Commit
•
04fb078
1
Parent(s):
ec39a9a
Model save
Browse files- README.md +29 -19
- adapter_model.safetensors +1 -1
- all_results.json +5 -5
- train_results.json +5 -5
- trainer_state.json +1906 -90
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
| 4 |
-
- alignment-handbook
|
| 5 |
- trl
|
| 6 |
- dpo
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
| 9 |
-
datasets:
|
| 10 |
-
- hugodk-sch/aftonposten_title_prefs
|
| 11 |
model-index:
|
| 12 |
- name: aftonposten-6b-align-scan
|
| 13 |
results: []
|
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
# aftonposten-6b-align-scan
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 6.
|
| 24 |
-
- Rewards/chosen: 0.
|
| 25 |
-
- Rewards/rejected: -0.
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -34.
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,15 +55,27 @@ The following hyperparameters were used during training:
|
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
| 59 |
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
-
- num_epochs:
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
-
| Training Loss | Epoch | Step |
|
| 65 |
-
|
| 66 |
-
| 5.6745 | 0.26 | 100 |
|
| 67 |
-
| 5.2135 | 0.52 | 200 |
|
| 68 |
-
| 4.3883 | 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: 6.0348
|
| 22 |
+
- Rewards/chosen: -0.0125
|
| 23 |
+
- Rewards/rejected: -0.0385
|
| 24 |
+
- Rewards/accuracies: 0.5835
|
| 25 |
+
- Rewards/margins: 0.0260
|
| 26 |
+
- Logps/rejected: -37.7091
|
| 27 |
+
- Logps/chosen: -34.0972
|
| 28 |
+
- Logits/rejected: -2.2004
|
| 29 |
+
- Logits/chosen: -2.2052
|
| 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 |
+
| 5.6745 | 0.26 | 100 | -2.2338 | -2.2290 | -34.0153 | -37.5181 | 6.2508 | 0.5461 | 0.0038 | 0.0041 | -0.0003 |
|
| 65 |
+
| 5.2135 | 0.52 | 200 | -2.2315 | -2.2267 | -34.0014 | -37.5042 | 6.2881 | 0.5403 | 0.0066 | 0.0041 | 0.0025 |
|
| 66 |
+
| 4.3883 | 0.78 | 300 | -2.2291 | -2.2243 | -34.0193 | -37.5325 | 6.2382 | 0.5166 | 0.0031 | 0.0062 | -0.0032 |
|
| 67 |
+
| 2.9753 | 1.04 | 400 | 6.0369 | 0.0069 | -0.0106 | 0.6034 | 0.0176 | -37.5698 | -33.9999 | -2.2093 | -2.2141 |
|
| 68 |
+
| 2.4163 | 1.3 | 500 | 6.0677 | -0.0149 | -0.0375 | 0.5801 | 0.0225 | -37.7039 | -34.1092 | -2.1858 | -2.1907 |
|
| 69 |
+
| 2.52 | 1.56 | 600 | 5.9990 | -0.0097 | -0.0348 | 0.5748 | 0.0251 | -37.6905 | -34.0832 | -2.1951 | -2.1999 |
|
| 70 |
+
| 2.9186 | 1.82 | 700 | 6.1696 | -0.0176 | -0.0364 | 0.5598 | 0.0188 | -37.6988 | -34.1227 | -2.2048 | -2.2097 |
|
| 71 |
+
| 1.2867 | 2.08 | 800 | 6.0594 | -0.0122 | -0.0361 | 0.5777 | 0.0239 | -37.6970 | -34.0957 | -2.2060 | -2.2109 |
|
| 72 |
+
| 0.8862 | 2.34 | 900 | 6.0621 | -0.0165 | -0.0403 | 0.5918 | 0.0237 | -37.7179 | -34.1172 | -2.2027 | -2.2076 |
|
| 73 |
+
| 1.2395 | 2.6 | 1000 | 6.0000 | -0.0163 | -0.0418 | 0.5864 | 0.0255 | -37.7257 | -34.1161 | -2.2002 | -2.2050 |
|
| 74 |
+
| 1.4312 | 2.86 | 1100 | 5.9905 | -0.0144 | -0.0409 | 0.5860 | 0.0264 | -37.7210 | -34.1067 | -2.1989 | -2.2038 |
|
| 75 |
+
| 1.0133 | 3.12 | 1200 | 6.1103 | -0.0167 | -0.0396 | 0.5889 | 0.0229 | -37.7146 | -34.1182 | -2.2000 | -2.2048 |
|
| 76 |
+
| 0.5152 | 3.38 | 1300 | 6.0578 | -0.0132 | -0.0383 | 0.5544 | 0.0251 | -37.7080 | -34.1004 | -2.2003 | -2.2051 |
|
| 77 |
+
| 0.8378 | 3.64 | 1400 | 6.0572 | -0.0138 | -0.0389 | 0.5748 | 0.0251 | -37.7113 | -34.1035 | -2.2004 | -2.2052 |
|
| 78 |
+
| 0.9599 | 3.9 | 1500 | 6.0348 | -0.0125 | -0.0385 | 0.5835 | 0.0260 | -37.7091 | -34.0972 | -2.2004 | -2.2052 |
|
| 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 sha256:5fe0989cf18557a7daa207439d8382e361ceff43a9e66402269ceda7d71f2a3e
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.22882342338562,
|
| 4 |
"eval_logits/rejected": -2.223994731903076,
|
| 5 |
"eval_logps/chosen": -34.021240234375,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.358,
|
| 15 |
"eval_steps_per_second": 0.296,
|
| 16 |
-
"train_loss":
|
| 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": -2.22882342338562,
|
| 4 |
"eval_logits/rejected": -2.223994731903076,
|
| 5 |
"eval_logps/chosen": -34.021240234375,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.358,
|
| 15 |
"eval_steps_per_second": 0.296,
|
| 16 |
+
"train_loss": 1.3887401007986688,
|
| 17 |
+
"train_runtime": 10806.9485,
|
| 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":
|
| 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": 1.3887401007986688,
|
| 4 |
+
"train_runtime": 10806.9485,
|
| 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": 5.
|
| 501 |
-
"rewards/accuracies": 0.
|
| 502 |
-
"rewards/chosen": 0.
|
| 503 |
-
"rewards/margins": 0.
|
| 504 |
-
"rewards/rejected": -0.
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
-
"learning_rate": 4.
|
| 510 |
-
"logits/chosen": -1.
|
| 511 |
-
"logits/rejected": -1.
|
| 512 |
-
"logps/chosen": -34.
|
| 513 |
-
"logps/rejected": -33.
|
| 514 |
-
"loss": 4.
|
| 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": -33.
|
| 527 |
-
"logps/rejected": -32.
|
| 528 |
-
"loss":
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": 0.
|
| 531 |
-
"rewards/margins": 0.
|
| 532 |
-
"rewards/rejected": -0.
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -33.
|
| 541 |
-
"logps/rejected": -33.
|
| 542 |
-
"loss":
|
| 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": 4.
|
| 557 |
-
"rewards/accuracies": 0.
|
| 558 |
-
"rewards/chosen": 0.
|
| 559 |
-
"rewards/margins": 0.
|
| 560 |
-
"rewards/rejected": -0.
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
-
"learning_rate":
|
| 566 |
-
"logits/chosen": -1.
|
| 567 |
-
"logits/rejected": -1.
|
| 568 |
-
"logps/chosen": -31.
|
| 569 |
-
"logps/rejected": -35.
|
| 570 |
-
"loss": 4.
|
| 571 |
-
"rewards/accuracies": 0.
|
| 572 |
-
"rewards/chosen": 0.
|
| 573 |
-
"rewards/margins": 0.
|
| 574 |
-
"rewards/rejected": -0.
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
-
"learning_rate":
|
| 580 |
-
"logits/chosen": -2.
|
| 581 |
-
"logits/rejected": -2.
|
| 582 |
-
"logps/chosen": -33.
|
| 583 |
-
"logps/rejected": -29.
|
| 584 |
-
"loss": 4.
|
| 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": 4.
|
| 599 |
-
"rewards/accuracies": 0.
|
| 600 |
-
"rewards/chosen": 0.
|
| 601 |
-
"rewards/margins": 0.
|
| 602 |
-
"rewards/rejected": -0.
