Commit
•
eadaeab
1
Parent(s):
5bb146a
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 +1904 -88
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
| 4 |
-
- alignment-handbook
|
| 5 |
- trl
|
| 6 |
- dpo
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
| 9 |
-
datasets:
|
| 10 |
-
- hugodk-sch/aftonposten_title_prefs
|
| 11 |
model-index:
|
| 12 |
- name: aftonposten-6b-align-scan
|
| 13 |
results: []
|
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
# aftonposten-6b-align-scan
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 0.
|
| 24 |
-
- Rewards/chosen: 0.
|
| 25 |
-
- Rewards/rejected: 0.
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -33.
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,15 +55,27 @@ The following hyperparameters were used during training:
|
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
| 59 |
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
-
- num_epochs:
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
-
| Training Loss | Epoch | Step |
|
| 65 |
-
|
| 66 |
-
| 0.4997 | 0.26 | 100 |
|
| 67 |
-
| 0.4992 | 0.52 | 200 |
|
| 68 |
-
| 0.4982 | 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: 0.4995
|
| 22 |
+
- Rewards/chosen: 0.0025
|
| 23 |
+
- Rewards/rejected: -0.0003
|
| 24 |
+
- Rewards/accuracies: 0.6034
|
| 25 |
+
- Rewards/margins: 0.0028
|
| 26 |
+
- Logps/rejected: -37.5427
|
| 27 |
+
- Logps/chosen: -33.7835
|
| 28 |
+
- Logits/rejected: -2.0908
|
| 29 |
+
- Logits/chosen: -2.0954
|
| 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 |
+
| 0.4997 | 0.26 | 100 | -2.2320 | -2.2272 | -33.8530 | -37.3521 | 0.5000 | 0.5183 | 0.0018 | 0.0002 | 0.0016 |
|
| 65 |
+
| 0.4992 | 0.52 | 200 | -2.2288 | -2.2240 | -33.7422 | -37.2812 | 0.4999 | 0.5748 | 0.0029 | 0.0006 | 0.0024 |
|
| 66 |
+
| 0.4982 | 0.78 | 300 | -2.2289 | -2.2241 | -33.7545 | -37.2741 | 0.4999 | 0.5042 | 0.0028 | 0.0004 | 0.0024 |
|
| 67 |
+
| 0.4967 | 1.04 | 400 | 0.4999 | 0.0036 | 0.0030 | 0.5623 | 0.0006 | -37.2152 | -33.6722 | -2.2047 | -2.2095 |
|
| 68 |
+
| 0.4954 | 1.3 | 500 | 0.4998 | 0.0042 | 0.0033 | 0.5772 | 0.0009 | -37.1891 | -33.6191 | -2.1759 | -2.1807 |
|
| 69 |
+
| 0.4964 | 1.56 | 600 | 0.4996 | 0.0044 | 0.0026 | 0.5685 | 0.0018 | -37.2574 | -33.5925 | -2.1522 | -2.1569 |
|
| 70 |
+
| 0.4942 | 1.82 | 700 | 0.4996 | 0.0034 | 0.0018 | 0.5797 | 0.0016 | -37.3406 | -33.6981 | -2.1370 | -2.1417 |
|
| 71 |
+
| 0.4926 | 2.08 | 800 | 0.4995 | 0.0037 | 0.0013 | 0.5947 | 0.0024 | -37.3864 | -33.6678 | -2.1276 | -2.1323 |
|
| 72 |
+
| 0.491 | 2.34 | 900 | 0.4995 | 0.0031 | 0.0006 | 0.5739 | 0.0025 | -37.4562 | -33.7214 | -2.1107 | -2.1154 |
|
| 73 |
+
| 0.4941 | 2.6 | 1000 | 0.4995 | 0.0028 | 0.0001 | 0.6059 | 0.0027 | -37.5025 | -33.7503 | -2.0960 | -2.1006 |
|
| 74 |
+
| 0.4943 | 2.86 | 1100 | 0.4995 | 0.0025 | -0.0002 | 0.5826 | 0.0028 | -37.5382 | -33.7800 | -2.0918 | -2.0965 |
|
| 75 |
+
| 0.4911 | 3.12 | 1200 | 0.4995 | 0.0026 | -0.0001 | 0.6009 | 0.0027 | -37.5257 | -33.7754 | -2.0913 | -2.0959 |
|
| 76 |
+
| 0.4906 | 3.38 | 1300 | 0.4995 | 0.0025 | -0.0002 | 0.6034 | 0.0027 | -37.5366 | -33.7817 | -2.0909 | -2.0955 |
|
| 77 |
+
| 0.4928 | 3.64 | 1400 | 0.4995 | 0.0025 | -0.0003 | 0.5922 | 0.0028 | -37.5423 | -33.7843 | -2.0903 | -2.0950 |
|
| 78 |
+
| 0.4912 | 3.9 | 1500 | 0.4995 | 0.0025 | -0.0003 | 0.6034 | 0.0028 | -37.5427 | -33.7835 | -2.0908 | -2.0954 |
|
| 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:ae05e30e298858cdca7a0c71dc8e5fe02939042f934bc13d1830d5b1bfc60782
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.228447198867798,
|
| 4 |
"eval_logits/rejected": -2.223632335662842,
|
| 5 |
"eval_logps/chosen": -33.74469757080078,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.357,
|
| 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": -2.228447198867798,
|
| 4 |
"eval_logits/rejected": -2.223632335662842,
|
| 5 |
"eval_logps/chosen": -33.74469757080078,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.357,
|
| 15 |
"eval_steps_per_second": 0.296,
|
| 16 |
+
"train_loss": 0.39746343532165923,
|
| 17 |
+
"train_runtime": 10799.7737,
|
| 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": 0.39746343532165923,
|
| 4 |
+
"train_runtime": 10799.7737,
|
| 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": -33.
|
| 513 |
-
"logps/rejected": -33.
|
| 514 |
"loss": 0.4988,
|
| 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": -1.
|
| 525 |
-
"logits/rejected": -1.
|
| 526 |
-
"logps/chosen": -32.
|
| 527 |
-
"logps/rejected": -32.
|
| 528 |
-
"loss": 0.
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": 0.
|
| 531 |
-
"rewards/margins": 0.
|
| 532 |
-
"rewards/rejected": 0.
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -33.
|
| 541 |
-
"logps/rejected": -32.
|
| 542 |
-
"loss": 0.
|
| 543 |
-
"rewards/accuracies": 0.
|
| 544 |
-
"rewards/chosen": 0.
|
| 545 |
-
"rewards/margins": 0.
|
| 546 |
-
"rewards/rejected": 0.
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
-
"learning_rate":
|
| 552 |
-
"logits/chosen": -1.
|
| 553 |
-
"logits/rejected": -1.
|
| 554 |
-
"logps/chosen": -32.
|
| 555 |
-
"logps/rejected": -32.
|
| 556 |
-
"loss": 0.
|
| 557 |
-
"rewards/accuracies": 0.
|
| 558 |
-
"rewards/chosen": 0.
|
| 559 |
-
"rewards/margins": 0.
|
| 560 |
-
"rewards/rejected": 0.
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
-
"learning_rate":
|
| 566 |
-
"logits/chosen": -1.
|
| 567 |
-
"logits/rejected": -1.
|
| 568 |
-
"logps/chosen": -31.
|
| 569 |
-
"logps/rejected": -
|
| 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": -32.
|
| 583 |
-
"logps/rejected": -28.
|
| 584 |
-
"loss": 0.
|
| 585 |
-
"rewards/accuracies": 0.
|
| 586 |
-
"rewards/chosen": 0.
|
| 587 |
-
"rewards/margins": 0.
|
| 588 |
-
"rewards/rejected": 0.
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
-
"learning_rate":
|
| 594 |
-
"logits/chosen": -1.
|
| 595 |
-
"logits/rejected": -1.
|
| 596 |
-
"logps/chosen": -33.
|
| 597 |
-
"logps/rejected": -30.
|
| 598 |
-
"loss": 0.
