| 1889 |
+
"eval_rewards/chosen": -0.27748534083366394,
|
| 1890 |
+
"eval_rewards/margins": 0.18390974402427673,
|
| 1891 |
+
"eval_rewards/rejected": -0.4613950550556183,
|
| 1892 |
+
"eval_runtime": 145.9474,
|
| 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": -1.7440783977508545,
|
| 1901 |
+
"logits/rejected": -1.7430063486099243,
|
| 1902 |
+
"logps/chosen": -34.408180236816406,
|
| 1903 |
+
"logps/rejected": -34.03881072998047,
|
| 1904 |
+
"loss": 0.1517,
|
| 1905 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1906 |
+
"rewards/chosen": 1.3154469728469849,
|
| 1907 |
+
"rewards/margins": 2.3790054321289062,
|
| 1908 |
+
"rewards/rejected": -1.0635584592819214,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -1.9152580499649048,
|
| 1915 |
+
"logits/rejected": -1.9091880321502686,
|
| 1916 |
+
"logps/chosen": -30.193584442138672,
|
| 1917 |
+
"logps/rejected": -34.31517791748047,
|
| 1918 |
+
"loss": 0.1516,
|
| 1919 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1920 |
+
"rewards/chosen": 1.2080823183059692,
|
| 1921 |
+
"rewards/margins": 2.357390880584717,
|
| 1922 |
+
"rewards/rejected": -1.149308443069458,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.7672933340072632,
|
| 1929 |
+
"logits/rejected": -1.7558486461639404,
|
| 1930 |
+
"logps/chosen": -32.647369384765625,
|
| 1931 |
+
"logps/rejected": -34.192893981933594,
|
| 1932 |
+
"loss": 0.1462,
|
| 1933 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1934 |
+
"rewards/chosen": 1.3906490802764893,
|
| 1935 |
+
"rewards/margins": 2.429901599884033,
|
| 1936 |
+
"rewards/rejected": -1.0392528772354126,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.856805443763733,
|
| 1943 |
+
"logits/rejected": -1.850097894668579,
|
| 1944 |
+
"logps/chosen": -29.289770126342773,
|
| 1945 |
+
"logps/rejected": -30.042938232421875,
|
| 1946 |
+
"loss": 0.1928,
|
| 1947 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1948 |
+
"rewards/chosen": 1.1283700466156006,
|
| 1949 |
+
"rewards/margins": 2.0686745643615723,
|
| 1950 |
+
"rewards/rejected": -0.9403042793273926,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -1.8852243423461914,
|
| 1957 |
+
"logits/rejected": -1.8920814990997314,
|
| 1958 |
+
"logps/chosen": -33.63641357421875,
|
| 1959 |
+
"logps/rejected": -32.55451202392578,
|
| 1960 |
+
"loss": 0.1642,
|
| 1961 |
+
"rewards/accuracies": 0.9375,
|
| 1962 |
+
"rewards/chosen": 1.4071805477142334,
|
| 1963 |
+
"rewards/margins": 2.2361385822296143,
|
| 1964 |
+
"rewards/rejected": -0.8289580345153809,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.8602968454360962,
|
| 1971 |
+
"logits/rejected": -1.8607813119888306,
|
| 1972 |
+
"logps/chosen": -31.587940216064453,
|
| 1973 |
+
"logps/rejected": -34.207176208496094,
|
| 1974 |
+
"loss": 0.1465,
|
| 1975 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1976 |
+
"rewards/chosen": 1.1524927616119385,
|
| 1977 |
+
"rewards/margins": 2.2663419246673584,
|
| 1978 |
+
"rewards/rejected": -1.1138492822647095,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.800450086593628,
|
| 1985 |
+
"logits/rejected": -1.7998168468475342,
|
| 1986 |
+
"logps/chosen": -31.380176544189453,
|
| 1987 |
+
"logps/rejected": -34.821746826171875,
|
| 1988 |
+
"loss": 0.151,
|
| 1989 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1990 |
+
"rewards/chosen": 1.3685301542282104,
|
| 1991 |
+
"rewards/margins": 2.4094412326812744,
|
| 1992 |
+
"rewards/rejected": -1.0409109592437744,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.686762809753418,
|
| 1999 |
+
"logits/rejected": -1.6821815967559814,