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
-
"epoch": 1.
|
| 607 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 608 |
"total_flos": 0.0,
|
| 609 |
-
"train_loss":
|
| 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": -1.9132274389266968,
|
| 497 |
+
"logits/rejected": -1.9099791049957275,
|
| 498 |
+
"logps/chosen": -31.283920288085938,
|
| 499 |
+
"logps/rejected": -33.752479553222656,
|
| 500 |
+
"loss": 5.1369,
|
| 501 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 502 |
+
"rewards/chosen": 0.06388597935438156,
|
| 503 |
+
"rewards/margins": 0.06595198810100555,
|
| 504 |
+
"rewards/rejected": -0.002066010609269142,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.961586356163025,
|
| 511 |
+
"logits/rejected": -1.949385404586792,
|
| 512 |
+
"logps/chosen": -34.29393768310547,
|
| 513 |
+
"logps/rejected": -33.63207244873047,
|
| 514 |
+
"loss": 4.8373,
|
| 515 |
+
"rewards/accuracies": 0.675000011920929,
|
| 516 |
+
"rewards/chosen": 0.0604279451072216,
|
| 517 |
+
"rewards/margins": 0.07907330244779587,
|
| 518 |
+
"rewards/rejected": -0.01864534802734852,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -1.9958088397979736,
|
| 525 |
+
"logits/rejected": -1.9943931102752686,
|
| 526 |
+
"logps/chosen": -33.11198043823242,
|
| 527 |
+
"logps/rejected": -32.48575973510742,
|
| 528 |
+
"loss": 4.8522,
|
| 529 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 530 |
+
"rewards/chosen": 0.07476169615983963,
|
| 531 |
+
"rewards/margins": 0.07599709928035736,
|
| 532 |
+
"rewards/rejected": -0.0012354092905297875,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.0807433128356934,
|
| 539 |
+
"logits/rejected": -2.0650699138641357,
|
| 540 |
+
"logps/chosen": -33.721527099609375,
|
| 541 |
+
"logps/rejected": -33.0389518737793,
|
| 542 |
+
"loss": 4.8887,
|
| 543 |
+
"rewards/accuracies": 0.675000011920929,
|
| 544 |
+
"rewards/chosen": 0.09157303720712662,
|
| 545 |
+
"rewards/margins": 0.07575628161430359,
|
| 546 |
+
"rewards/rejected": 0.015816759318113327,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.9550502300262451,
|
| 553 |
+
"logits/rejected": -1.9541908502578735,
|
| 554 |
+
"logps/chosen": -32.77659606933594,
|
| 555 |
+
"logps/rejected": -32.46695327758789,
|
| 556 |
+
"loss": 4.6833,
|
| 557 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 558 |
+
"rewards/chosen": 0.10109980404376984,
|
| 559 |
+
"rewards/margins": 0.10146143287420273,
|
| 560 |
+
"rewards/rejected": -0.0003616276371758431,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.9093825817108154,
|
| 567 |
+
"logits/rejected": -1.919638991355896,
|
| 568 |
+
"logps/chosen": -31.725839614868164,
|
| 569 |
+
"logps/rejected": -35.360939025878906,
|
| 570 |
+
"loss": 4.4471,
|
| 571 |
+
"rewards/accuracies": 0.762499988079071,
|
| 572 |
+
"rewards/chosen": 0.09577047824859619,
|
| 573 |
+
"rewards/margins": 0.11020632833242416,
|
| 574 |
+
"rewards/rejected": -0.014435847289860249,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.0445544719696045,
|
| 581 |
+
"logits/rejected": -2.038135051727295,
|
| 582 |
+
"logps/chosen": -33.22490692138672,
|
| 583 |
+
"logps/rejected": -29.234344482421875,
|
| 584 |
+
"loss": 4.6128,
|
| 585 |
+
"rewards/accuracies": 0.762499988079071,
|
| 586 |
+
"rewards/chosen": 0.08541258424520493,
|
| 587 |
+
"rewards/margins": 0.08699630200862885,
|
| 588 |
+
"rewards/rejected": -0.0015837062383070588,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.9023174047470093,
|
| 595 |
+
"logits/rejected": -1.9045337438583374,
|
| 596 |
+
"logps/chosen": -33.59349060058594,
|
| 597 |
+
"logps/rejected": -30.87937355041504,
|
| 598 |
+
"loss": 4.2966,
|
| 599 |
+
"rewards/accuracies": 0.800000011920929,
|
| 600 |
+
"rewards/chosen": 0.12911152839660645,
|
| 601 |
+
"rewards/margins": 0.13046079874038696,
|
| 602 |
+
"rewards/rejected": -0.0013492825673893094,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9307386875152588,
|
| 609 |
+
"logits/rejected": -1.9295110702514648,
|
| 610 |
+
"logps/chosen": -33.762290954589844,
|
| 611 |
+
"logps/rejected": -35.92414474487305,
|
| 612 |
+
"loss": 3.4964,
|
| 613 |
+
"rewards/accuracies": 0.8333333730697632,
|
| 614 |
+
"rewards/chosen": 0.1244986280798912,
|
| 615 |
+
"rewards/margins": 0.1781271994113922,
|
| 616 |
+
"rewards/rejected": -0.05362857133150101,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.8655402660369873,
|
| 623 |
+
"logits/rejected": -1.8571646213531494,
|
| 624 |
+
"logps/chosen": -31.09210205078125,
|
| 625 |
+
"logps/rejected": -36.32701873779297,
|
| 626 |
+
"loss": 2.9753,
|
| 627 |
+
"rewards/accuracies": 0.887499988079071,
|
| 628 |
+
"rewards/chosen": 0.13637451827526093,
|
| 629 |
+
"rewards/margins": 0.2136588841676712,
|
| 630 |
+
"rewards/rejected": -0.07728435099124908,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.2141356468200684,
|
| 636 |
+
"eval_logits/rejected": -2.209277629852295,
|
| 637 |
+
"eval_logps/chosen": -33.99992752075195,
|
| 638 |
+
"eval_logps/rejected": -37.56977081298828,
|
| 639 |
+
"eval_loss": 6.036898612976074,
|
| 640 |
+
"eval_rewards/accuracies": 0.6034052968025208,
|
| 641 |
+
"eval_rewards/chosen": 0.006925266236066818,
|
| 642 |
+
"eval_rewards/margins": 0.01755475252866745,
|
| 643 |
+
"eval_rewards/rejected": -0.010629487223923206,
|
| 644 |
+
"eval_runtime": 146.4885,
|
| 645 |
+
"eval_samples_per_second": 2.341,
|
| 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": -2.038944721221924,
|
| 653 |
+
"logits/rejected": -2.0418522357940674,
|
| 654 |
+
"logps/chosen": -32.3203010559082,
|
| 655 |
+
"logps/rejected": -35.14866638183594,
|
| 656 |
+
"loss": 2.7334,
|
| 657 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 658 |
+
"rewards/chosen": 0.1467280387878418,
|
| 659 |
+
"rewards/margins": 0.2178923636674881,
|
| 660 |
+
"rewards/rejected": -0.07116430997848511,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.8793596029281616,
|
| 667 |
+
"logits/rejected": -1.877995252609253,
|
| 668 |
+
"logps/chosen": -28.413448333740234,
|
| 669 |
+
"logps/rejected": -32.73353958129883,
|
| 670 |
+
"loss": 2.6582,
|
| 671 |
+
"rewards/accuracies": 0.9375,
|
| 672 |
+
"rewards/chosen": 0.1308210790157318,
|
| 673 |
+
"rewards/margins": 0.21399688720703125,
|
| 674 |
+
"rewards/rejected": -0.08317580074071884,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.8368927240371704,
|
| 681 |
+
"logits/rejected": -1.8299328088760376,
|
| 682 |
+
"logps/chosen": -33.15730667114258,
|
| 683 |
+
"logps/rejected": -34.441612243652344,
|
| 684 |
+
"loss": 2.4992,
|
| 685 |
+
"rewards/accuracies": 0.949999988079071,
|
| 686 |
+
"rewards/chosen": 0.18335571885108948,
|
| 687 |
+
"rewards/margins": 0.2226235568523407,
|
| 688 |
+
"rewards/rejected": -0.039267830550670624,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -2.0078189373016357,
|
| 695 |
+
"logits/rejected": -2.0026912689208984,
|
| 696 |
+
"logps/chosen": -30.97007179260254,
|
| 697 |
+
"logps/rejected": -32.39707565307617,