|
| 599 |
"rewards/accuracies": 0.7250000238418579,
|
| 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": -1.914014458656311,
|
| 497 |
+
"logits/rejected": -1.9108896255493164,
|
| 498 |
+
"logps/chosen": -31.016780853271484,
|
| 499 |
+
"logps/rejected": -33.5390739440918,
|
| 500 |
+
"loss": 0.4991,
|
| 501 |
+
"rewards/accuracies": 0.699999988079071,
|
| 502 |
+
"rewards/chosen": 0.005865688901394606,
|
| 503 |
+
"rewards/margins": 0.0038349279202520847,
|
| 504 |
+
"rewards/rejected": 0.002030761446803808,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.9622209072113037,
|
| 511 |
+
"logits/rejected": -1.9501073360443115,
|
| 512 |
+
"logps/chosen": -33.98822784423828,
|
| 513 |
+
"logps/rejected": -33.42522048950195,
|
| 514 |
"loss": 0.4988,
|
| 515 |
+
"rewards/accuracies": 0.737500011920929,
|
| 516 |
+
"rewards/chosen": 0.0060784295201301575,
|
| 517 |
+
"rewards/margins": 0.00494221830740571,
|
| 518 |
+
"rewards/rejected": 0.0011362109798938036,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -1.997842788696289,
|
| 525 |
+
"logits/rejected": -1.9965225458145142,
|
| 526 |
+
"logps/chosen": -32.7388801574707,
|
| 527 |
+
"logps/rejected": -32.22772216796875,
|
| 528 |
+
"loss": 0.4988,
|
| 529 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 530 |
+
"rewards/chosen": 0.00746917212381959,
|
| 531 |
+
"rewards/margins": 0.0049505168572068214,
|
| 532 |
+
"rewards/rejected": 0.002518654800951481,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.0849790573120117,
|
| 539 |
+
"logits/rejected": -2.06939435005188,
|
| 540 |
+
"logps/chosen": -33.35835266113281,
|
| 541 |
+
"logps/rejected": -32.73654556274414,
|
| 542 |
+
"loss": 0.4989,
|
| 543 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 544 |
+
"rewards/chosen": 0.008210408501327038,
|
| 545 |
+
"rewards/margins": 0.004395507741719484,
|
| 546 |
+
"rewards/rejected": 0.0038149021565914154,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.9578001499176025,
|
| 553 |
+
"logits/rejected": -1.9570014476776123,
|
| 554 |
+
"logps/chosen": -32.38352966308594,
|
| 555 |
+
"logps/rejected": -32.16341781616211,
|
| 556 |
+
"loss": 0.4985,
|
| 557 |
+
"rewards/accuracies": 0.699999988079071,
|
| 558 |
+
"rewards/chosen": 0.008985674008727074,
|
| 559 |
+
"rewards/margins": 0.005968388635665178,
|
| 560 |
+
"rewards/rejected": 0.0030172846745699644,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.9111387729644775,
|
| 567 |
+
"logits/rejected": -1.9213663339614868,
|
| 568 |
+
"logps/chosen": -31.330036163330078,
|
| 569 |
+
"logps/rejected": -34.964256286621094,
|
| 570 |
+
"loss": 0.4986,
|
| 571 |
+
"rewards/accuracies": 0.675000011920929,
|
| 572 |
+
"rewards/chosen": 0.008746545761823654,
|
| 573 |
+
"rewards/margins": 0.005501560866832733,
|
| 574 |
+
"rewards/rejected": 0.003244984894990921,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.0449702739715576,
|
| 581 |
+
"logits/rejected": -2.0385282039642334,
|
| 582 |
+
"logps/chosen": -32.85245132446289,
|
| 583 |
+
"logps/rejected": -28.954187393188477,
|
| 584 |
+
"loss": 0.4987,
|
| 585 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 586 |
+
"rewards/chosen": 0.0079951835796237,
|
| 587 |
+
"rewards/margins": 0.005272808950394392,
|
| 588 |
+
"rewards/rejected": 0.002722373930737376,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.901015281677246,
|
| 595 |
+
"logits/rejected": -1.9031999111175537,
|
| 596 |
+
"logps/chosen": -33.20323944091797,
|
| 597 |
+
"logps/rejected": -30.677053451538086,
|
| 598 |
+
"loss": 0.498,
|
| 599 |
"rewards/accuracies": 0.7250000238418579,
|
| 600 |
+
"rewards/chosen": 0.010358055122196674,
|
| 601 |
+
"rewards/margins": 0.00840230192989111,
|
| 602 |
+
"rewards/rejected": 0.001955752493813634,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9289333820343018,
|
| 609 |
+
"logits/rejected": -1.9276466369628906,
|
| 610 |
+
"logps/chosen": -33.32128143310547,
|
| 611 |
+
"logps/rejected": -35.63185119628906,
|
| 612 |
+
"loss": 0.4975,
|
| 613 |
+
"rewards/accuracies": 0.7583333253860474,
|
| 614 |
+
"rewards/chosen": 0.01063510961830616,
|
| 615 |
+
"rewards/margins": 0.010393550619482994,
|
| 616 |
+
"rewards/rejected": 0.00024155918799806386,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.863186240196228,
|
| 623 |
+
"logits/rejected": -1.8549854755401611,
|
| 624 |
+
"logps/chosen": -30.498249053955078,
|
| 625 |
+
"logps/rejected": -36.088096618652344,
|
| 626 |
+
"loss": 0.4967,
|
| 627 |
+
"rewards/accuracies": 0.875,
|
| 628 |
+
"rewards/chosen": 0.012757278978824615,
|
| 629 |
+
"rewards/margins": 0.014232242479920387,
|
| 630 |
+
"rewards/rejected": -0.001474961405619979,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.2094738483428955,
|
| 636 |
+
"eval_logits/rejected": -2.2046518325805664,
|
| 637 |
+
"eval_logps/chosen": -33.672157287597656,
|
| 638 |
+
"eval_logps/rejected": -37.215171813964844,
|
| 639 |
+
"eval_loss": 0.4998500943183899,
|
| 640 |
+
"eval_rewards/accuracies": 0.5622923374176025,
|
| 641 |
+
"eval_rewards/chosen": 0.0036239090841263533,
|
| 642 |
+
"eval_rewards/margins": 0.0006094546988606453,
|
| 643 |
+
"eval_rewards/rejected": 0.003014454385265708,
|
| 644 |
+
"eval_runtime": 146.2447,
|
| 645 |
+
"eval_samples_per_second": 2.345,
|
| 646 |
+
"eval_steps_per_second": 0.294,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 1.06,
|
| 651 |
+
"learning_rate": 4.590793060955158e-06,
|
| 652 |
+
"logits/chosen": -2.0367393493652344,
|
| 653 |
+
"logits/rejected": -2.0395472049713135,
|
| 654 |
+
"logps/chosen": -31.686859130859375,
|
| 655 |
+
"logps/rejected": -34.91094970703125,
|
| 656 |
+
"loss": 0.4966,
|
| 657 |
+
"rewards/accuracies": 0.862500011920929,
|
| 658 |
+
"rewards/chosen": 0.013670777902007103,
|
| 659 |
+
"rewards/margins": 0.0148518281057477,
|
| 660 |
+
"rewards/rejected": -0.0011810490395873785,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.8766273260116577,
|
| 667 |
+
"logits/rejected": -1.8752492666244507,
|
| 668 |
+
"logps/chosen": -27.93256187438965,
|
| 669 |
+
"logps/rejected": -32.38922882080078,
|
| 670 |
+
"loss": 0.4973,
|
| 671 |
+
"rewards/accuracies": 0.925000011920929,
|
| 672 |
+
"rewards/chosen": 0.011349949985742569,
|
| 673 |
+
"rewards/margins": 0.012065614573657513,
|
| 674 |
+
"rewards/rejected": -0.0007156648789532483,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.834745168685913,
|
| 681 |
+
"logits/rejected": -1.8277431726455688,
|
| 682 |
+
"logps/chosen": -32.58226013183594,
|
| 683 |
+
"logps/rejected": -34.0352783203125,
|
| 684 |
+
"loss": 0.4969,
|
| 685 |
+
"rewards/accuracies": 0.9375,
|
| 686 |
+
"rewards/chosen": 0.014918235130608082,
|
| 687 |
+
"rewards/margins": 0.012818296439945698,
|
| 688 |
+
"rewards/rejected": 0.0020999389234930277,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -2.005305767059326,
|
| 695 |
+
"logits/rejected": -2.0002236366271973,
|
| 696 |
+
"logps/chosen": -30.26675033569336,
|
| 697 |
+
"logps/rejected": -32.066688537597656,
|
| 698 |
+
"loss": 0.4962,
|
| 699 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 700 |
+
"rewards/chosen": 0.015165433287620544,
|
| 701 |
+
"rewards/margins": 0.01615620031952858,
|
| 702 |
+
"rewards/rejected": -0.0009907669154927135,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.8884214162826538,
|
| 709 |
+
"logits/rejected": -1.892564058303833,
|
| 710 |
+
"logps/chosen": -32.953224182128906,
|
| 711 |
+
"logps/rejected": -32.72943878173828,
|
| 712 |
+
"loss": 0.4961,
|
| 713 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 714 |
+
"rewards/chosen": 0.01562627963721752,
|
| 715 |
+
"rewards/margins": 0.017034271731972694,
|
| 716 |
+
"rewards/rejected": -0.0014079909306019545,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.8940017223358154,
|
| 723 |
+
"logits/rejected": -1.888357162475586,
|
| 724 |
+
"logps/chosen": -33.757293701171875,
|
| 725 |
+
"logps/rejected": -35.25668716430664,
|
| 726 |
+
"loss": 0.4956,
|
| 727 |
+
"rewards/accuracies": 0.875,
|
| 728 |
+
"rewards/chosen": 0.016622314229607582,
|
| 729 |
+
"rewards/margins": 0.01969819888472557,
|
| 730 |
+
"rewards/rejected": -0.003075886517763138,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.9210342168807983,
|
| 737 |
+
"logits/rejected": -1.920659065246582,
|
| 738 |
+
"logps/chosen": -32.557655334472656,
|
| 739 |
+
"logps/rejected": -34.18804168701172,
|
| 740 |
+
"loss": 0.4962,
|
| 741 |
+
"rewards/accuracies": 0.887499988079071,
|
| 742 |
+
"rewards/chosen": 0.01610674150288105,
|
| 743 |
+
"rewards/margins": 0.01613234356045723,
|
| 744 |
+
"rewards/rejected": -2.560380380600691e-05,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -1.9694734811782837,
|
| 751 |
+
"logits/rejected": -1.9510208368301392,
|
| 752 |
+
"logps/chosen": -32.425758361816406,
|
| 753 |
+
"logps/rejected": -33.256771087646484,
|
| 754 |
+
"loss": 0.4956,
|
| 755 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 756 |
+
"rewards/chosen": 0.01735007017850876,
|
| 757 |
+
"rewards/margins": 0.01921800896525383,
|
| 758 |
+
"rewards/rejected": -0.001867939019575715,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.934520959854126,
|
| 765 |
+
"logits/rejected": -1.934033751487732,
|
| 766 |
+
"logps/chosen": -29.948516845703125,
|
| 767 |
+
"logps/rejected": -31.202651977539062,
|
| 768 |
+
"loss": 0.4961,
|
| 769 |
+
"rewards/accuracies": 0.824999988079071,
|
| 770 |
+
"rewards/chosen": 0.016627881675958633,