|
| 2000 |
+
"logps/chosen": -33.52080535888672,
|
| 2001 |
+
"logps/rejected": -32.1031494140625,
|
| 2002 |
+
"loss": 0.1522,
|
| 2003 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2004 |
+
"rewards/chosen": 1.3733112812042236,
|
| 2005 |
+
"rewards/margins": 2.314640760421753,
|
| 2006 |
+
"rewards/rejected": -0.9413297772407532,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -1.8902908563613892,
|
| 2013 |
+
"logits/rejected": -1.8748180866241455,
|
| 2014 |
+
"logps/chosen": -32.840858459472656,
|
| 2015 |
+
"logps/rejected": -36.358482360839844,
|
| 2016 |
+
"loss": 0.1254,
|
| 2017 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2018 |
+
"rewards/chosen": 1.2053766250610352,
|
| 2019 |
+
"rewards/margins": 2.4799063205718994,
|
| 2020 |
+
"rewards/rejected": -1.274530053138733,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.83368718624115,
|
| 2027 |
+
"logits/rejected": -1.8386014699935913,
|
| 2028 |
+
"logps/chosen": -31.894006729125977,
|
| 2029 |
+
"logps/rejected": -37.62579345703125,
|
| 2030 |
+
"loss": 0.1199,
|
| 2031 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2032 |
+
"rewards/chosen": 1.298893690109253,
|
| 2033 |
+
"rewards/margins": 2.5169498920440674,
|
| 2034 |
+
"rewards/rejected": -1.2180559635162354,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.131683349609375,
|
| 2040 |
+
"eval_logits/rejected": -2.1269302368164062,
|
| 2041 |
+
"eval_logps/chosen": -34.37530517578125,
|
| 2042 |
+
"eval_logps/rejected": -38.09721374511719,
|
| 2043 |
+
"eval_loss": 0.6739928126335144,
|
| 2044 |
+
"eval_rewards/accuracies": 0.5980066657066345,
|
| 2045 |
+
"eval_rewards/chosen": -0.2726014256477356,
|
| 2046 |
+
"eval_rewards/margins": 0.19187170267105103,
|
| 2047 |
+
"eval_rewards/rejected": -0.464473158121109,
|
| 2048 |
+
"eval_runtime": 145.964,
|
| 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": -1.8072227239608765,
|
| 2057 |
+
"logits/rejected": -1.8072770833969116,
|
| 2058 |
+
"logps/chosen": -34.529273986816406,
|
| 2059 |
+
"logps/rejected": -38.355064392089844,
|
| 2060 |
+
"loss": 0.165,
|
| 2061 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2062 |
+
"rewards/chosen": 1.37179696559906,
|
| 2063 |
+
"rewards/margins": 2.388556957244873,
|
| 2064 |
+
"rewards/rejected": -1.0167598724365234,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -1.8941234350204468,
|
| 2071 |
+
"logits/rejected": -1.8927080631256104,
|
| 2072 |
+
"logps/chosen": -30.5273494720459,
|
| 2073 |
+
"logps/rejected": -32.267601013183594,
|
| 2074 |
+
"loss": 0.1563,
|
| 2075 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2076 |
+
"rewards/chosen": 1.2532024383544922,
|
| 2077 |
+
"rewards/margins": 2.304598093032837,
|
| 2078 |
+
"rewards/rejected": -1.0513951778411865,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.8009631633758545,
|
| 2085 |
+
"logits/rejected": -1.8024797439575195,
|
| 2086 |
+
"logps/chosen": -32.74066925048828,
|
| 2087 |
+
"logps/rejected": -34.96391296386719,
|
| 2088 |
+
"loss": 0.1456,
|
| 2089 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2090 |
+
"rewards/chosen": 1.306864619255066,
|
| 2091 |
+
"rewards/margins": 2.3590540885925293,
|
| 2092 |
+
"rewards/rejected": -1.052189588546753,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.8623323440551758,
|
| 2099 |
+
"logits/rejected": -1.8611522912979126,
|
| 2100 |
+
"logps/chosen": -29.76908302307129,
|
| 2101 |
+
"logps/rejected": -33.332550048828125,
|
| 2102 |
+
"loss": 0.1572,
|
| 2103 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2104 |
+
"rewards/chosen": 1.1436678171157837,