|
| 698 |
+
"loss": 2.5198,
|
| 699 |
+
"rewards/accuracies": 0.925000011920929,
|
| 700 |
+
"rewards/chosen": 0.16264474391937256,
|
| 701 |
+
"rewards/margins": 0.24853797256946564,
|
| 702 |
+
"rewards/rejected": -0.08589322865009308,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.8930528163909912,
|
| 709 |
+
"logits/rejected": -1.897268295288086,
|
| 710 |
+
"logps/chosen": -33.69664764404297,
|
| 711 |
+
"logps/rejected": -33.11066818237305,
|
| 712 |
+
"loss": 2.0077,
|
| 713 |
+
"rewards/accuracies": 0.949999988079071,
|
| 714 |
+
"rewards/chosen": 0.16384032368659973,
|
| 715 |
+
"rewards/margins": 0.26824674010276794,
|
| 716 |
+
"rewards/rejected": -0.1044064313173294,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.9001489877700806,
|
| 723 |
+
"logits/rejected": -1.8944313526153564,
|
| 724 |
+
"logps/chosen": -34.53886032104492,
|
| 725 |
+
"logps/rejected": -35.5705451965332,
|
| 726 |
+
"loss": 2.054,
|
| 727 |
+
"rewards/accuracies": 0.949999988079071,
|
| 728 |
+
"rewards/chosen": 0.17613370716571808,
|
| 729 |
+
"rewards/margins": 0.30042320489883423,
|
| 730 |
+
"rewards/rejected": -0.12428952753543854,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.929231882095337,
|
| 737 |
+
"logits/rejected": -1.9289064407348633,
|
| 738 |
+
"logps/chosen": -33.301231384277344,
|
| 739 |
+
"logps/rejected": -34.58076858520508,
|
| 740 |
+
"loss": 2.5855,
|
| 741 |
+
"rewards/accuracies": 0.949999988079071,
|
| 742 |
+
"rewards/chosen": 0.17341957986354828,
|
| 743 |
+
"rewards/margins": 0.2524762451648712,
|
| 744 |
+
"rewards/rejected": -0.07905664294958115,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -1.9759957790374756,
|
| 751 |
+
"logits/rejected": -1.9571311473846436,
|
| 752 |
+
"logps/chosen": -33.301368713378906,
|
| 753 |
+
"logps/rejected": -33.618614196777344,
|
| 754 |
+
"loss": 2.3815,
|
| 755 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 756 |
+
"rewards/chosen": 0.171879380941391,
|
| 757 |
+
"rewards/margins": 0.28160572052001953,
|
| 758 |
+
"rewards/rejected": -0.10972632467746735,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.9400146007537842,
|
| 765 |
+
"logits/rejected": -1.9393583536148071,
|
| 766 |
+
"logps/chosen": -30.776290893554688,
|
| 767 |
+
"logps/rejected": -31.59432029724121,
|
| 768 |
+
"loss": 2.4535,
|
| 769 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 770 |
+
"rewards/chosen": 0.16700208187103271,
|
| 771 |
+
"rewards/margins": 0.2366572916507721,
|
| 772 |
+
"rewards/rejected": -0.06965517997741699,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.9039013385772705,
|
| 779 |
+
"logits/rejected": -1.8969634771347046,
|
| 780 |
+
"logps/chosen": -31.851303100585938,
|
| 781 |
+
"logps/rejected": -35.185157775878906,
|
| 782 |
+
"loss": 2.4163,
|
| 783 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 784 |
+
"rewards/chosen": 0.16376493871212006,
|
| 785 |
+
"rewards/margins": 0.2754477858543396,
|
| 786 |
+
"rewards/rejected": -0.11168281733989716,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.190660238265991,
|
| 792 |
+
"eval_logits/rejected": -2.185832977294922,
|
| 793 |
+
"eval_logps/chosen": -34.10919189453125,
|
| 794 |
+
"eval_logps/rejected": -37.70389938354492,
|
| 795 |
+
"eval_loss": 6.067742347717285,
|
| 796 |
+
"eval_rewards/accuracies": 0.5801494717597961,
|
| 797 |
+
"eval_rewards/chosen": -0.014928131364285946,
|
| 798 |
+
"eval_rewards/margins": 0.02252793498337269,
|
| 799 |
+
"eval_rewards/rejected": -0.03745606541633606,
|
| 800 |
+
"eval_runtime": 145.8919,
|
| 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": -1.9019960165023804,
|
| 809 |
+
"logits/rejected": -1.9057096242904663,
|
| 810 |
+
"logps/chosen": -28.43143081665039,
|
| 811 |
+
"logps/rejected": -33.44213104248047,
|
| 812 |
+
"loss": 2.4231,
|
| 813 |
+
"rewards/accuracies": 0.887499988079071,
|
| 814 |
+
"rewards/chosen": 0.1486787497997284,
|
| 815 |
+
"rewards/margins": 0.2652469575405121,
|
| 816 |
+
"rewards/rejected": -0.11656824499368668,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.9110679626464844,
|
| 823 |
+
"logits/rejected": -1.921726942062378,
|
| 824 |
+
"logps/chosen": -32.49781036376953,
|
| 825 |
+
"logps/rejected": -31.106653213500977,
|
| 826 |
+
"loss": 2.6543,
|
| 827 |
+
"rewards/accuracies": 0.887499988079071,
|
| 828 |
+
"rewards/chosen": 0.15290473401546478,
|
| 829 |
+
"rewards/margins": 0.2438981980085373,
|
| 830 |
+
"rewards/rejected": -0.0909934788942337,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.8584115505218506,
|
| 837 |
+
"logits/rejected": -1.8560092449188232,
|
| 838 |
+
"logps/chosen": -30.80838394165039,
|
| 839 |
+
"logps/rejected": -30.5560302734375,
|
| 840 |
+
"loss": 2.4362,
|
| 841 |
+
"rewards/accuracies": 0.887499988079071,
|
| 842 |
+
"rewards/chosen": 0.17173974215984344,
|
| 843 |
+
"rewards/margins": 0.2475927770137787,
|
| 844 |
+
"rewards/rejected": -0.07585303485393524,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.9417930841445923,
|
| 851 |
+
"logits/rejected": -1.9347444772720337,
|
| 852 |
+
"logps/chosen": -33.26526641845703,
|
| 853 |
+
"logps/rejected": -33.53062057495117,
|
| 854 |
+
"loss": 2.2112,
|
| 855 |
+
"rewards/accuracies": 0.887499988079071,
|
| 856 |
+
"rewards/chosen": 0.19351224601268768,
|
| 857 |
+
"rewards/margins": 0.2961391806602478,
|
| 858 |
+
"rewards/rejected": -0.10262690484523773,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.9108800888061523,
|
| 865 |
+
"logits/rejected": -1.9172462224960327,
|
| 866 |
+
"logps/chosen": -30.884307861328125,
|
| 867 |
+
"logps/rejected": -33.133323669433594,
|
| 868 |
+
"loss": 2.3198,
|
| 869 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 870 |
+
"rewards/chosen": 0.17377518117427826,
|
| 871 |
+
"rewards/margins": 0.248693585395813,
|
| 872 |
+
"rewards/rejected": -0.07491841167211533,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -1.9825620651245117,
|
| 879 |
+
"logits/rejected": -1.9594837427139282,
|
| 880 |
+
"logps/chosen": -30.586624145507812,
|
| 881 |
+
"logps/rejected": -33.408687591552734,
|
| 882 |
+
"loss": 2.6429,
|
| 883 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 884 |
+
"rewards/chosen": 0.17018920183181763,
|
| 885 |
+
"rewards/margins": 0.25271496176719666,
|
| 886 |
+
"rewards/rejected": -0.08252575248479843,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.9179880619049072,
|
| 893 |
+
"logits/rejected": -1.9130771160125732,
|
| 894 |
+
"logps/chosen": -31.49106216430664,
|
| 895 |
+
"logps/rejected": -32.453086853027344,
|
| 896 |
+
"loss": 2.7291,
|
| 897 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 898 |
+
"rewards/chosen": 0.18513426184654236,
|
| 899 |
+
"rewards/margins": 0.25156813859939575,
|
| 900 |
+
"rewards/rejected": -0.0664338767528534,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -1.9604476690292358,
|
| 907 |
+
"logits/rejected": -1.9623527526855469,
|
| 908 |
+
"logps/chosen": -33.557342529296875,
|
| 909 |
+