|
| 771 |
+
"rewards/margins": 0.016193937510252,
|
| 772 |
+
"rewards/rejected": 0.0004339427687227726,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.8985036611557007,
|
| 779 |
+
"logits/rejected": -1.891710638999939,
|
| 780 |
+
"logps/chosen": -30.8389892578125,
|
| 781 |
+
"logps/rejected": -34.76033020019531,
|
| 782 |
+
"loss": 0.4954,
|
| 783 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 784 |
+
"rewards/chosen": 0.018311362713575363,
|
| 785 |
+
"rewards/margins": 0.01964724436402321,
|
| 786 |
+
"rewards/rejected": -0.0013358818832784891,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.1807026863098145,
|
| 792 |
+
"eval_logits/rejected": -2.1759257316589355,
|
| 793 |
+
"eval_logps/chosen": -33.619056701660156,
|
| 794 |
+
"eval_logps/rejected": -37.18906021118164,
|
| 795 |
+
"eval_loss": 0.4997919797897339,
|
| 796 |
+
"eval_rewards/accuracies": 0.5772424936294556,
|
| 797 |
+
"eval_rewards/chosen": 0.0041549475863575935,
|
| 798 |
+
"eval_rewards/margins": 0.0008793265442363918,
|
| 799 |
+
"eval_rewards/rejected": 0.003275620751082897,
|
| 800 |
+
"eval_runtime": 145.8925,
|
| 801 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.892907738685608,
|
| 809 |
+
"logits/rejected": -1.8967831134796143,
|
| 810 |
+
"logps/chosen": -27.704700469970703,
|
| 811 |
+
"logps/rejected": -32.9903450012207,
|
| 812 |
+
"loss": 0.4963,
|
| 813 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 814 |
+
"rewards/chosen": 0.014701254665851593,
|
| 815 |
+
"rewards/margins": 0.016011826694011688,
|
| 816 |
+
"rewards/rejected": -0.001310571446083486,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.8976653814315796,
|
| 823 |
+
"logits/rejected": -1.9082101583480835,
|
| 824 |
+
"logps/chosen": -31.580432891845703,
|
| 825 |
+
"logps/rejected": -30.9519100189209,
|
| 826 |
+
"loss": 0.4955,
|
| 827 |
+
"rewards/accuracies": 0.875,
|
| 828 |
+
"rewards/chosen": 0.016818998381495476,
|
| 829 |
+
"rewards/margins": 0.01982123777270317,
|
| 830 |
+
"rewards/rejected": -0.003002240788191557,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.8399875164031982,
|
| 837 |
+
"logits/rejected": -1.8377044200897217,
|
| 838 |
+
"logps/chosen": -29.908676147460938,
|
| 839 |
+
"logps/rejected": -30.362207412719727,
|
| 840 |
+
"loss": 0.4955,
|
| 841 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 842 |
+
"rewards/chosen": 0.01758403517305851,
|
| 843 |
+
"rewards/margins": 0.019438493996858597,
|
| 844 |
+
"rewards/rejected": -0.0018544571939855814,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.920372724533081,
|
| 851 |
+
"logits/rejected": -1.9134242534637451,
|
| 852 |
+
"logps/chosen": -32.44926071166992,
|
| 853 |
+
"logps/rejected": -33.200767517089844,
|
| 854 |
+
"loss": 0.4955,
|
| 855 |
+
"rewards/accuracies": 0.925000011920929,
|
| 856 |
+
"rewards/chosen": 0.017835671082139015,
|
| 857 |
+
"rewards/margins": 0.019668519496917725,
|
| 858 |
+
"rewards/rejected": -0.0018328496953472495,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.885063886642456,
|
| 865 |
+
"logits/rejected": -1.8914177417755127,
|
| 866 |
+
"logps/chosen": -30.19329261779785,
|
| 867 |
+
"logps/rejected": -32.799842834472656,
|
| 868 |
+
"loss": 0.4963,
|
| 869 |
+
"rewards/accuracies": 0.8125,
|
| 870 |
+
"rewards/chosen": 0.015598910860717297,
|
| 871 |
+
"rewards/margins": 0.01600998267531395,
|
| 872 |
+
"rewards/rejected": -0.0004110717272851616,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -1.9558032751083374,
|
| 879 |
+
"logits/rejected": -1.9331070184707642,
|
| 880 |
+
"logps/chosen": -29.753009796142578,
|
| 881 |
+
"logps/rejected": -33.04769515991211,
|
| 882 |
+
"loss": 0.4959,
|
| 883 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 884 |
+
"rewards/chosen": 0.0168455857783556,
|
| 885 |
+
"rewards/margins": 0.017361946403980255,
|
| 886 |
+
"rewards/rejected": -0.0005163621390238404,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.8903181552886963,
|
| 893 |
+
"logits/rejected": -1.8854230642318726,
|
| 894 |
+
"logps/chosen": -30.62668228149414,
|
| 895 |
+
"logps/rejected": -32.05100631713867,
|
| 896 |
+
"loss": 0.4959,
|
| 897 |
+
"rewards/accuracies": 0.8125,
|
| 898 |
+
"rewards/chosen": 0.01790049858391285,
|
| 899 |
+
"rewards/margins": 0.01720140501856804,
|
| 900 |
+
"rewards/rejected": 0.0006990955444052815,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -1.930395483970642,
|
| 907 |
+
"logits/rejected": -1.932220220565796,
|
| 908 |
+
"logps/chosen": -32.589778900146484,
|
| 909 |
+
"logps/rejected": -35.16767883300781,
|
| 910 |
+
"loss": 0.4952,
|
| 911 |
+
"rewards/accuracies": 0.925000011920929,
|
| 912 |
+
"rewards/chosen": 0.017860129475593567,
|
| 913 |
+
"rewards/margins": 0.021527757868170738,
|
| 914 |
+
"rewards/rejected": -0.0036676295567303896,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.9129533767700195,
|
| 921 |
+
"logits/rejected": -1.9123092889785767,
|
| 922 |
+
"logps/chosen": -28.789093017578125,
|
| 923 |
+
"logps/rejected": -31.536624908447266,
|
| 924 |
+
"loss": 0.4956,
|
| 925 |
+
"rewards/accuracies": 0.887499988079071,
|
| 926 |
+
"rewards/chosen": 0.018029652535915375,
|
| 927 |
+
"rewards/margins": 0.01877954974770546,
|
| 928 |
+
"rewards/rejected": -0.0007498954655602574,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -1.9223308563232422,
|
| 935 |
+
"logits/rejected": -1.9202518463134766,
|
| 936 |
+
"logps/chosen": -31.292776107788086,
|
| 937 |
+
"logps/rejected": -32.150634765625,
|
| 938 |
+
"loss": 0.4964,
|
| 939 |
+
"rewards/accuracies": 0.925000011920929,
|
| 940 |
+
"rewards/chosen": 0.017055068165063858,
|
| 941 |
+
"rewards/margins": 0.015561839565634727,
|
| 942 |
+
"rewards/rejected": 0.0014932285994291306,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.1569008827209473,
|
| 948 |
+
"eval_logits/rejected": -2.1521544456481934,
|
| 949 |
+
"eval_logps/chosen": -33.592525482177734,
|
| 950 |
+
"eval_logps/rejected": -37.25735855102539,
|
| 951 |
+
"eval_loss": 0.4995650053024292,
|
| 952 |
+
"eval_rewards/accuracies": 0.5685215592384338,
|
| 953 |
+
"eval_rewards/chosen": 0.004420237150043249,
|
| 954 |
+
"eval_rewards/margins": 0.0018276458140462637,
|
| 955 |
+
"eval_rewards/rejected": 0.0025925911031663418,
|
| 956 |
+
"eval_runtime": 145.7896,
|
| 957 |
+
"eval_samples_per_second": 2.353,
|
| 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": -1.9622409343719482,
|
| 965 |
+
"logits/rejected": -1.968703031539917,
|
| 966 |
+
"logps/chosen": -30.59027099609375,
|
| 967 |
+
"logps/rejected": -32.09941864013672,
|
| 968 |
+
"loss": 0.4961,
|
| 969 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 970 |
+
"rewards/chosen": 0.016471117734909058,
|
| 971 |
+
"rewards/margins": 0.016833433881402016,
|
| 972 |
+
"rewards/rejected": -0.0003623150405474007,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -1.9154777526855469,
|
| 979 |
+
"logits/rejected": -1.913214921951294,
|
| 980 |
+
"logps/chosen": -32.721229553222656,
|
| 981 |
+
"logps/rejected": -31.03836441040039,
|
| 982 |
+
"loss": 0.4951,
|
| 983 |
+
"rewards/accuracies": 0.862500011920929,
|
| 984 |
+
"rewards/chosen": 0.0203465037047863,
|
| 985 |
+
"rewards/margins": 0.020875070244073868,
|
| 986 |
+
"rewards/rejected": -0.000528569333255291,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.8615707159042358,
|
| 993 |
+
"logits/rejected": -1.8549633026123047,
|
| 994 |
+
"logps/chosen": -33.12920379638672,
|
| 995 |
+
"logps/rejected": -32.32065963745117,
|
| 996 |
+
"loss": 0.4946,
|
| 997 |
+
"rewards/accuracies": 0.875,
|
| 998 |
+
"rewards/chosen": 0.023350173607468605,
|
| 999 |
+
"rewards/margins": 0.02298470214009285,
|
| 1000 |
+
"rewards/rejected": 0.00036546969204209745,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.8696269989013672,
|
| 1007 |
+
"logits/rejected": -1.8758395910263062,
|
| 1008 |
+
"logps/chosen": -32.165340423583984,
|
| 1009 |
+
"logps/rejected": -33.2056770324707,
|
| 1010 |
+
"loss": 0.4958,
|
| 1011 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1012 |
+
"rewards/chosen": 0.018139610067009926,
|
| 1013 |
+
"rewards/margins": 0.01779373362660408,
|
| 1014 |
+
"rewards/rejected": 0.0003458770806901157,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -1.952419638633728,
|
| 1021 |
+
"logits/rejected": -1.966408371925354,
|
| 1022 |
+
"logps/chosen": -30.081417083740234,
|
| 1023 |
+
"logps/rejected": -31.9990291595459,
|
| 1024 |
+
"loss": 0.4958,
|
| 1025 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1026 |
+
"rewards/chosen": 0.017710469663143158,
|
| 1027 |
+
"rewards/margins": 0.017723901197314262,
|
| 1028 |
+
"rewards/rejected": -1.3430044418782927e-05,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.828285574913025,
|
| 1035 |
+
"logits/rejected": -1.8253133296966553,
|
| 1036 |
+
"logps/chosen": -31.497600555419922,
|
| 1037 |
+
"logps/rejected": -35.03656005859375,
|
| 1038 |
+
"loss": 0.4936,
|
| 1039 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1040 |
+
"rewards/chosen": 0.024253567680716515,
|
| 1041 |
+
"rewards/margins": 0.02761664055287838,
|
| 1042 |
+
"rewards/rejected": -0.0033630705438554287,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -1.9160382747650146,
|
| 1049 |
+
"logits/rejected": -1.916337251663208,
|
| 1050 |
+
"logps/chosen": -29.65035057067871,
|
| 1051 |
+
"logps/rejected": -34.16764831542969,
|
| 1052 |
+
"loss": 0.4957,
|