|
| 2105 |
+
"rewards/margins": 2.21612811088562,
|
| 2106 |
+
"rewards/rejected": -1.072460412979126,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -1.877587914466858,
|
| 2113 |
+
"logits/rejected": -1.862841248512268,
|
| 2114 |
+
"logps/chosen": -32.96019744873047,
|
| 2115 |
+
"logps/rejected": -37.72450637817383,
|
| 2116 |
+
"loss": 0.153,
|
| 2117 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2118 |
+
"rewards/chosen": 1.0851805210113525,
|
| 2119 |
+
"rewards/margins": 2.253241539001465,
|
| 2120 |
+
"rewards/rejected": -1.1680610179901123,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -1.8752319812774658,
|
| 2127 |
+
"logits/rejected": -1.8875735998153687,
|
| 2128 |
+
"logps/chosen": -30.424901962280273,
|
| 2129 |
+
"logps/rejected": -33.42066192626953,
|
| 2130 |
+
"loss": 0.1772,
|
| 2131 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2132 |
+
"rewards/chosen": 1.3148540258407593,
|
| 2133 |
+
"rewards/margins": 2.2595856189727783,
|
| 2134 |
+
"rewards/rejected": -0.9447315335273743,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -1.9719635248184204,
|
| 2141 |
+
"logits/rejected": -1.971278429031372,
|
| 2142 |
+
"logps/chosen": -30.404891967773438,
|
| 2143 |
+
"logps/rejected": -37.444129943847656,
|
| 2144 |
+
"loss": 0.1268,
|
| 2145 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2146 |
+
"rewards/chosen": 1.300431489944458,
|
| 2147 |
+
"rewards/margins": 2.5910286903381348,
|
| 2148 |
+
"rewards/rejected": -1.2905975580215454,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.8381322622299194,
|
| 2155 |
+
"logits/rejected": -1.845894455909729,
|
| 2156 |
+
"logps/chosen": -32.695735931396484,
|
| 2157 |
+
"logps/rejected": -38.82695388793945,
|
| 2158 |
+
"loss": 0.1845,
|
| 2159 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2160 |
+
"rewards/chosen": 1.1877927780151367,
|
| 2161 |
+
"rewards/margins": 2.2896358966827393,
|
| 2162 |
+
"rewards/rejected": -1.101843237876892,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.7921359539031982,
|
| 2169 |
+
"logits/rejected": -1.7879730463027954,
|
| 2170 |
+
"logps/chosen": -32.99811553955078,
|
| 2171 |
+
"logps/rejected": -32.61560821533203,
|
| 2172 |
+
"loss": 0.1972,
|
| 2173 |
+
"rewards/accuracies": 0.9375,
|
| 2174 |
+
"rewards/chosen": 1.4152759313583374,
|
| 2175 |
+
"rewards/margins": 2.229846715927124,
|
| 2176 |
+
"rewards/rejected": -0.8145707845687866,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -1.9499454498291016,
|
| 2183 |
+
"logits/rejected": -1.9474738836288452,
|
| 2184 |
+
"logps/chosen": -29.77544593811035,
|
| 2185 |
+
"logps/rejected": -33.36935806274414,
|
| 2186 |
+
"loss": 0.1578,
|
| 2187 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2188 |
+
"rewards/chosen": 1.2168843746185303,
|
| 2189 |
+
"rewards/margins": 2.1967949867248535,
|
| 2190 |
+
"rewards/rejected": -0.979910671710968,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.131394863128662,
|
| 2196 |
+
"eval_logits/rejected": -2.126645803451538,
|
| 2197 |
+
"eval_logps/chosen": -34.39291763305664,
|
| 2198 |
+
"eval_logps/rejected": -38.091670989990234,
|
| 2199 |
+
"eval_loss": 0.6838791370391846,
|
| 2200 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 2201 |
+
"eval_rewards/chosen": -0.28668972849845886,
|
| 2202 |
+
"eval_rewards/margins": 0.17335131764411926,
|
| 2203 |
+
"eval_rewards/rejected": -0.4600410461425781,
|
| 2204 |
+
"eval_runtime": 145.9578,
|
| 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": -1.8938592672348022,
|
| 2213 |
+
"logits/rejected": -1.8896372318267822,