"logps/rejected": -35.4670524597168,
|
| 910 |
+
"loss": 2.2236,
|
| 911 |
+
"rewards/accuracies": 0.9375,
|
| 912 |
+
"rewards/chosen": 0.1636902391910553,
|
| 913 |
+
"rewards/margins": 0.29691803455352783,
|
| 914 |
+
"rewards/rejected": -0.13322779536247253,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.9461857080459595,
|
| 921 |
+
"logits/rejected": -1.9455292224884033,
|
| 922 |
+
"logps/chosen": -29.717754364013672,
|
| 923 |
+
"logps/rejected": -31.953426361083984,
|
| 924 |
+
"loss": 2.4743,
|
| 925 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 926 |
+
"rewards/chosen": 0.17486092448234558,
|
| 927 |
+
"rewards/margins": 0.2732195258140564,
|
| 928 |
+
"rewards/rejected": -0.09835861623287201,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -1.9628851413726807,
|
| 935 |
+
"logits/rejected": -1.960695505142212,
|
| 936 |
+
"logps/chosen": -32.0013542175293,
|
| 937 |
+
"logps/rejected": -32.742637634277344,
|
| 938 |
+
"loss": 2.52,
|
| 939 |
+
"rewards/accuracies": 0.9375,
|
| 940 |
+
"rewards/chosen": 0.19938594102859497,
|
| 941 |
+
"rewards/margins": 0.2879222333431244,
|
| 942 |
+
"rewards/rejected": -0.08853629976511002,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.1998937129974365,
|
| 948 |
+
"eval_logits/rejected": -2.195075750350952,
|
| 949 |
+
"eval_logps/chosen": -34.08315658569336,
|
| 950 |
+
"eval_logps/rejected": -37.69052505493164,
|
| 951 |
+
"eval_loss": 5.999026775360107,
|
| 952 |
+
"eval_rewards/accuracies": 0.5747508406639099,
|
| 953 |
+
"eval_rewards/chosen": -0.009720847941935062,
|
| 954 |
+
"eval_rewards/margins": 0.025059644132852554,
|
| 955 |
+
"eval_rewards/rejected": -0.03478049114346504,
|
| 956 |
+
"eval_runtime": 145.9294,
|
| 957 |
+
"eval_samples_per_second": 2.35,
|
| 958 |
+
"eval_steps_per_second": 0.295,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 1.58,
|
| 963 |
+
"learning_rate": 3.7793282895240927e-06,
|
| 964 |
+
"logits/chosen": -2.0093092918395996,
|
| 965 |
+
"logits/rejected": -2.0159478187561035,
|
| 966 |
+
"logps/chosen": -31.2258358001709,
|
| 967 |
+
"logps/rejected": -32.541290283203125,
|
| 968 |
+
"loss": 2.1606,
|
| 969 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 970 |
+
"rewards/chosen": 0.20230896770954132,
|
| 971 |
+
"rewards/margins": 0.29792895913124084,
|
| 972 |
+
"rewards/rejected": -0.09561996161937714,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -1.9672517776489258,
|
| 979 |
+
"logits/rejected": -1.9648239612579346,
|
| 980 |
+
"logps/chosen": -33.72132110595703,
|
| 981 |
+
"logps/rejected": -31.379138946533203,
|
| 982 |
+
"loss": 2.198,
|
| 983 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 984 |
+
"rewards/chosen": 0.20691220462322235,
|
| 985 |
+
"rewards/margins": 0.2856389880180359,
|
| 986 |
+
"rewards/rejected": -0.07872680574655533,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.912590742111206,
|
| 993 |
+
"logits/rejected": -1.9057750701904297,
|
| 994 |
+
"logps/chosen": -34.36874008178711,
|
| 995 |
+
"logps/rejected": -32.927589416503906,
|
| 996 |
+
"loss": 2.2377,
|
| 997 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 998 |
+
"rewards/chosen": 0.2190968096256256,
|
| 999 |
+
"rewards/margins": 0.3331736922264099,
|
| 1000 |
+
"rewards/rejected": -0.11407686769962311,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.9224069118499756,
|
| 1007 |
+
"logits/rejected": -1.9286394119262695,
|
| 1008 |
+
"logps/chosen": -33.01720428466797,
|
| 1009 |
+
"logps/rejected": -33.65998840332031,
|
| 1010 |
+
"loss": 2.4802,
|
| 1011 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1012 |
+
"rewards/chosen": 0.19241946935653687,
|
| 1013 |
+
"rewards/margins": 0.2763652503490448,
|
| 1014 |
+
"rewards/rejected": -0.08394578844308853,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -2.0105245113372803,
|
| 1021 |
+
"logits/rejected": -2.024742603302002,
|
| 1022 |
+
"logps/chosen": -30.962276458740234,
|
| 1023 |
+
"logps/rejected": -32.379337310791016,
|
| 1024 |
+
"loss": 2.7242,
|
| 1025 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1026 |
+
"rewards/chosen": 0.17803707718849182,
|
| 1027 |
+
"rewards/margins": 0.25436651706695557,
|
| 1028 |
+
"rewards/rejected": -0.07632941007614136,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.8864635229110718,
|
| 1035 |
+
"logits/rejected": -1.8834316730499268,
|
| 1036 |
+
"logps/chosen": -32.803226470947266,
|
| 1037 |
+
"logps/rejected": -35.24103546142578,
|
| 1038 |
+
"loss": 2.0371,
|
| 1039 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1040 |
+
"rewards/chosen": 0.22394683957099915,
|
| 1041 |
+
"rewards/margins": 0.3321036398410797,
|
| 1042 |
+
"rewards/rejected": -0.10815682262182236,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -1.9782884120941162,
|
| 1049 |
+
"logits/rejected": -1.9784494638442993,
|
| 1050 |
+
"logps/chosen": -30.333904266357422,
|
| 1051 |
+
"logps/rejected": -34.28630828857422,
|
| 1052 |
+
"loss": 2.1727,
|
| 1053 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1054 |
+
"rewards/chosen": 0.19033712148666382,
|
| 1055 |
+
"rewards/margins": 0.28232187032699585,
|
| 1056 |
+
"rewards/rejected": -0.09198474884033203,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.9290120601654053,
|
| 1063 |
+
"logits/rejected": -1.9259191751480103,
|
| 1064 |
+
"logps/chosen": -29.953277587890625,
|
| 1065 |
+
"logps/rejected": -33.294857025146484,
|
| 1066 |
+
"loss": 2.5836,
|
| 1067 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1068 |
+
"rewards/chosen": 0.18381401896476746,
|
| 1069 |
+
"rewards/margins": 0.27922549843788147,
|
| 1070 |
+
"rewards/rejected": -0.09541147202253342,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -1.9663722515106201,
|
| 1077 |
+
"logits/rejected": -1.9663045406341553,
|
| 1078 |
+
"logps/chosen": -29.003856658935547,
|
| 1079 |
+
"logps/rejected": -31.04767417907715,
|
| 1080 |
+
"loss": 2.0189,
|
| 1081 |
+
"rewards/accuracies": 0.9375,
|
| 1082 |
+
"rewards/chosen": 0.23184648156166077,
|
| 1083 |
+
"rewards/margins": 0.30728113651275635,
|
| 1084 |
+
"rewards/rejected": -0.07543464004993439,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.8771839141845703,
|
| 1091 |
+
"logits/rejected": -1.8805001974105835,
|
| 1092 |
+
"logps/chosen": -33.498069763183594,
|
| 1093 |
+
"logps/rejected": -31.84822654724121,
|
| 1094 |
+
"loss": 2.9186,
|
| 1095 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1096 |
+
"rewards/chosen": 0.19608056545257568,
|
| 1097 |
+
"rewards/margins": 0.272787868976593,
|
| 1098 |
+
"rewards/rejected": -0.07670729607343674,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.2096614837646484,
|
| 1104 |
+
"eval_logits/rejected": -2.2048423290252686,
|
| 1105 |
+
"eval_logps/chosen": -34.122745513916016,
|
| 1106 |
+
"eval_logps/rejected": -37.69883728027344,
|
| 1107 |
+
"eval_loss": 6.169560432434082,
|
| 1108 |
+
"eval_rewards/accuracies": 0.5598006844520569,
|
| 1109 |
+
"eval_rewards/chosen": -0.01763882488012314,
|
| 1110 |
+