| 1053 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1054 |
+
"rewards/chosen": 0.01635240949690342,
|
| 1055 |
+
"rewards/margins": 0.01976504735648632,
|
| 1056 |
+
"rewards/rejected": -0.003412640420719981,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.8645700216293335,
|
| 1063 |
+
"logits/rejected": -1.861588716506958,
|
| 1064 |
+
"logps/chosen": -29.15572166442871,
|
| 1065 |
+
"logps/rejected": -33.00225067138672,
|
| 1066 |
+
"loss": 0.4957,
|
| 1067 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1068 |
+
"rewards/chosen": 0.01716628111898899,
|
| 1069 |
+
"rewards/margins": 0.01901078037917614,
|
| 1070 |
+
"rewards/rejected": -0.0018444998422637582,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -1.8963727951049805,
|
| 1077 |
+
"logits/rejected": -1.8964437246322632,
|
| 1078 |
+
"logps/chosen": -27.984289169311523,
|
| 1079 |
+
"logps/rejected": -30.99312973022461,
|
| 1080 |
+
"loss": 0.4943,
|
| 1081 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1082 |
+
"rewards/chosen": 0.02178804762661457,
|
| 1083 |
+
"rewards/margins": 0.025014396756887436,
|
| 1084 |
+
"rewards/rejected": -0.0032263505272567272,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.8081165552139282,
|
| 1091 |
+
"logits/rejected": -1.81146240234375,
|
| 1092 |
+
"logps/chosen": -32.05728530883789,
|
| 1093 |
+
"logps/rejected": -31.581283569335938,
|
| 1094 |
+
"loss": 0.4942,
|
| 1095 |
+
"rewards/accuracies": 0.8125,
|
| 1096 |
+
"rewards/chosen": 0.024211909621953964,
|
| 1097 |
+
"rewards/margins": 0.025377869606018066,
|
| 1098 |
+
"rewards/rejected": -0.001165962778031826,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.1417086124420166,
|
| 1104 |
+
"eval_logits/rejected": -2.1370012760162354,
|
| 1105 |
+
"eval_logps/chosen": -33.69810485839844,
|
| 1106 |
+
"eval_logps/rejected": -37.340553283691406,
|
| 1107 |
+
"eval_loss": 0.49963295459747314,
|
| 1108 |
+
"eval_rewards/accuracies": 0.579734206199646,
|
| 1109 |
+
"eval_rewards/chosen": 0.0033644884824752808,
|
| 1110 |
+
"eval_rewards/margins": 0.0016038385219871998,
|
| 1111 |
+
"eval_rewards/rejected": 0.0017606498440727592,
|
| 1112 |
+
"eval_runtime": 145.8796,
|
| 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": -1.8899507522583008,
|
| 1121 |
+
"logits/rejected": -1.8811132907867432,
|
| 1122 |
+
"logps/chosen": -34.376182556152344,
|
| 1123 |
+
"logps/rejected": -32.027015686035156,
|
| 1124 |
+
"loss": 0.4948,
|
| 1125 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1126 |
+
"rewards/chosen": 0.021358376368880272,
|
| 1127 |
+
"rewards/margins": 0.02257036231458187,
|
| 1128 |
+
"rewards/rejected": -0.0012119871098548174,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -1.9003782272338867,
|
| 1135 |
+
"logits/rejected": -1.9015012979507446,
|
| 1136 |
+
"logps/chosen": -34.15888214111328,
|
| 1137 |
+
"logps/rejected": -33.193199157714844,
|
| 1138 |
+
"loss": 0.4944,
|
| 1139 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1140 |
+
"rewards/chosen": 0.022765014320611954,
|
| 1141 |
+
"rewards/margins": 0.023936782032251358,
|
| 1142 |
+
"rewards/rejected": -0.0011717682937160134,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -1.9175533056259155,
|
| 1149 |
+
"logits/rejected": -1.9151859283447266,
|
| 1150 |
+
"logps/chosen": -30.27239990234375,
|
| 1151 |
+
"logps/rejected": -33.10813522338867,
|
| 1152 |
+
"loss": 0.4944,
|
| 1153 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1154 |
+
"rewards/chosen": 0.02290928363800049,
|
| 1155 |
+
"rewards/margins": 0.024021681398153305,
|
| 1156 |
+
"rewards/rejected": -0.0011123981093987823,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -1.8999090194702148,
|
| 1163 |
+
"logits/rejected": -1.9051405191421509,
|
| 1164 |
+
"logps/chosen": -29.600543975830078,
|
| 1165 |
+
"logps/rejected": -33.40522384643555,
|
| 1166 |
+
"loss": 0.4944,
|
| 1167 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1168 |
+
"rewards/chosen": 0.021851813420653343,
|
| 1169 |
+
"rewards/margins": 0.02435588836669922,
|
| 1170 |
+
"rewards/rejected": -0.0025040716864168644,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.8462251424789429,
|
| 1177 |
+
"logits/rejected": -1.8450710773468018,
|
| 1178 |
+
"logps/chosen": -30.871395111083984,
|
| 1179 |
+
"logps/rejected": -34.77960205078125,
|
| 1180 |
+
"loss": 0.4947,
|
| 1181 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1182 |
+
"rewards/chosen": 0.022535843774676323,
|
| 1183 |
+
"rewards/margins": 0.02347227931022644,
|
| 1184 |
+
"rewards/rejected": -0.0009364362922497094,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.720664381980896,
|
| 1191 |
+
"logits/rejected": -1.7157522439956665,
|
| 1192 |
+
"logps/chosen": -33.3805046081543,
|
| 1193 |
+
"logps/rejected": -35.51945877075195,
|
| 1194 |
+
"loss": 0.4932,
|
| 1195 |
+
"rewards/accuracies": 0.875,
|
| 1196 |
+
"rewards/chosen": 0.026179391890764236,
|
| 1197 |
+
"rewards/margins": 0.02928983047604561,
|
| 1198 |
+
"rewards/rejected": -0.003110440680757165,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.8583590984344482,
|
| 1205 |
+
"logits/rejected": -1.8622472286224365,
|
| 1206 |
+
"logps/chosen": -33.08837127685547,
|
| 1207 |
+
"logps/rejected": -33.761695861816406,
|
| 1208 |
+
"loss": 0.4948,
|
| 1209 |
+
"rewards/accuracies": 0.8541666269302368,
|
| 1210 |
+
"rewards/chosen": 0.021252866834402084,
|
| 1211 |
+
"rewards/margins": 0.022921394556760788,
|
| 1212 |
+
"rewards/rejected": -0.0016685245791450143,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.803155541419983,
|
| 1219 |
+
"logits/rejected": -1.8014135360717773,
|
| 1220 |
+
"logps/chosen": -30.931873321533203,
|
| 1221 |
+
"logps/rejected": -35.34140396118164,
|
| 1222 |
+
"loss": 0.4914,
|
| 1223 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1224 |
+
"rewards/chosen": 0.029697230085730553,
|
| 1225 |
+
"rewards/margins": 0.039064671844244,
|
| 1226 |
+
"rewards/rejected": -0.00936744175851345,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.8476654291152954,
|
| 1233 |
+
"logits/rejected": -1.8463468551635742,
|
| 1234 |
+
"logps/chosen": -30.539403915405273,
|
| 1235 |
+
"logps/rejected": -33.56371307373047,
|
| 1236 |
+
"loss": 0.4925,
|
| 1237 |
+
"rewards/accuracies": 0.9375,
|
| 1238 |
+
"rewards/chosen": 0.02836332656443119,
|
| 1239 |
+
"rewards/margins": 0.033134736120700836,
|
| 1240 |
+
"rewards/rejected": -0.004771408159285784,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.7794796228408813,
|
| 1247 |
+
"logits/rejected": -1.773779273033142,
|
| 1248 |
+
"logps/chosen": -31.581567764282227,
|
| 1249 |
+
"logps/rejected": -31.875568389892578,
|
| 1250 |
+
"loss": 0.4926,
|
| 1251 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1252 |
+
"rewards/chosen": 0.028015125542879105,
|
| 1253 |
+
"rewards/margins": 0.03212819993495941,
|
| 1254 |
+
"rewards/rejected": -0.004113074392080307,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.132253885269165,
|
| 1260 |
+
"eval_logits/rejected": -2.1275634765625,
|
| 1261 |
+
"eval_logps/chosen": -33.6678466796875,
|
| 1262 |
+
"eval_logps/rejected": -37.38642120361328,
|
| 1263 |
+
"eval_loss": 0.499472975730896,
|
| 1264 |
+
"eval_rewards/accuracies": 0.594684362411499,
|
| 1265 |
+
"eval_rewards/chosen": 0.0036670216359198093,
|
| 1266 |
+
"eval_rewards/margins": 0.002365043619647622,
|
| 1267 |
+
"eval_rewards/rejected": 0.0013019782491028309,
|
| 1268 |
+
"eval_runtime": 145.9155,
|
| 1269 |
+
"eval_samples_per_second": 2.351,
|
| 1270 |
+
"eval_steps_per_second": 0.295,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.1,
|
| 1275 |
+
"learning_rate": 2.7094204786572254e-06,
|
| 1276 |
+
"logits/chosen": -1.875432014465332,
|
| 1277 |
+
"logits/rejected": -1.8828538656234741,
|
| 1278 |
+
"logps/chosen": -29.421600341796875,
|
| 1279 |
+
"logps/rejected": -34.89478302001953,
|
| 1280 |
+
"loss": 0.4926,
|
| 1281 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1282 |
+
"rewards/chosen": 0.027066577225923538,
|
| 1283 |
+
"rewards/margins": 0.033455990254879,
|
| 1284 |
+
"rewards/rejected": -0.006389413960278034,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.8495336771011353,
|
| 1291 |
+
"logits/rejected": -1.852142572402954,
|
| 1292 |
+
"logps/chosen": -30.283716201782227,
|
| 1293 |
+
"logps/rejected": -33.47504425048828,
|
| 1294 |
+
"loss": 0.4942,
|
| 1295 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1296 |
+
"rewards/chosen": 0.022873621433973312,
|
| 1297 |
+
"rewards/margins": 0.02532605454325676,
|
| 1298 |
+
"rewards/rejected": -0.00245243264362216,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.748708963394165,
|
| 1305 |
+
"logits/rejected": -1.7528718709945679,
|
| 1306 |
+
"logps/chosen": -29.96748924255371,
|
| 1307 |
+
"logps/rejected": -35.57471466064453,
|
| 1308 |
+
"loss": 0.4927,
|
| 1309 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1310 |
+
"rewards/chosen": 0.026239663362503052,
|
| 1311 |
+
"rewards/margins": 0.032885000109672546,
|
| 1312 |
+
"rewards/rejected": -0.0066453381441533566,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.8163249492645264,
|
| 1319 |
+
"logits/rejected": -1.8123382329940796,
|
| 1320 |
+
"logps/chosen": -30.351696014404297,
|
| 1321 |
+
"logps/rejected": -33.14829635620117,
|
| 1322 |
+
"loss": 0.4935,
|
| 1323 |
+
"rewards/accuracies": 0.875,
|
| 1324 |
+
"rewards/chosen": 0.026169534772634506,
|
| 1325 |
+
"rewards/margins": 0.027706000953912735,