|
| 2214 |
+
"logps/chosen": -32.033363342285156,
|
| 2215 |
+
"logps/rejected": -30.647994995117188,
|
| 2216 |
+
"loss": 0.1724,
|
| 2217 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2218 |
+
"rewards/chosen": 1.2873361110687256,
|
| 2219 |
+
"rewards/margins": 2.162166118621826,
|
| 2220 |
+
"rewards/rejected": -0.8748300671577454,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.8500086069107056,
|
| 2227 |
+
"logits/rejected": -1.8511066436767578,
|
| 2228 |
+
"logps/chosen": -31.751754760742188,
|
| 2229 |
+
"logps/rejected": -35.97992706298828,
|
| 2230 |
+
"loss": 0.1799,
|
| 2231 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2232 |
+
"rewards/chosen": 1.29049813747406,
|
| 2233 |
+
"rewards/margins": 2.1352338790893555,
|
| 2234 |
+
"rewards/rejected": -0.8447355031967163,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.8036447763442993,
|
| 2241 |
+
"logits/rejected": -1.7969812154769897,
|
| 2242 |
+
"logps/chosen": -32.511940002441406,
|
| 2243 |
+
"logps/rejected": -38.00764846801758,
|
| 2244 |
+
"loss": 0.1569,
|
| 2245 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2246 |
+
"rewards/chosen": 1.303862452507019,
|
| 2247 |
+
"rewards/margins": 2.358455181121826,
|
| 2248 |
+
"rewards/rejected": -1.0545929670333862,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.8134548664093018,
|
| 2255 |
+
"logits/rejected": -1.8079309463500977,
|
| 2256 |
+
"logps/chosen": -27.75140380859375,
|
| 2257 |
+
"logps/rejected": -32.93821334838867,
|
| 2258 |
+
"loss": 0.1916,
|
| 2259 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2260 |
+
"rewards/chosen": 1.2494693994522095,
|
| 2261 |
+
"rewards/margins": 2.2155184745788574,
|
| 2262 |
+
"rewards/rejected": -0.9660491943359375,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.8413394689559937,
|
| 2269 |
+
"logits/rejected": -1.8308031558990479,
|
| 2270 |
+
"logps/chosen": -31.208871841430664,
|
| 2271 |
+
"logps/rejected": -36.86098861694336,
|
| 2272 |
+
"loss": 0.1533,
|
| 2273 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2274 |
+
"rewards/chosen": 1.0181248188018799,
|
| 2275 |
+
"rewards/margins": 2.2786247730255127,
|
| 2276 |
+
"rewards/rejected": -1.260500192642212,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.8131767511367798,
|
| 2283 |
+
"logits/rejected": -1.8103145360946655,
|
| 2284 |
+
"logps/chosen": -31.82034683227539,
|
| 2285 |
+
"logps/rejected": -31.843425750732422,
|
| 2286 |
+
"loss": 0.2074,
|
| 2287 |
+
"rewards/accuracies": 0.9375,
|
| 2288 |
+
"rewards/chosen": 1.1915438175201416,
|
| 2289 |
+
"rewards/margins": 1.9536281824111938,
|
| 2290 |
+
"rewards/rejected": -0.7620843052864075,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -1.8745816946029663,
|
| 2297 |
+
"logits/rejected": -1.8810014724731445,
|
| 2298 |
+
"logps/chosen": -30.35885238647461,
|
| 2299 |
+
"logps/rejected": -34.228118896484375,
|
| 2300 |
+
"loss": 0.1567,
|
| 2301 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2302 |
+
"rewards/chosen": 1.3682581186294556,
|
| 2303 |
+
"rewards/margins": 2.380812883377075,
|
| 2304 |
+
"rewards/rejected": -1.01255464553833,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -1.8742589950561523,
|
| 2311 |
+
"logits/rejected": -1.8845386505126953,
|
| 2312 |
+
"logps/chosen": -30.218097686767578,
|
| 2313 |
+
"logps/rejected": -30.740955352783203,
|
| 2314 |
+
"loss": 0.168,
|
| 2315 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2316 |
+
"rewards/chosen": 1.1817275285720825,
|
| 2317 |
+
"rewards/margins": 2.0716397762298584,
|
| 2318 |
+