"eval_rewards/margins": 0.01880452036857605,
|
| 1111 |
+
"eval_rewards/rejected": -0.03644334897398949,
|
| 1112 |
+
"eval_runtime": 145.9673,
|
| 1113 |
+
"eval_samples_per_second": 2.35,
|
| 1114 |
+
"eval_steps_per_second": 0.295,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 1.84,
|
| 1119 |
+
"learning_rate": 3.2639145321045933e-06,
|
| 1120 |
+
"logits/chosen": -1.9619114398956299,
|
| 1121 |
+
"logits/rejected": -1.9527734518051147,
|
| 1122 |
+
"logps/chosen": -35.47718811035156,
|
| 1123 |
+
"logps/rejected": -32.34703826904297,
|
| 1124 |
+
"loss": 2.0062,
|
| 1125 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1126 |
+
"rewards/chosen": 0.20696695148944855,
|
| 1127 |
+
"rewards/margins": 0.2952113747596741,
|
| 1128 |
+
"rewards/rejected": -0.08824445307254791,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -1.9771125316619873,
|
| 1135 |
+
"logits/rejected": -1.9781534671783447,
|
| 1136 |
+
"logps/chosen": -35.389427185058594,
|
| 1137 |
+
"logps/rejected": -33.445552825927734,
|
| 1138 |
+
"loss": 2.2199,
|
| 1139 |
+
"rewards/accuracies": 0.9375,
|
| 1140 |
+
"rewards/chosen": 0.20919255912303925,
|
| 1141 |
+
"rewards/margins": 0.28309938311576843,
|
| 1142 |
+
"rewards/rejected": -0.07390682399272919,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -1.9929240942001343,
|
| 1149 |
+
"logits/rejected": -1.9904762506484985,
|
| 1150 |
+
"logps/chosen": -31.665563583374023,
|
| 1151 |
+
"logps/rejected": -33.510894775390625,
|
| 1152 |
+
"loss": 2.1505,
|
| 1153 |
+
"rewards/accuracies": 0.9375,
|
| 1154 |
+
"rewards/chosen": 0.17955288290977478,
|
| 1155 |
+
"rewards/margins": 0.28235217928886414,
|
| 1156 |
+
"rewards/rejected": -0.10279928147792816,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -1.9763247966766357,
|
| 1163 |
+
"logits/rejected": -1.981488823890686,
|
| 1164 |
+
"logps/chosen": -30.730270385742188,
|
| 1165 |
+
"logps/rejected": -33.69140625,
|
| 1166 |
+
"loss": 1.8054,
|
| 1167 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1168 |
+
"rewards/chosen": 0.21109160780906677,
|
| 1169 |
+
"rewards/margins": 0.3184094727039337,
|
| 1170 |
+
"rewards/rejected": -0.10731786489486694,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.9217841625213623,
|
| 1177 |
+
"logits/rejected": -1.9206024408340454,
|
| 1178 |
+
"logps/chosen": -32.20301818847656,
|
| 1179 |
+
"logps/rejected": -35.21897888183594,
|
| 1180 |
+
"loss": 2.2296,
|
| 1181 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1182 |
+
"rewards/chosen": 0.18439263105392456,
|
| 1183 |
+
"rewards/margins": 0.29099735617637634,
|
| 1184 |
+
"rewards/rejected": -0.10660471022129059,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.801102638244629,
|
| 1191 |
+
"logits/rejected": -1.7960617542266846,
|
| 1192 |
+
"logps/chosen": -35.06943130493164,
|
| 1193 |
+
"logps/rejected": -35.63484573364258,
|
| 1194 |
+
"loss": 2.4449,
|
| 1195 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1196 |
+
"rewards/chosen": 0.1858016401529312,
|
| 1197 |
+
"rewards/margins": 0.27108871936798096,
|
| 1198 |
+
"rewards/rejected": -0.08528711646795273,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.941235899925232,
|
| 1205 |
+
"logits/rejected": -1.9449260234832764,
|
| 1206 |
+
"logps/chosen": -34.17765808105469,
|
| 1207 |
+
"logps/rejected": -33.9752311706543,
|
| 1208 |
+
"loss": 2.3588,
|
| 1209 |
+
"rewards/accuracies": 0.8833333849906921,
|
| 1210 |
+
"rewards/chosen": 0.2072003185749054,
|
| 1211 |
+
"rewards/margins": 0.2832789123058319,
|
| 1212 |
+
"rewards/rejected": -0.07607860863208771,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.884577751159668,
|
| 1219 |
+
"logits/rejected": -1.8827228546142578,
|
| 1220 |
+
"logps/chosen": -32.60283279418945,
|
| 1221 |
+
"logps/rejected": -35.19242858886719,
|
| 1222 |
+
"loss": 0.9771,
|
| 1223 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1224 |
+
"rewards/chosen": 0.25975239276885986,
|
| 1225 |
+
"rewards/margins": 0.41730666160583496,
|
| 1226 |
+
"rewards/rejected": -0.1575542539358139,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.933282494544983,
|
| 1233 |
+
"logits/rejected": -1.931807279586792,
|
| 1234 |
+
"logps/chosen": -32.13750076293945,
|
| 1235 |
+
"logps/rejected": -33.67426300048828,
|
| 1236 |
+
"loss": 1.1284,
|
| 1237 |
+
"rewards/accuracies": 0.9375,
|
| 1238 |
+
"rewards/chosen": 0.24764792621135712,
|
| 1239 |
+
"rewards/margins": 0.3651863932609558,
|
| 1240 |
+
"rewards/rejected": -0.11753849685192108,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.8644397258758545,
|
| 1247 |
+
"logits/rejected": -1.8584182262420654,
|
| 1248 |
+
"logps/chosen": -33.09177780151367,
|
| 1249 |
+
"logps/rejected": -32.095069885253906,
|
| 1250 |
+
"loss": 1.2867,
|
| 1251 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1252 |
+
"rewards/chosen": 0.2582601010799408,
|
| 1253 |
+
"rewards/margins": 0.38442230224609375,
|
| 1254 |
+
"rewards/rejected": -0.12616217136383057,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.2108798027038574,
|
| 1260 |
+
"eval_logits/rejected": -2.2060329914093018,
|
| 1261 |
+
"eval_logps/chosen": -34.095699310302734,
|
| 1262 |
+
"eval_logps/rejected": -37.697025299072266,
|
| 1263 |
+
"eval_loss": 6.059365272521973,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5776578187942505,
|
| 1265 |
+
"eval_rewards/chosen": -0.012229476124048233,
|
| 1266 |
+
"eval_rewards/margins": 0.023852398619055748,
|
| 1267 |
+
"eval_rewards/rejected": -0.03608187288045883,
|
| 1268 |
+
"eval_runtime": 145.9095,
|
| 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": -1.965205430984497,
|
| 1277 |
+
"logits/rejected": -1.972726821899414,
|
| 1278 |
+
"logps/chosen": -30.922103881835938,
|
| 1279 |
+
"logps/rejected": -34.932464599609375,
|
| 1280 |
+
"loss": 1.1344,
|
| 1281 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1282 |
+
"rewards/chosen": 0.24123027920722961,
|
| 1283 |
+
"rewards/margins": 0.37655484676361084,
|
| 1284 |
+
"rewards/rejected": -0.13532456755638123,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.9390350580215454,
|
| 1291 |
+
"logits/rejected": -1.9418513774871826,
|
| 1292 |
+
"logps/chosen": -31.403156280517578,
|
| 1293 |
+
"logps/rejected": -33.695457458496094,
|
| 1294 |
+
"loss": 1.4773,
|
| 1295 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1296 |
+
"rewards/chosen": 0.23358435928821564,
|
| 1297 |
+
"rewards/margins": 0.32671505212783813,
|
| 1298 |
+
"rewards/rejected": -0.09313070774078369,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.8422248363494873,
|
| 1305 |
+
"logits/rejected": -1.8463550806045532,
|
| 1306 |
+
"logps/chosen": -31.341222763061523,
|
| 1307 |
+
"logps/rejected": -35.71354675292969,
|
| 1308 |
+
"loss": 0.9876,
|
| 1309 |
+
"rewards/accuracies": 1.0,
|
| 1310 |
+
"rewards/chosen": 0.2500464916229248,
|
| 1311 |
+
"rewards/margins": 0.4107195734977722,
|
| 1312 |
+
"rewards/rejected": -0.16067302227020264,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.904828429222107,