|
| 1326 |
+
"rewards/rejected": -0.0015364640858024359,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -1.8956245183944702,
|
| 1333 |
+
"logits/rejected": -1.8976259231567383,
|
| 1334 |
+
"logps/chosen": -29.960979461669922,
|
| 1335 |
+
"logps/rejected": -36.380279541015625,
|
| 1336 |
+
"loss": 0.4921,
|
| 1337 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1338 |
+
"rewards/chosen": 0.02814355492591858,
|
| 1339 |
+
"rewards/margins": 0.03532697632908821,
|
| 1340 |
+
"rewards/rejected": -0.007183422800153494,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.7016551494598389,
|
| 1347 |
+
"logits/rejected": -1.6961944103240967,
|
| 1348 |
+
"logps/chosen": -33.43014144897461,
|
| 1349 |
+
"logps/rejected": -33.36838150024414,
|
| 1350 |
+
"loss": 0.4925,
|
| 1351 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1352 |
+
"rewards/chosen": 0.028515327721834183,
|
| 1353 |
+
"rewards/margins": 0.03328012302517891,
|
| 1354 |
+
"rewards/rejected": -0.004764794372022152,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.7582767009735107,
|
| 1361 |
+
"logits/rejected": -1.7583192586898804,
|
| 1362 |
+
"logps/chosen": -32.9239616394043,
|
| 1363 |
+
"logps/rejected": -35.554019927978516,
|
| 1364 |
+
"loss": 0.4919,
|
| 1365 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1366 |
+
"rewards/chosen": 0.029188454151153564,
|
| 1367 |
+
"rewards/margins": 0.036265987902879715,
|
| 1368 |
+
"rewards/rejected": -0.007077532354742289,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.8046823740005493,
|
| 1375 |
+
"logits/rejected": -1.7934318780899048,
|
| 1376 |
+
"logps/chosen": -31.343753814697266,
|
| 1377 |
+
"logps/rejected": -33.549922943115234,
|
| 1378 |
+
"loss": 0.4925,
|
| 1379 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1380 |
+
"rewards/chosen": 0.027628283947706223,
|
| 1381 |
+
"rewards/margins": 0.033907197415828705,
|
| 1382 |
+
"rewards/rejected": -0.006278916262090206,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -1.9076982736587524,
|
| 1389 |
+
"logits/rejected": -1.9121280908584595,
|
| 1390 |
+
"logps/chosen": -30.531978607177734,
|
| 1391 |
+
"logps/rejected": -33.27050018310547,
|
| 1392 |
+
"loss": 0.492,
|
| 1393 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1394 |
+
"rewards/chosen": 0.027979601174592972,
|
| 1395 |
+
"rewards/margins": 0.0363197848200798,
|
| 1396 |
+
"rewards/rejected": -0.008340183645486832,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.826246976852417,
|
| 1403 |
+
"logits/rejected": -1.8192167282104492,
|
| 1404 |
+
"logps/chosen": -31.618709564208984,
|
| 1405 |
+
"logps/rejected": -33.17729949951172,
|
| 1406 |
+
"loss": 0.491,
|
| 1407 |
+
"rewards/accuracies": 0.9375,
|
| 1408 |
+
"rewards/chosen": 0.032353680580854416,
|
| 1409 |
+
"rewards/margins": 0.04010532423853874,
|
| 1410 |
+
"rewards/rejected": -0.007751642260700464,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.1153886318206787,
|
| 1416 |
+
"eval_logits/rejected": -2.1107325553894043,
|
| 1417 |
+
"eval_logps/chosen": -33.72140884399414,
|
| 1418 |
+
"eval_logps/rejected": -37.456180572509766,
|
| 1419 |
+
"eval_loss": 0.4994627833366394,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5739202499389648,
|
| 1421 |
+
"eval_rewards/chosen": 0.003131402190774679,
|
| 1422 |
+
"eval_rewards/margins": 0.0025269899051636457,
|
| 1423 |
+
"eval_rewards/rejected": 0.0006044124602340162,
|
| 1424 |
+
"eval_runtime": 145.8928,
|
| 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": -1.8176963329315186,
|
| 1433 |
+
"logits/rejected": -1.8180383443832397,
|
| 1434 |
+
"logps/chosen": -28.452808380126953,
|
| 1435 |
+
"logps/rejected": -35.59280014038086,
|
| 1436 |
+
"loss": 0.4921,
|
| 1437 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1438 |
+
"rewards/chosen": 0.027143558487296104,
|
| 1439 |
+
"rewards/margins": 0.036174654960632324,
|
| 1440 |
+
"rewards/rejected": -0.00903109647333622,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.772467017173767,
|
| 1447 |
+
"logits/rejected": -1.764556646347046,
|
| 1448 |
+
"logps/chosen": -29.524948120117188,
|
| 1449 |
+
"logps/rejected": -34.33641815185547,
|
| 1450 |
+
"loss": 0.4934,
|
| 1451 |
+
"rewards/accuracies": 0.9375,
|
| 1452 |
+
"rewards/chosen": 0.023139983415603638,
|
| 1453 |
+
"rewards/margins": 0.030790451914072037,
|
| 1454 |
+
"rewards/rejected": -0.007650467567145824,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -1.9374393224716187,
|
| 1461 |
+
"logits/rejected": -1.9376726150512695,
|
| 1462 |
+
"logps/chosen": -27.288341522216797,
|
| 1463 |
+
"logps/rejected": -33.29496383666992,
|
| 1464 |
+
"loss": 0.493,
|
| 1465 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1466 |
+
"rewards/chosen": 0.027086907997727394,
|
| 1467 |
+
"rewards/margins": 0.03138797730207443,
|
| 1468 |
+
"rewards/rejected": -0.0043010651133954525,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.8147531747817993,
|
| 1475 |
+
"logits/rejected": -1.8243385553359985,
|
| 1476 |
+
"logps/chosen": -31.1882266998291,
|
| 1477 |
+
"logps/rejected": -33.76173400878906,
|
| 1478 |
+
"loss": 0.491,
|
| 1479 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1480 |
+
"rewards/chosen": 0.03236837312579155,
|
| 1481 |
+
"rewards/margins": 0.040089257061481476,
|
| 1482 |
+
"rewards/rejected": -0.007720877416431904,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -1.8579801321029663,
|
| 1489 |
+
"logits/rejected": -1.861673355102539,
|
| 1490 |
+
"logps/chosen": -28.247013092041016,
|
| 1491 |
+
"logps/rejected": -32.32887268066406,
|
| 1492 |
+
"loss": 0.4933,
|
| 1493 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1494 |
+
"rewards/chosen": 0.024007126688957214,
|
| 1495 |
+
"rewards/margins": 0.02993503212928772,
|
| 1496 |
+
"rewards/rejected": -0.0059279040433466434,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -1.8509804010391235,
|
| 1503 |
+
"logits/rejected": -1.8471883535385132,
|
| 1504 |
+
"logps/chosen": -31.077259063720703,
|
| 1505 |
+
"logps/rejected": -34.63513946533203,
|
| 1506 |
+
"loss": 0.4918,
|
| 1507 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1508 |
+
"rewards/chosen": 0.029979607090353966,
|
| 1509 |
+
"rewards/margins": 0.037267137318849564,
|
| 1510 |
+
"rewards/rejected": -0.0072875297628343105,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.7676801681518555,
|
| 1517 |
+
"logits/rejected": -1.7701276540756226,
|
| 1518 |
+
"logps/chosen": -29.73238754272461,
|
| 1519 |
+
"logps/rejected": -35.80852127075195,
|
| 1520 |
+
"loss": 0.4919,
|
| 1521 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1522 |
+
"rewards/chosen": 0.030803611502051353,
|
| 1523 |
+
"rewards/margins": 0.03707288205623627,
|
| 1524 |
+
"rewards/rejected": -0.0062692672945559025,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.8003818988800049,
|
| 1531 |
+
"logits/rejected": -1.7940855026245117,
|
| 1532 |
+
"logps/chosen": -32.70087814331055,
|
| 1533 |
+
"logps/rejected": -37.19306945800781,
|
| 1534 |
+
"loss": 0.4921,
|
| 1535 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1536 |
+
"rewards/chosen": 0.027041900902986526,
|
| 1537 |
+
"rewards/margins": 0.03635306656360626,
|
| 1538 |
+
"rewards/rejected": -0.009311167523264885,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.776210069656372,
|
| 1545 |
+
"logits/rejected": -1.7711480855941772,
|
| 1546 |
+
"logps/chosen": -28.2900390625,
|
| 1547 |
+
"logps/rejected": -36.815948486328125,
|
| 1548 |
+
"loss": 0.4922,
|
| 1549 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1550 |
+
"rewards/chosen": 0.02712910808622837,
|
| 1551 |
+
"rewards/margins": 0.03538975864648819,
|
| 1552 |
+
"rewards/rejected": -0.008260652422904968,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.7549470663070679,
|
| 1559 |
+
"logits/rejected": -1.7445472478866577,
|
| 1560 |
+
"logps/chosen": -29.535968780517578,
|
| 1561 |
+
"logps/rejected": -30.032791137695312,
|
| 1562 |
+
"loss": 0.4941,
|
| 1563 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1564 |
+
"rewards/chosen": 0.022127199918031693,
|
| 1565 |
+
"rewards/margins": 0.026457766070961952,
|
| 1566 |
+
"rewards/rejected": -0.004330565221607685,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.1005983352661133,
|
| 1572 |
+
"eval_logits/rejected": -2.095959424972534,
|
| 1573 |
+
"eval_logps/chosen": -33.75029373168945,
|
| 1574 |
+
"eval_logps/rejected": -37.502525329589844,
|
| 1575 |
+
"eval_loss": 0.49945273995399475,
|
| 1576 |
+
"eval_rewards/accuracies": 0.6058970093727112,
|
| 1577 |
+
"eval_rewards/chosen": 0.0028425909113138914,
|
| 1578 |
+
"eval_rewards/margins": 0.002701645949855447,
|
| 1579 |
+
"eval_rewards/rejected": 0.00014094497601035982,
|
| 1580 |
+
"eval_runtime": 145.8673,
|
| 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": -1.8609699010849,
|
| 1589 |
+
"logits/rejected": -1.8604313135147095,
|
| 1590 |
+
"logps/chosen": -29.16253662109375,
|
| 1591 |
+
"logps/rejected": -30.912038803100586,
|
| 1592 |
+
"loss": 0.4927,
|
| 1593 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1594 |
+
"rewards/chosen": 0.026616577059030533,
|
| 1595 |
+
"rewards/margins": 0.03315214067697525,
|
| 1596 |
+
"rewards/rejected": -0.006535563617944717,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -1.9033771753311157,
|
| 1603 |
+
"logits/rejected": -1.9041751623153687,
|
| 1604 |
+
"logps/chosen": -29.415115356445312,
|
| 1605 |
+