"rewards/rejected": -0.889912486076355,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.8073488473892212,
|
| 2325 |
+
"logits/rejected": -1.800286054611206,
|
| 2326 |
+
"logps/chosen": -32.302825927734375,
|
| 2327 |
+
"logps/rejected": -33.47874069213867,
|
| 2328 |
+
"loss": 0.1842,
|
| 2329 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2330 |
+
"rewards/chosen": 1.082926869392395,
|
| 2331 |
+
"rewards/margins": 2.097539186477661,
|
| 2332 |
+
"rewards/rejected": -1.0146124362945557,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.8606147766113281,
|
| 2339 |
+
"logits/rejected": -1.8575241565704346,
|
| 2340 |
+
"logps/chosen": -32.205726623535156,
|
| 2341 |
+
"logps/rejected": -34.68817138671875,
|
| 2342 |
+
"loss": 0.1614,
|
| 2343 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2344 |
+
"rewards/chosen": 1.1358753442764282,
|
| 2345 |
+
"rewards/margins": 2.2782340049743652,
|
| 2346 |
+
"rewards/rejected": -1.142358660697937,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.1312177181243896,
|
| 2352 |
+
"eval_logits/rejected": -2.1264684200286865,
|
| 2353 |
+
"eval_logps/chosen": -34.386234283447266,
|
| 2354 |
+
"eval_logps/rejected": -38.08888244628906,
|
| 2355 |
+
"eval_loss": 0.6819679737091064,
|
| 2356 |
+
"eval_rewards/accuracies": 0.574335515499115,
|
| 2357 |
+
"eval_rewards/chosen": -0.2813448905944824,
|
| 2358 |
+
"eval_rewards/margins": 0.17646357417106628,
|
| 2359 |
+
"eval_rewards/rejected": -0.4578084647655487,
|
| 2360 |
+
"eval_runtime": 145.9567,
|
| 2361 |
+
"eval_samples_per_second": 2.35,
|
| 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": -1.9368705749511719,
|
| 2369 |
+
"logits/rejected": -1.9374831914901733,
|
| 2370 |
+
"logps/chosen": -28.38448715209961,
|
| 2371 |
+
"logps/rejected": -34.06378173828125,
|
| 2372 |
+
"loss": 0.2003,
|
| 2373 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2374 |
+
"rewards/chosen": 1.145431399345398,
|
| 2375 |
+
"rewards/margins": 2.204097032546997,
|
| 2376 |
+
"rewards/rejected": -1.0586655139923096,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.7932342290878296,
|
| 2383 |
+
"logits/rejected": -1.7930355072021484,
|
| 2384 |
+
"logps/chosen": -31.60140037536621,
|
| 2385 |
+
"logps/rejected": -35.09097671508789,
|
| 2386 |
+
"loss": 0.2089,
|
| 2387 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2388 |
+
"rewards/chosen": 1.1523739099502563,
|
| 2389 |
+
"rewards/margins": 2.1261672973632812,
|
| 2390 |
+
"rewards/rejected": -0.9737933874130249,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -1.943830132484436,
|
| 2397 |
+
"logits/rejected": -1.9411506652832031,
|
| 2398 |
+
"logps/chosen": -30.156597137451172,
|
| 2399 |
+
"logps/rejected": -32.75658416748047,
|
| 2400 |
+
"loss": 0.1839,
|
| 2401 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2402 |
+
"rewards/chosen": 1.053972840309143,
|
| 2403 |
+
"rewards/margins": 2.0723206996917725,
|
| 2404 |
+
"rewards/rejected": -1.018348217010498,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -1.9267715215682983,
|
| 2411 |
+
"logits/rejected": -1.9278953075408936,
|
| 2412 |
+
"logps/chosen": -29.27459716796875,
|
| 2413 |
+
"logps/rejected": -30.95477294921875,
|
| 2414 |
+
"loss": 0.1802,
|
| 2415 |
+
"rewards/accuracies": 0.98333340883255,
|
| 2416 |
+
"rewards/chosen": 1.0762757062911987,
|
| 2417 |
+
"rewards/margins": 2.007948398590088,
|
| 2418 |
+
"rewards/rejected": -0.9316726922988892,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 0.20605388523696305,
|
| 2426 |
+
"train_runtime": 10804.1609,
|
| 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,
|