|
| 1319 |
+
"logits/rejected": -1.9009695053100586,
|
| 1320 |
+
"logps/chosen": -31.746740341186523,
|
| 1321 |
+
"logps/rejected": -33.4873161315918,
|
| 1322 |
+
"loss": 1.6306,
|
| 1323 |
+
"rewards/accuracies": 0.9375,
|
| 1324 |
+
"rewards/chosen": 0.2443816363811493,
|
| 1325 |
+
"rewards/margins": 0.3429158329963684,
|
| 1326 |
+
"rewards/rejected": -0.09853418916463852,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -1.98971426486969,
|
| 1333 |
+
"logits/rejected": -1.991711974143982,
|
| 1334 |
+
"logps/chosen": -31.50253677368164,
|
| 1335 |
+
"logps/rejected": -36.3741569519043,
|
| 1336 |
+
"loss": 0.9313,
|
| 1337 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1338 |
+
"rewards/chosen": 0.2545596957206726,
|
| 1339 |
+
"rewards/margins": 0.39700326323509216,
|
| 1340 |
+
"rewards/rejected": -0.14244356751441956,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.7937977313995361,
|
| 1347 |
+
"logits/rejected": -1.7878059148788452,
|
| 1348 |
+
"logps/chosen": -35.0206184387207,
|
| 1349 |
+
"logps/rejected": -33.611572265625,
|
| 1350 |
+
"loss": 1.1249,
|
| 1351 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1352 |
+
"rewards/chosen": 0.2522119879722595,
|
| 1353 |
+
"rewards/margins": 0.39614588022232056,
|
| 1354 |
+
"rewards/rejected": -0.14393387734889984,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.8565397262573242,
|
| 1361 |
+
"logits/rejected": -1.85663640499115,
|
| 1362 |
+
"logps/chosen": -34.53380584716797,
|
| 1363 |
+
"logps/rejected": -35.518733978271484,
|
| 1364 |
+
"loss": 1.1272,
|
| 1365 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1366 |
+
"rewards/chosen": 0.2618001401424408,
|
| 1367 |
+
"rewards/margins": 0.396293580532074,
|
| 1368 |
+
"rewards/rejected": -0.13449345529079437,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.9093639850616455,
|
| 1375 |
+
"logits/rejected": -1.897884726524353,
|
| 1376 |
+
"logps/chosen": -32.93827819824219,
|
| 1377 |
+
"logps/rejected": -33.690372467041016,
|
| 1378 |
+
"loss": 0.865,
|
| 1379 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1380 |
+
"rewards/chosen": 0.23366093635559082,
|
| 1381 |
+
"rewards/margins": 0.3873291015625,
|
| 1382 |
+
"rewards/rejected": -0.15366819500923157,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -2.010274648666382,
|
| 1389 |
+
"logits/rejected": -2.014864444732666,
|
| 1390 |
+
"logps/chosen": -32.14999771118164,
|
| 1391 |
+
"logps/rejected": -33.120323181152344,
|
| 1392 |
+
"loss": 1.1326,
|
| 1393 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1394 |
+
"rewards/chosen": 0.23598790168762207,
|
| 1395 |
+
"rewards/margins": 0.3727555572986603,
|
| 1396 |
+
"rewards/rejected": -0.1367676854133606,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.9262173175811768,
|
| 1403 |
+
"logits/rejected": -1.9189611673355103,
|
| 1404 |
+
"logps/chosen": -33.60108184814453,
|
| 1405 |
+
"logps/rejected": -33.117454528808594,
|
| 1406 |
+
"loss": 0.8862,
|
| 1407 |
+
"rewards/accuracies": 1.0,
|
| 1408 |
+
"rewards/chosen": 0.2505994439125061,
|
| 1409 |
+
"rewards/margins": 0.39366382360458374,
|
| 1410 |
+
"rewards/rejected": -0.14306436479091644,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.207599639892578,
|
| 1416 |
+
"eval_logits/rejected": -2.202746629714966,
|
| 1417 |
+
"eval_logps/chosen": -34.117210388183594,
|
| 1418 |
+
"eval_logps/rejected": -37.7178955078125,
|
| 1419 |
+
"eval_loss": 6.062132835388184,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5917773842811584,
|
| 1421 |
+
"eval_rewards/chosen": -0.01653219200670719,
|
| 1422 |
+
"eval_rewards/margins": 0.02372238226234913,
|
| 1423 |
+
"eval_rewards/rejected": -0.04025457799434662,
|
| 1424 |
+
"eval_runtime": 145.9709,
|
| 1425 |
+
"eval_samples_per_second": 2.35,
|
| 1426 |
+
"eval_steps_per_second": 0.295,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 2.36,
|
| 1431 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 1432 |
+
"logits/chosen": -1.9199193716049194,
|
| 1433 |
+
"logits/rejected": -1.9204641580581665,
|
| 1434 |
+
"logps/chosen": -30.064319610595703,
|
| 1435 |
+
"logps/rejected": -35.48960494995117,
|
| 1436 |
+
"loss": 1.2088,
|
| 1437 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1438 |
+
"rewards/chosen": 0.22056813538074493,
|
| 1439 |
+
"rewards/margins": 0.38055095076560974,
|
| 1440 |
+
"rewards/rejected": -0.15998278558254242,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.8791996240615845,
|
| 1447 |
+
"logits/rejected": -1.871145248413086,
|
| 1448 |
+
"logps/chosen": -30.649662017822266,
|
| 1449 |
+
"logps/rejected": -34.292842864990234,
|
| 1450 |
+
"loss": 1.1982,
|
| 1451 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1452 |
+
"rewards/chosen": 0.23785707354545593,
|
| 1453 |
+
"rewards/margins": 0.3821527063846588,
|
| 1454 |
+
"rewards/rejected": -0.14429563283920288,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -2.044569969177246,
|
| 1461 |
+
"logits/rejected": -2.045048236846924,
|
| 1462 |
+
"logps/chosen": -28.890884399414062,
|
| 1463 |
+
"logps/rejected": -33.5275993347168,
|
| 1464 |
+
"loss": 1.6139,
|
| 1465 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1466 |
+
"rewards/chosen": 0.22123007476329803,
|
| 1467 |
+
"rewards/margins": 0.3537783920764923,
|
| 1468 |
+
"rewards/rejected": -0.13254830241203308,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.9257404804229736,
|
| 1475 |
+
"logits/rejected": -1.935403823852539,
|
| 1476 |
+
"logps/chosen": -33.113128662109375,
|
| 1477 |
+
"logps/rejected": -33.752899169921875,
|
| 1478 |
+
"loss": 1.3358,
|
| 1479 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1480 |
+
"rewards/chosen": 0.2623876631259918,
|
| 1481 |
+
"rewards/margins": 0.415038526058197,
|
| 1482 |
+
"rewards/rejected": -0.15265092253684998,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -1.9671064615249634,
|
| 1489 |
+
"logits/rejected": -1.9709781408309937,
|
| 1490 |
+
"logps/chosen": -29.538061141967773,
|
| 1491 |
+
"logps/rejected": -32.42096710205078,
|
| 1492 |
+
"loss": 1.1713,
|
| 1493 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1494 |
+
"rewards/chosen": 0.22193200886249542,
|
| 1495 |
+
"rewards/margins": 0.3589089512825012,
|
| 1496 |
+
"rewards/rejected": -0.13697698712348938,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -1.9648845195770264,
|
| 1503 |
+
"logits/rejected": -1.96088445186615,
|
| 1504 |
+
"logps/chosen": -32.74964141845703,
|
| 1505 |
+
"logps/rejected": -34.44969940185547,
|
| 1506 |
+
"loss": 1.3059,
|
| 1507 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1508 |
+
"rewards/chosen": 0.26511532068252563,
|
| 1509 |
+
"rewards/margins": 0.3737781345844269,
|
| 1510 |
+
"rewards/rejected": -0.10866282880306244,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.8809791803359985,
|
| 1517 |
+
"logits/rejected": -1.883603811264038,
|
| 1518 |
+
"logps/chosen": -31.52639389038086,
|
| 1519 |
+
"logps/rejected": -35.896881103515625,