"logps/rejected": -31.9776611328125,
|
| 1606 |
+
"loss": 0.4927,
|
| 1607 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1608 |
+
"rewards/chosen": 0.025284653529524803,
|
| 1609 |
+
"rewards/margins": 0.033403180539608,
|
| 1610 |
+
"rewards/rejected": -0.008118532598018646,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -1.870643973350525,
|
| 1617 |
+
"logits/rejected": -1.8740373849868774,
|
| 1618 |
+
"logps/chosen": -29.589406967163086,
|
| 1619 |
+
"logps/rejected": -28.370319366455078,
|
| 1620 |
+
"loss": 0.4929,
|
| 1621 |
+
"rewards/accuracies": 0.875,
|
| 1622 |
+
"rewards/chosen": 0.02690659835934639,
|
| 1623 |
+
"rewards/margins": 0.030772637575864792,
|
| 1624 |
+
"rewards/rejected": -0.003866040613502264,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -1.8781054019927979,
|
| 1631 |
+
"logits/rejected": -1.8704578876495361,
|
| 1632 |
+
"logps/chosen": -28.706920623779297,
|
| 1633 |
+
"logps/rejected": -34.413326263427734,
|
| 1634 |
+
"loss": 0.4928,
|
| 1635 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1636 |
+
"rewards/chosen": 0.02470974065363407,
|
| 1637 |
+
"rewards/margins": 0.03384070843458176,
|
| 1638 |
+
"rewards/rejected": -0.009130971506237984,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.818250298500061,
|
| 1645 |
+
"logits/rejected": -1.8290458917617798,
|
| 1646 |
+
"logps/chosen": -30.612024307250977,
|
| 1647 |
+
"logps/rejected": -32.82752227783203,
|
| 1648 |
+
"loss": 0.492,
|
| 1649 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1650 |
+
"rewards/chosen": 0.02941080555319786,
|
| 1651 |
+
"rewards/margins": 0.035235948860645294,
|
| 1652 |
+
"rewards/rejected": -0.005825136322528124,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -1.8953624963760376,
|
| 1659 |
+
"logits/rejected": -1.8862526416778564,
|
| 1660 |
+
"logps/chosen": -30.589797973632812,
|
| 1661 |
+
"logps/rejected": -37.13585662841797,
|
| 1662 |
+
"loss": 0.4921,
|
| 1663 |
+
"rewards/accuracies": 0.875,
|
| 1664 |
+
"rewards/chosen": 0.029109802097082138,
|
| 1665 |
+
"rewards/margins": 0.03715446963906288,
|
| 1666 |
+
"rewards/rejected": -0.008044666610658169,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.8016560077667236,
|
| 1673 |
+
"logits/rejected": -1.8037437200546265,
|
| 1674 |
+
"logps/chosen": -32.77046203613281,
|
| 1675 |
+
"logps/rejected": -33.44564437866211,
|
| 1676 |
+
"loss": 0.4911,
|
| 1677 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1678 |
+
"rewards/chosen": 0.03218822926282883,
|
| 1679 |
+
"rewards/margins": 0.039665985852479935,
|
| 1680 |
+
"rewards/rejected": -0.007477754261344671,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -1.9271634817123413,
|
| 1687 |
+
"logits/rejected": -1.9280281066894531,
|
| 1688 |
+
"logps/chosen": -30.242328643798828,
|
| 1689 |
+
"logps/rejected": -35.20170211791992,
|
| 1690 |
+
"loss": 0.4917,
|
| 1691 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1692 |
+
"rewards/chosen": 0.028153110295534134,
|
| 1693 |
+
"rewards/margins": 0.03901328891515732,
|
| 1694 |
+
"rewards/rejected": -0.010860181413590908,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -1.9012151956558228,
|
| 1701 |
+
"logits/rejected": -1.9054481983184814,
|
| 1702 |
+
"logps/chosen": -28.445846557617188,
|
| 1703 |
+
"logps/rejected": -34.31275177001953,
|
| 1704 |
+
"loss": 0.493,
|
| 1705 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1706 |
+
"rewards/chosen": 0.02447468414902687,
|
| 1707 |
+
"rewards/margins": 0.03212384134531021,
|
| 1708 |
+
"rewards/rejected": -0.0076491571962833405,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -1.8907396793365479,
|
| 1715 |
+
"logits/rejected": -1.895939588546753,
|
| 1716 |
+
"logps/chosen": -31.212976455688477,
|
| 1717 |
+
"logps/rejected": -30.585351943969727,
|
| 1718 |
+
"loss": 0.4943,
|
| 1719 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1720 |
+
"rewards/chosen": 0.0235602967441082,
|
| 1721 |
+
"rewards/margins": 0.02539706788957119,
|
| 1722 |
+
"rewards/rejected": -0.001836769632063806,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.0964860916137695,
|
| 1728 |
+
"eval_logits/rejected": -2.091848373413086,
|
| 1729 |
+
"eval_logps/chosen": -33.77998352050781,
|
| 1730 |
+
"eval_logps/rejected": -37.53815841674805,
|
| 1731 |
+
"eval_loss": 0.4994538724422455,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5826411843299866,
|
| 1733 |
+
"eval_rewards/chosen": 0.0025456694420427084,
|
| 1734 |
+
"eval_rewards/margins": 0.0027610675897449255,
|
| 1735 |
+
"eval_rewards/rejected": -0.0002153978421119973,
|
| 1736 |
+
"eval_runtime": 145.7726,
|
| 1737 |
+
"eval_samples_per_second": 2.353,
|
| 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": -1.8822052478790283,
|
| 1745 |
+
"logits/rejected": -1.8824293613433838,
|
| 1746 |
+
"logps/chosen": -30.01492691040039,
|
| 1747 |
+
"logps/rejected": -31.45803451538086,
|
| 1748 |
+
"loss": 0.4923,
|
| 1749 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1750 |
+
"rewards/chosen": 0.029252270236611366,
|
| 1751 |
+
"rewards/margins": 0.03367346525192261,
|
| 1752 |
+
"rewards/rejected": -0.004421197809278965,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.7276147603988647,
|
| 1759 |
+
"logits/rejected": -1.7294237613677979,
|
| 1760 |
+
"logps/chosen": -31.513330459594727,
|
| 1761 |
+
"logps/rejected": -33.66464614868164,
|
| 1762 |
+
"loss": 0.4919,
|
| 1763 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1764 |
+
"rewards/chosen": 0.030516481027007103,
|
| 1765 |
+
"rewards/margins": 0.03675025328993797,
|
| 1766 |
+
"rewards/rejected": -0.006233775056898594,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.7628084421157837,
|
| 1773 |
+
"logits/rejected": -1.760263442993164,
|
| 1774 |
+
"logps/chosen": -28.515111923217773,
|
| 1775 |
+
"logps/rejected": -31.614593505859375,
|
| 1776 |
+
"loss": 0.4935,
|
| 1777 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1778 |
+
"rewards/chosen": 0.024966087192296982,
|
| 1779 |
+
"rewards/margins": 0.02945462428033352,
|
| 1780 |
+
"rewards/rejected": -0.00448854174464941,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -1.8981006145477295,
|
| 1787 |
+
"logits/rejected": -1.8926544189453125,
|
| 1788 |
+
"logps/chosen": -30.10824966430664,
|
| 1789 |
+
"logps/rejected": -33.54905319213867,
|
| 1790 |
+
"loss": 0.4922,
|
| 1791 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1792 |
+
"rewards/chosen": 0.02931547723710537,
|
| 1793 |
+
"rewards/margins": 0.035265903919935226,
|
| 1794 |
+
"rewards/rejected": -0.005950425285845995,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.759209394454956,
|
| 1801 |
+
"logits/rejected": -1.7495086193084717,
|
| 1802 |
+
"logps/chosen": -30.550457000732422,
|
| 1803 |
+
"logps/rejected": -31.75992202758789,
|
| 1804 |
+
"loss": 0.4925,
|
| 1805 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1806 |
+
"rewards/chosen": 0.02901277504861355,
|
| 1807 |
+
"rewards/margins": 0.03368794918060303,
|
| 1808 |
+
"rewards/rejected": -0.004675174597650766,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.8278675079345703,
|
| 1815 |
+
"logits/rejected": -1.8278887271881104,
|
| 1816 |
+
"logps/chosen": -30.80059814453125,
|
| 1817 |
+
"logps/rejected": -31.803054809570312,
|
| 1818 |
+
"loss": 0.4921,
|
| 1819 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1820 |
+
"rewards/chosen": 0.028811518102884293,
|
| 1821 |
+
"rewards/margins": 0.0348275825381279,
|
| 1822 |
+
"rewards/rejected": -0.00601606722921133,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -1.902727484703064,
|
| 1829 |
+
"logits/rejected": -1.8974792957305908,
|
| 1830 |
+
"logps/chosen": -30.870708465576172,
|
| 1831 |
+
"logps/rejected": -33.48421859741211,
|
| 1832 |
+
"loss": 0.4928,
|
| 1833 |
+
"rewards/accuracies": 0.9375,
|
| 1834 |
+
"rewards/chosen": 0.026552587747573853,
|
| 1835 |
+
"rewards/margins": 0.03198261931538582,
|
| 1836 |
+
"rewards/rejected": -0.005430030170828104,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.7634168863296509,
|
| 1843 |
+
"logits/rejected": -1.7711273431777954,
|
| 1844 |
+
"logps/chosen": -30.915302276611328,
|
| 1845 |
+
"logps/rejected": -33.02967071533203,
|
| 1846 |
+
"loss": 0.4932,
|
| 1847 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1848 |
+
"rewards/chosen": 0.024176979437470436,
|
| 1849 |
+
"rewards/margins": 0.030859291553497314,
|
| 1850 |
+
"rewards/rejected": -0.006682313047349453,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -1.8290866613388062,
|
| 1857 |
+
"logits/rejected": -1.8232570886611938,
|
| 1858 |
+
"logps/chosen": -29.006099700927734,
|
| 1859 |
+
"logps/rejected": -34.01789093017578,
|
| 1860 |
+
"loss": 0.4914,
|
| 1861 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1862 |
+
"rewards/chosen": 0.03167518228292465,
|
| 1863 |
+
"rewards/margins": 0.03901957720518112,
|
| 1864 |
+
"rewards/rejected": -0.007344390265643597,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.7752304077148438,
|
| 1871 |
+
"logits/rejected": -1.7772098779678345,
|
| 1872 |
+
"logps/chosen": -30.416156768798828,
|
| 1873 |
+
"logps/rejected": -33.775291442871094,
|
| 1874 |
+
"loss": 0.4911,
|
| 1875 |
+
"rewards/accuracies": 0.9375,
|
| 1876 |
+
"rewards/chosen": 0.03074287250638008,
|
| 1877 |
+
"rewards/margins": 0.0408952571451664,
|
| 1878 |
+
"rewards/rejected": -0.010152382776141167,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.0958993434906006,
|