|
| 1520 |
+
"loss": 0.8379,
|
| 1521 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1522 |
+
"rewards/chosen": 0.2572704553604126,
|
| 1523 |
+
"rewards/margins": 0.40032801032066345,
|
| 1524 |
+
"rewards/rejected": -0.14305754005908966,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.9164276123046875,
|
| 1531 |
+
"logits/rejected": -1.9099451303482056,
|
| 1532 |
+
"logps/chosen": -34.2746696472168,
|
| 1533 |
+
"logps/rejected": -36.992088317871094,
|
| 1534 |
+
"loss": 1.5424,
|
| 1535 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1536 |
+
"rewards/chosen": 0.2260798215866089,
|
| 1537 |
+
"rewards/margins": 0.3721070885658264,
|
| 1538 |
+
"rewards/rejected": -0.14602726697921753,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.890472412109375,
|
| 1545 |
+
"logits/rejected": -1.8852546215057373,
|
| 1546 |
+
"logps/chosen": -29.865234375,
|
| 1547 |
+
"logps/rejected": -36.65885543823242,
|
| 1548 |
+
"loss": 1.2374,
|
| 1549 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1550 |
+
"rewards/chosen": 0.22754326462745667,
|
| 1551 |
+
"rewards/margins": 0.3613375723361969,
|
| 1552 |
+
"rewards/rejected": -0.13379430770874023,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.8666467666625977,
|
| 1559 |
+
"logits/rejected": -1.8558076620101929,
|
| 1560 |
+
"logps/chosen": -30.546798706054688,
|
| 1561 |
+
"logps/rejected": -30.192230224609375,
|
| 1562 |
+
"loss": 1.2395,
|
| 1563 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1564 |
+
"rewards/chosen": 0.2403787076473236,
|
| 1565 |
+
"rewards/margins": 0.35887858271598816,
|
| 1566 |
+
"rewards/rejected": -0.11849988996982574,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.2049996852874756,
|
| 1572 |
+
"eval_logits/rejected": -2.2001616954803467,
|
| 1573 |
+
"eval_logps/chosen": -34.11613845825195,
|
| 1574 |
+
"eval_logps/rejected": -37.72568130493164,
|
| 1575 |
+
"eval_loss": 6.000017166137695,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5863787531852722,
|
| 1577 |
+
"eval_rewards/chosen": -0.016317714005708694,
|
| 1578 |
+
"eval_rewards/margins": 0.025494439527392387,
|
| 1579 |
+
"eval_rewards/rejected": -0.04181215539574623,
|
| 1580 |
+
"eval_runtime": 145.9275,
|
| 1581 |
+
"eval_samples_per_second": 2.35,
|
| 1582 |
+
"eval_steps_per_second": 0.295,
|
| 1583 |
+
"step": 1000
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 2.62,
|
| 1587 |
+
"learning_rate": 1.5972066659083796e-06,
|
| 1588 |
+
"logits/chosen": -1.9777822494506836,
|
| 1589 |
+
"logits/rejected": -1.977115273475647,
|
| 1590 |
+
"logps/chosen": -30.621231079101562,
|
| 1591 |
+
"logps/rejected": -30.826059341430664,
|
| 1592 |
+
"loss": 1.3924,
|
| 1593 |
+
"rewards/accuracies": 0.9375,
|
| 1594 |
+
"rewards/chosen": 0.2405930459499359,
|
| 1595 |
+
"rewards/margins": 0.35410913825035095,
|
| 1596 |
+
"rewards/rejected": -0.11351609230041504,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -2.015455961227417,
|
| 1603 |
+
"logits/rejected": -2.0161612033843994,
|
| 1604 |
+
"logps/chosen": -30.704198837280273,
|
| 1605 |
+
"logps/rejected": -31.847875595092773,
|
| 1606 |
+
"loss": 1.1721,
|
| 1607 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1608 |
+
"rewards/chosen": 0.24787600338459015,
|
| 1609 |
+
"rewards/margins": 0.3842898905277252,
|
| 1610 |
+
"rewards/rejected": -0.1364138424396515,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -1.9872283935546875,
|
| 1617 |
+
"logits/rejected": -1.9903920888900757,
|
| 1618 |
+
"logps/chosen": -31.16965675354004,
|
| 1619 |
+
"logps/rejected": -28.573312759399414,
|
| 1620 |
+
"loss": 1.2521,
|
| 1621 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1622 |
+
"rewards/chosen": 0.2220829278230667,
|
| 1623 |
+
"rewards/margins": 0.3400026857852936,
|
| 1624 |
+
"rewards/rejected": -0.11791972070932388,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -2.000020980834961,
|
| 1631 |
+
"logits/rejected": -1.9921737909317017,
|
| 1632 |
+
"logps/chosen": -29.917871475219727,
|
| 1633 |
+
"logps/rejected": -34.14818572998047,
|
| 1634 |
+
"loss": 1.036,
|
| 1635 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1636 |
+
"rewards/chosen": 0.25200483202934265,
|
| 1637 |
+
"rewards/margins": 0.38159602880477905,
|
| 1638 |
+
"rewards/rejected": -0.12959113717079163,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.9298603534698486,
|
| 1645 |
+
"logits/rejected": -1.940890908241272,
|
| 1646 |
+
"logps/chosen": -32.330406188964844,
|
| 1647 |
+
"logps/rejected": -32.83357238769531,
|
| 1648 |
+
"loss": 1.158,
|
| 1649 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1650 |
+
"rewards/chosen": 0.2445407211780548,
|
| 1651 |
+
"rewards/margins": 0.3622528612613678,
|
| 1652 |
+
"rewards/rejected": -0.1177120953798294,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -2.009378433227539,
|
| 1659 |
+
"logits/rejected": -1.9999854564666748,
|
| 1660 |
+
"logps/chosen": -32.38091278076172,
|
| 1661 |
+
"logps/rejected": -37.074951171875,
|
| 1662 |
+
"loss": 1.4986,
|
| 1663 |
+
"rewards/accuracies": 0.9375,
|
| 1664 |
+
"rewards/chosen": 0.22397327423095703,
|
| 1665 |
+
"rewards/margins": 0.37268632650375366,
|
| 1666 |
+
"rewards/rejected": -0.1487130969762802,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.9161951541900635,
|
| 1673 |
+
"logits/rejected": -1.9179637432098389,
|
| 1674 |
+
"logps/chosen": -34.65896987915039,
|
| 1675 |
+
"logps/rejected": -33.29723358154297,
|
| 1676 |
+
"loss": 1.1095,
|
| 1677 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1678 |
+
"rewards/chosen": 0.2660633623600006,
|
| 1679 |
+
"rewards/margins": 0.3859363794326782,
|
| 1680 |
+
"rewards/rejected": -0.11987300962209702,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -2.0510499477386475,
|
| 1687 |
+
"logits/rejected": -2.052177906036377,
|
| 1688 |
+
"logps/chosen": -31.77374839782715,
|
| 1689 |
+
"logps/rejected": -34.914920806884766,
|
| 1690 |
+
"loss": 1.054,
|
| 1691 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1692 |
+
"rewards/chosen": 0.2567782998085022,
|
| 1693 |
+
"rewards/margins": 0.4166257381439209,
|
| 1694 |
+
"rewards/rejected": -0.15984737873077393,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -2.019425868988037,
|
| 1701 |
+
"logits/rejected": -2.023533344268799,
|
| 1702 |
+
"logps/chosen": -29.67460060119629,
|
| 1703 |
+
"logps/rejected": -34.19086456298828,
|
| 1704 |
+
"loss": 1.3668,
|
| 1705 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1706 |
+
"rewards/chosen": 0.2437433898448944,
|
| 1707 |
+
"rewards/margins": 0.37235018610954285,
|
| 1708 |
+
"rewards/rejected": -0.12860681116580963,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -2.0084166526794434,
|
| 1715 |
+
"logits/rejected": -2.0137572288513184,
|
| 1716 |
+
"logps/chosen": -32.371253967285156,
|
| 1717 |
+
"logps/rejected": -30.9876651763916,
|
| 1718 |
+
"loss": 1.4312,
|
| 1719 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1720 |
+
"rewards/chosen": 0.23955067992210388,