| 1884 |
+
"eval_logits/rejected": -2.0912692546844482,
|
| 1885 |
+
"eval_logps/chosen": -33.775428771972656,
|
| 1886 |
+
"eval_logps/rejected": -37.52570724487305,
|
| 1887 |
+
"eval_loss": 0.4994668662548065,
|
| 1888 |
+
"eval_rewards/accuracies": 0.6009136438369751,
|
| 1889 |
+
"eval_rewards/chosen": 0.0025912297423928976,
|
| 1890 |
+
"eval_rewards/margins": 0.002682073274627328,
|
| 1891 |
+
"eval_rewards/rejected": -9.084340126719326e-05,
|
| 1892 |
+
"eval_runtime": 145.907,
|
| 1893 |
+
"eval_samples_per_second": 2.351,
|
| 1894 |
+
"eval_steps_per_second": 0.295,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 3.14,
|
| 1899 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1900 |
+
"logits/chosen": -1.7069997787475586,
|
| 1901 |
+
"logits/rejected": -1.7061100006103516,
|
| 1902 |
+
"logps/chosen": -32.673728942871094,
|
| 1903 |
+
"logps/rejected": -33.5372314453125,
|
| 1904 |
+
"loss": 0.4906,
|
| 1905 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1906 |
+
"rewards/chosen": 0.0337875559926033,
|
| 1907 |
+
"rewards/margins": 0.042066238820552826,
|
| 1908 |
+
"rewards/rejected": -0.008278685621917248,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -1.8786895275115967,
|
| 1915 |
+
"logits/rejected": -1.8728017807006836,
|
| 1916 |
+
"logps/chosen": -28.626781463623047,
|
| 1917 |
+
"logps/rejected": -33.97628402709961,
|
| 1918 |
+
"loss": 0.491,
|
| 1919 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1920 |
+
"rewards/chosen": 0.030769089236855507,
|
| 1921 |
+
"rewards/margins": 0.04174653813242912,
|
| 1922 |
+
"rewards/rejected": -0.010977448895573616,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.7307627201080322,
|
| 1929 |
+
"logits/rejected": -1.7194693088531494,
|
| 1930 |
+
"logps/chosen": -31.0261173248291,
|
| 1931 |
+
"logps/rejected": -33.56462478637695,
|
| 1932 |
+
"loss": 0.4912,
|
| 1933 |
+
"rewards/accuracies": 0.9375,
|
| 1934 |
+
"rewards/chosen": 0.033595647662878036,
|
| 1935 |
+
"rewards/margins": 0.040303632616996765,
|
| 1936 |
+
"rewards/rejected": -0.006707982625812292,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.8212391138076782,
|
| 1943 |
+
"logits/rejected": -1.8145866394042969,
|
| 1944 |
+
"logps/chosen": -28.291839599609375,
|
| 1945 |
+
"logps/rejected": -29.779611587524414,
|
| 1946 |
+
"loss": 0.4929,
|
| 1947 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1948 |
+
"rewards/chosen": 0.024083953350782394,
|
| 1949 |
+
"rewards/margins": 0.033204466104507446,
|
| 1950 |
+
"rewards/rejected": -0.009120511822402477,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -1.8461530208587646,
|
| 1957 |
+
"logits/rejected": -1.853011131286621,
|
| 1958 |
+
"logps/chosen": -32.12201690673828,
|
| 1959 |
+
"logps/rejected": -32.083274841308594,
|
| 1960 |
+
"loss": 0.4913,
|
| 1961 |
+
"rewards/accuracies": 0.9375,
|
| 1962 |
+
"rewards/chosen": 0.032733749598264694,
|
| 1963 |
+
"rewards/margins": 0.0383833646774292,
|
| 1964 |
+
"rewards/rejected": -0.005649610888212919,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.8253049850463867,
|
| 1971 |
+
"logits/rejected": -1.8258377313613892,
|
| 1972 |
+
"logps/chosen": -30.2758846282959,
|
| 1973 |
+
"logps/rejected": -33.74597930908203,
|
| 1974 |
+
"loss": 0.492,
|
| 1975 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1976 |
+
"rewards/chosen": 0.027526726946234703,
|
| 1977 |
+
"rewards/margins": 0.03683784604072571,
|
| 1978 |
+
"rewards/rejected": -0.00931111816316843,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.7631717920303345,
|
| 1985 |
+
"logits/rejected": -1.7627891302108765,
|
| 1986 |
+
"logps/chosen": -30.03763771057129,
|
| 1987 |
+
"logps/rejected": -34.385276794433594,
|
| 1988 |
+
"loss": 0.4914,
|
| 1989 |
+
"rewards/accuracies": 0.9375,
|
| 1990 |
+
"rewards/chosen": 0.03053201362490654,
|
| 1991 |
+
"rewards/margins": 0.039178695529699326,
|
| 1992 |
+
"rewards/rejected": -0.008646685630083084,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.648390531539917,
|
| 1999 |
+
"logits/rejected": -1.6438019275665283,
|
| 2000 |
+
"logps/chosen": -32.16303253173828,
|
| 2001 |
+
"logps/rejected": -31.598995208740234,
|
| 2002 |
+
"loss": 0.4915,
|
| 2003 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2004 |
+
"rewards/chosen": 0.0307441595941782,
|
| 2005 |
+
"rewards/margins": 0.03746918961405754,
|
| 2006 |
+
"rewards/rejected": -0.00672503188252449,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -1.8533127307891846,
|
| 2013 |
+
"logits/rejected": -1.838148832321167,
|
| 2014 |
+
"logps/chosen": -31.7786865234375,
|
| 2015 |
+
"logps/rejected": -35.894744873046875,
|
| 2016 |
+
"loss": 0.4922,
|
| 2017 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2018 |
+
"rewards/chosen": 0.025688907131552696,
|
| 2019 |
+
"rewards/margins": 0.03698316216468811,
|
| 2020 |
+
"rewards/rejected": -0.011294253170490265,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.7974460124969482,
|
| 2027 |
+
"logits/rejected": -1.802198052406311,
|
| 2028 |
+
"logps/chosen": -30.211551666259766,
|
| 2029 |
+
"logps/rejected": -37.193172454833984,
|
| 2030 |
+
"loss": 0.4906,
|
| 2031 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2032 |
+
"rewards/chosen": 0.0330607071518898,
|
| 2033 |
+
"rewards/margins": 0.04396023601293564,
|
| 2034 |
+
"rewards/rejected": -0.010899528861045837,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.0954809188842773,
|
| 2040 |
+
"eval_logits/rejected": -2.0908546447753906,
|
| 2041 |
+
"eval_logps/chosen": -33.78174591064453,
|
| 2042 |
+
"eval_logps/rejected": -37.53657150268555,
|
| 2043 |
+
"eval_loss": 0.49945539236068726,
|
| 2044 |
+
"eval_rewards/accuracies": 0.6034052968025208,
|
| 2045 |
+
"eval_rewards/chosen": 0.0025280837435275316,
|
| 2046 |
+
"eval_rewards/margins": 0.002727580489590764,
|
| 2047 |
+
"eval_rewards/rejected": -0.0001994967897189781,
|
| 2048 |
+
"eval_runtime": 145.8892,
|
| 2049 |
+
"eval_samples_per_second": 2.351,
|
| 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": -1.770353078842163,
|
| 2057 |
+
"logits/rejected": -1.770532250404358,
|
| 2058 |
+
"logps/chosen": -32.75363540649414,
|
| 2059 |
+
"logps/rejected": -37.96880340576172,
|
| 2060 |
+
"loss": 0.4904,
|
| 2061 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2062 |
+
"rewards/chosen": 0.03490384668111801,
|
| 2063 |
+
"rewards/margins": 0.04375072568655014,
|
| 2064 |
+
"rewards/rejected": -0.008846879936754704,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -1.8576828241348267,
|
| 2071 |
+
"logits/rejected": -1.8563992977142334,
|
| 2072 |
+
"logps/chosen": -29.168567657470703,
|
| 2073 |
+
"logps/rejected": -31.714990615844727,
|
| 2074 |
+
"loss": 0.492,
|
| 2075 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2076 |
+
"rewards/chosen": 0.02925288677215576,
|
| 2077 |
+
"rewards/margins": 0.036869172006845474,
|
| 2078 |
+
"rewards/rejected": -0.007616284303367138,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.766406774520874,
|
| 2085 |
+
"logits/rejected": -1.7676417827606201,
|
| 2086 |
+
"logps/chosen": -31.04288673400879,
|
| 2087 |
+
"logps/rejected": -34.331932067871094,
|
| 2088 |
+
"loss": 0.491,
|
| 2089 |
+
"rewards/accuracies": 0.9375,
|
| 2090 |
+
"rewards/chosen": 0.03331366181373596,
|
| 2091 |
+
"rewards/margins": 0.04014625400304794,
|
| 2092 |
+
"rewards/rejected": -0.006832593586295843,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.8259414434432983,
|
| 2099 |
+
"logits/rejected": -1.8247743844985962,
|
| 2100 |
+
"logps/chosen": -28.54180335998535,
|
| 2101 |
+
"logps/rejected": -32.83352279663086,
|
| 2102 |
+
"loss": 0.4923,
|
| 2103 |
+
"rewards/accuracies": 0.9375,
|
| 2104 |
+
"rewards/chosen": 0.026568632572889328,
|
| 2105 |
+
"rewards/margins": 0.03498411923646927,
|
| 2106 |
+
"rewards/rejected": -0.008415484800934792,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -1.8422014713287354,
|
| 2113 |
+
"logits/rejected": -1.8278350830078125,
|
| 2114 |
+
"logps/chosen": -31.403839111328125,
|
| 2115 |
+
"logps/rejected": -37.33091354370117,
|
| 2116 |
+
"loss": 0.4914,
|
| 2117 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2118 |
+
"rewards/chosen": 0.02912832237780094,
|
| 2119 |
+
"rewards/margins": 0.039793141186237335,
|
| 2120 |
+
"rewards/rejected": -0.010664817877113819,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -1.834238052368164,
|
| 2127 |
+
"logits/rejected": -1.8465309143066406,
|
| 2128 |
+
"logps/chosen": -29.14828872680664,
|
| 2129 |
+
"logps/rejected": -32.92185592651367,
|
| 2130 |
+
"loss": 0.492,
|
| 2131 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2132 |
+
"rewards/chosen": 0.029201796278357506,
|
| 2133 |
+
"rewards/margins": 0.03602294623851776,
|
| 2134 |
+
"rewards/rejected": -0.006821149028837681,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -1.937150239944458,
|
| 2141 |
+
"logits/rejected": -1.9364910125732422,
|
| 2142 |
+
"logps/chosen": -28.896930694580078,
|
| 2143 |
+
"logps/rejected": -37.03318786621094,
|
| 2144 |
+
"loss": 0.4907,
|
| 2145 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2146 |
+
"rewards/chosen": 0.03133498877286911,
|
| 2147 |
+
"rewards/margins": 0.043358031660318375,
|
| 2148 |
+
"rewards/rejected": -0.012023041024804115,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.7997983694076538,
|
| 2155 |
+
"logits/rejected": -1.8073842525482178,
|
| 2156 |
+
"logps/chosen": -30.809356689453125,
|
| 2157 |
+
"logps/rejected": -38.418704986572266,
|
| 2158 |
+
"loss": 0.4906,
|
| 2159 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2160 |
+
"rewards/chosen": 0.03371125087141991,
|
| 2161 |
+
"rewards/margins": 0.04340182989835739,
|
| 2162 |
+
"rewards/rejected": -0.009690576232969761,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.756156325340271,
|
| 2169 |
+
"logits/rejected": -1.7522109746932983,
|
| 2170 |
+
"logps/chosen": -31.52239418029785,
|
| 2171 |
+
"logps/rejected": -32.12908172607422,
|
| 2172 |
+
"loss": 0.4915,
|
| 2173 |
+
"rewards/accuracies": 0.875,
|
| 2174 |
+
"rewards/chosen": 0.0324481800198555,
|
| 2175 |
+
"rewards/margins": 0.03776510804891586,
|
| 2176 |
+
"rewards/rejected": -0.005316923372447491,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -1.9121891260147095,
|
| 2183 |
+
"logits/rejected": -1.90972101688385,
|
| 2184 |
+
"logps/chosen": -28.551815032958984,
|
| 2185 |
+
"logps/rejected": -32.69978713989258,
|
| 2186 |
+
"loss": 0.4928,
|
| 2187 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2188 |
+
"rewards/chosen": 0.027447333559393883,
|
| 2189 |
+
"rewards/margins": 0.03300053998827934,
|
| 2190 |
+
"rewards/rejected": -0.0055532073602080345,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.0949766635894775,
|
| 2196 |
+
"eval_logits/rejected": -2.0903432369232178,
|
| 2197 |
+
"eval_logps/chosen": -33.78425216674805,
|
| 2198 |
+
"eval_logps/rejected": -37.54228591918945,
|
| 2199 |
+
"eval_loss": 0.4994584321975708,
|
| 2200 |
+
"eval_rewards/accuracies": 0.5921927094459534,
|
| 2201 |
+
"eval_rewards/chosen": 0.0025030241813510656,
|
| 2202 |
+
"eval_rewards/margins": 0.0027596852742135525,
|
| 2203 |
+
"eval_rewards/rejected": -0.0002566613839007914,
|
| 2204 |
+
"eval_runtime": 145.7016,
|
| 2205 |
+
"eval_samples_per_second": 2.354,
|
| 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": -1.8566265106201172,
|
| 2213 |
+
"logits/rejected": -1.8526551723480225,
|
| 2214 |
+
"logps/chosen": -30.65496826171875,
|
| 2215 |
+
"logps/rejected": -30.23004722595215,
|
| 2216 |
+
"loss": 0.4918,
|
| 2217 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2218 |
+
"rewards/chosen": 0.029875636100769043,
|
| 2219 |
+
"rewards/margins": 0.036631517112255096,
|
| 2220 |
+
"rewards/rejected": -0.0067558870650827885,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.8138599395751953,
|
| 2227 |
+
"logits/rejected": -1.8149265050888062,
|
| 2228 |
+
"logps/chosen": -30.453847885131836,
|
| 2229 |
+
"logps/rejected": -35.29817581176758,
|
| 2230 |
+
"loss": 0.4926,
|
| 2231 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2232 |
+
"rewards/chosen": 0.029110288247466087,
|
| 2233 |
+
"rewards/margins": 0.0328519269824028,
|
| 2234 |
+
"rewards/rejected": -0.0037416405975818634,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.7643918991088867,
|
| 2241 |
+
"logits/rejected": -1.7578294277191162,
|
| 2242 |
+
"logps/chosen": -31.150136947631836,
|
| 2243 |
+
"logps/rejected": -37.63862228393555,
|
| 2244 |
+
"loss": 0.4914,
|
| 2245 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2246 |
+
"rewards/chosen": 0.0299163106828928,
|
| 2247 |
+
"rewards/margins": 0.039408471435308456,
|
| 2248 |
+
"rewards/rejected": -0.009492164477705956,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.7766668796539307,
|
| 2255 |
+
"logits/rejected": -1.7712846994400024,
|
| 2256 |
+
"logps/chosen": -26.59188461303711,
|
| 2257 |
+
"logps/rejected": -32.42449188232422,
|
| 2258 |
+
"loss": 0.4925,
|
| 2259 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2260 |
+
"rewards/chosen": 0.027213579043745995,
|
| 2261 |
+
"rewards/margins": 0.03415192291140556,
|
| 2262 |
+
"rewards/rejected": -0.006938344333320856,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.8049747943878174,
|
| 2269 |
+
"logits/rejected": -1.7948274612426758,
|
| 2270 |
+
"logps/chosen": -30.14188575744629,
|
| 2271 |
+
"logps/rejected": -36.50958251953125,
|
| 2272 |
+
"loss": 0.4926,
|
| 2273 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2274 |
+
"rewards/chosen": 0.023396430537104607,
|
| 2275 |
+
"rewards/margins": 0.03563863784074783,
|
| 2276 |
+
"rewards/rejected": -0.012242205440998077,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.776219129562378,
|
| 2283 |
+
"logits/rejected": -1.773436188697815,
|
| 2284 |
+
"logps/chosen": -30.742990493774414,
|
| 2285 |
+
"logps/rejected": -31.301162719726562,
|
| 2286 |
+
"loss": 0.4933,
|
| 2287 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2288 |
+
"rewards/chosen": 0.025667887181043625,
|
| 2289 |
+
"rewards/margins": 0.029771283268928528,
|
| 2290 |
+
"rewards/rejected": -0.004103394225239754,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -1.835860013961792,
|
| 2297 |
+
"logits/rejected": -1.8421319723129272,
|
| 2298 |
+
"logps/chosen": -28.727758407592773,
|
| 2299 |
+
"logps/rejected": -33.723846435546875,
|
| 2300 |
+
"loss": 0.4909,
|
| 2301 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2302 |
+
"rewards/chosen": 0.033414192497730255,
|
| 2303 |
+
"rewards/margins": 0.04102843254804611,
|
| 2304 |
+
"rewards/rejected": -0.007614238653331995,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -1.8385893106460571,
|
| 2311 |
+
"logits/rejected": -1.8487203121185303,
|
| 2312 |
+
"logps/chosen": -28.86956787109375,
|
| 2313 |
+
"logps/rejected": -30.35264015197754,
|
| 2314 |
+
"loss": 0.4921,
|
| 2315 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2316 |
+
"rewards/chosen": 0.02825688198208809,
|
| 2317 |
+
"rewards/margins": 0.03549762815237045,
|
| 2318 |
+
"rewards/rejected": -0.007240750826895237,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.7700088024139404,
|
| 2325 |
+
"logits/rejected": -1.7631587982177734,
|
| 2326 |
+
"logps/chosen": -31.02837562561035,
|
| 2327 |
+
"logps/rejected": -33.01637649536133,
|
| 2328 |
+
"loss": 0.4924,
|
| 2329 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2330 |
+
"rewards/chosen": 0.02628110721707344,
|
| 2331 |
+
"rewards/margins": 0.03434007614850998,
|
| 2332 |
+
"rewards/rejected": -0.008058969862759113,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.823885202407837,
|
| 2339 |
+
"logits/rejected": -1.8207124471664429,
|
| 2340 |
+
"logps/chosen": -30.59462547302246,
|
| 2341 |
+
"logps/rejected": -34.283512115478516,
|
| 2342 |
+
"loss": 0.4912,
|
| 2343 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2344 |
+
"rewards/chosen": 0.030309468507766724,
|
| 2345 |
+
"rewards/margins": 0.04054233804345131,
|
| 2346 |
+
"rewards/rejected": -0.010232868604362011,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.0954155921936035,
|
| 2352 |
+
"eval_logits/rejected": -2.090785264968872,
|
| 2353 |
+
"eval_logps/chosen": -33.78346252441406,
|
| 2354 |
+
"eval_logps/rejected": -37.54273223876953,
|
| 2355 |
+
"eval_loss": 0.49945247173309326,
|
| 2356 |
+
"eval_rewards/accuracies": 0.6034052968025208,
|
| 2357 |
+
"eval_rewards/chosen": 0.0025108722038567066,
|
| 2358 |
+
"eval_rewards/margins": 0.0027720185462385416,
|
| 2359 |
+
"eval_rewards/rejected": -0.00026114637148566544,
|
| 2360 |
+
"eval_runtime": 145.8069,
|
| 2361 |
+
"eval_samples_per_second": 2.352,
|
| 2362 |
+
"eval_steps_per_second": 0.295,
|
| 2363 |
+
"step": 1500
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 3.92,
|
| 2367 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2368 |
+
"logits/chosen": -1.900957465171814,
|
| 2369 |
+
"logits/rejected": -1.9016454219818115,
|
| 2370 |
+
"logps/chosen": -26.679061889648438,
|
| 2371 |
+
"logps/rejected": -33.479774475097656,
|
| 2372 |
+
"loss": 0.4914,
|
| 2373 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 2374 |
+
"rewards/chosen": 0.03137214854359627,
|
| 2375 |
+
"rewards/margins": 0.03876541927456856,
|
| 2376 |
+
"rewards/rejected": -0.007393266074359417,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.7551101446151733,
|
| 2383 |
+
"logits/rejected": -1.7550216913223267,
|
| 2384 |
+
"logps/chosen": -30.129684448242188,
|
| 2385 |
+
"logps/rejected": -34.65699005126953,
|
| 2386 |
+
"loss": 0.4918,
|
| 2387 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2388 |
+
"rewards/chosen": 0.029121821746230125,
|
| 2389 |
+
"rewards/margins": 0.036954350769519806,
|
| 2390 |
+
"rewards/rejected": -0.007832523435354233,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -1.9060754776000977,
|
| 2397 |
+
"logits/rejected": -1.9034589529037476,
|
| 2398 |
+
"logps/chosen": -28.90531349182129,
|
| 2399 |
+
"logps/rejected": -32.31726837158203,
|
| 2400 |
+
"loss": 0.4926,
|
| 2401 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2402 |
+
"rewards/chosen": 0.02568744495511055,
|
| 2403 |
+
"rewards/margins": 0.03402363136410713,
|
| 2404 |
+
"rewards/rejected": -0.008336183615028858,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -1.890435814857483,
|
| 2411 |
+
"logits/rejected": -1.8915026187896729,
|
| 2412 |
+
"logps/chosen": -28.199377059936523,
|
| 2413 |
+
"logps/rejected": -30.4346981048584,
|
| 2414 |
+
"loss": 0.4932,
|
| 2415 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 2416 |
+
"rewards/chosen": 0.024205626919865608,
|
| 2417 |
+
"rewards/margins": 0.030650783330202103,
|
| 2418 |
+
"rewards/rejected": -0.006445156875997782,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 0.39746343532165923,
|
| 2426 |
+
"train_runtime": 10799.7737,
|
| 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,
|