|
| 1721 |
+
"rewards/margins": 0.35674941539764404,
|
| 1722 |
+
"rewards/rejected": -0.11719872802495956,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.2037713527679443,
|
| 1728 |
+
"eval_logits/rejected": -2.1989424228668213,
|
| 1729 |
+
"eval_logps/chosen": -34.106727600097656,
|
| 1730 |
+
"eval_logps/rejected": -37.72097396850586,
|
| 1731 |
+
"eval_loss": 5.990525722503662,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5859634280204773,
|
| 1733 |
+
"eval_rewards/chosen": -0.01443566381931305,
|
| 1734 |
+
"eval_rewards/margins": 0.02643514797091484,
|
| 1735 |
+
"eval_rewards/rejected": -0.04087081179022789,
|
| 1736 |
+
"eval_runtime": 145.9163,
|
| 1737 |
+
"eval_samples_per_second": 2.351,
|
| 1738 |
+
"eval_steps_per_second": 0.295,
|
| 1739 |
+
"step": 1100
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 2.88,
|
| 1743 |
+
"learning_rate": 1.0963854773524548e-06,
|
| 1744 |
+
"logits/chosen": -2.000826835632324,
|
| 1745 |
+
"logits/rejected": -2.000603199005127,
|
| 1746 |
+
"logps/chosen": -31.66360092163086,
|
| 1747 |
+
"logps/rejected": -31.520954132080078,
|
| 1748 |
+
"loss": 1.1158,
|
| 1749 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1750 |
+
"rewards/chosen": 0.2553100287914276,
|
| 1751 |
+
"rewards/margins": 0.3563171923160553,
|
| 1752 |
+
"rewards/rejected": -0.1010071411728859,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.8451296091079712,
|
| 1759 |
+
"logits/rejected": -1.8467060327529907,
|
| 1760 |
+
"logps/chosen": -33.390525817871094,
|
| 1761 |
+
"logps/rejected": -33.623374938964844,
|
| 1762 |
+
"loss": 1.7743,
|
| 1763 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1764 |
+
"rewards/chosen": 0.2348897010087967,
|
| 1765 |
+
"rewards/margins": 0.3513115346431732,
|
| 1766 |
+
"rewards/rejected": -0.11642180383205414,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.884866714477539,
|
| 1773 |
+
"logits/rejected": -1.882128357887268,
|
| 1774 |
+
"logps/chosen": -29.975570678710938,
|
| 1775 |
+
"logps/rejected": -31.670913696289062,
|
| 1776 |
+
"loss": 2.1098,
|
| 1777 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1778 |
+
"rewards/chosen": 0.20722976326942444,
|
| 1779 |
+
"rewards/margins": 0.3082646131515503,
|
| 1780 |
+
"rewards/rejected": -0.10103483498096466,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -2.0206472873687744,
|
| 1787 |
+
"logits/rejected": -2.0149998664855957,
|
| 1788 |
+
"logps/chosen": -31.671985626220703,
|
| 1789 |
+
"logps/rejected": -33.523948669433594,
|
| 1790 |
+
"loss": 1.0808,
|
| 1791 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1792 |
+
"rewards/chosen": 0.27356189489364624,
|
| 1793 |
+
"rewards/margins": 0.3875501751899719,
|
| 1794 |
+
"rewards/rejected": -0.11398820579051971,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.8714290857315063,
|
| 1801 |
+
"logits/rejected": -1.8613548278808594,
|
| 1802 |
+
"logps/chosen": -32.308326721191406,
|
| 1803 |
+
"logps/rejected": -31.882171630859375,
|
| 1804 |
+
"loss": 1.3498,
|
| 1805 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1806 |
+
"rewards/chosen": 0.22868108749389648,
|
| 1807 |
+
"rewards/margins": 0.34663525223731995,
|
| 1808 |
+
"rewards/rejected": -0.11795412003993988,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.9497884511947632,
|
| 1815 |
+
"logits/rejected": -1.9496195316314697,
|
| 1816 |
+
"logps/chosen": -32.30144119262695,
|
| 1817 |
+
"logps/rejected": -31.75592613220215,
|
| 1818 |
+
"loss": 0.782,
|
| 1819 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1820 |
+
"rewards/chosen": 0.2760619819164276,
|
| 1821 |
+
"rewards/margins": 0.3869572877883911,
|
| 1822 |
+
"rewards/rejected": -0.1108952984213829,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -2.022653818130493,
|
| 1829 |
+
"logits/rejected": -2.017073154449463,
|
| 1830 |
+
"logps/chosen": -32.25920867919922,
|
| 1831 |
+
"logps/rejected": -33.56011962890625,
|
| 1832 |
+
"loss": 0.9637,
|
| 1833 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1834 |
+
"rewards/chosen": 0.25335273146629333,
|
| 1835 |
+
"rewards/margins": 0.37713244557380676,
|
| 1836 |
+
"rewards/rejected": -0.12377973645925522,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.8782762289047241,
|
| 1843 |
+
"logits/rejected": -1.8860442638397217,
|
| 1844 |
+
"logps/chosen": -32.099754333496094,
|
| 1845 |
+
"logps/rejected": -33.019081115722656,
|
| 1846 |
+
"loss": 0.8431,
|
| 1847 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1848 |
+
"rewards/chosen": 0.24664826691150665,
|
| 1849 |
+
"rewards/margins": 0.3781769871711731,
|
| 1850 |
+
"rewards/rejected": -0.13152867555618286,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -1.954120397567749,
|
| 1857 |
+
"logits/rejected": -1.9483168125152588,
|
| 1858 |
+
"logps/chosen": -30.81561851501465,
|
| 1859 |
+
"logps/rejected": -33.87861251831055,
|
| 1860 |
+
"loss": 0.9223,
|
| 1861 |
+
"rewards/accuracies": 0.9375,
|
| 1862 |
+
"rewards/chosen": 0.2715989053249359,
|
| 1863 |
+
"rewards/margins": 0.3906315267086029,
|
| 1864 |
+
"rewards/rejected": -0.11903263628482819,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.893996000289917,
|
| 1871 |
+
"logits/rejected": -1.8961610794067383,
|
| 1872 |
+
"logps/chosen": -32.268707275390625,
|
| 1873 |
+
"logps/rejected": -33.56250762939453,
|
| 1874 |
+
"loss": 1.0133,
|
| 1875 |
+
"rewards/accuracies": 0.9375,
|
| 1876 |
+
"rewards/chosen": 0.24434716999530792,
|
| 1877 |
+
"rewards/margins": 0.4048388600349426,
|
| 1878 |
+
"rewards/rejected": -0.1604916751384735,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.204813241958618,
|
| 1884 |
+
"eval_logits/rejected": -2.1999785900115967,
|
| 1885 |
+
"eval_logps/chosen": -34.118247985839844,
|
| 1886 |
+
"eval_logps/rejected": -37.71463394165039,
|
| 1887 |
+
"eval_loss": 6.110255241394043,
|
| 1888 |
+
"eval_rewards/accuracies": 0.5888704061508179,
|
| 1889 |
+
"eval_rewards/chosen": -0.016739506274461746,
|
| 1890 |
+
"eval_rewards/margins": 0.022863931953907013,
|
| 1891 |
+
"eval_rewards/rejected": -0.03960343450307846,
|
| 1892 |
+
"eval_runtime": 145.9534,
|
| 1893 |
+
"eval_samples_per_second": 2.35,
|
| 1894 |
+
"eval_steps_per_second": 0.295,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 3.14,
|
| 1899 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1900 |
+
"logits/chosen": -1.8266493082046509,
|
| 1901 |
+
"logits/rejected": -1.8255360126495361,
|
| 1902 |
+
"logps/chosen": -34.7088623046875,
|
| 1903 |
+
"logps/rejected": -33.4069709777832,
|
| 1904 |
+
"loss": 0.8168,
|
| 1905 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1906 |
+
"rewards/chosen": 0.268724262714386,
|
| 1907 |
+
"rewards/margins": 0.40824633836746216,
|
| 1908 |
+
"rewards/rejected": -0.13952204585075378,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -2.001779317855835,
|
| 1915 |
+
"logits/rejected": -1.9957544803619385,
|
| 1916 |
+
"logps/chosen": -30.487533569335938,
|
| 1917 |
+
"logps/rejected": -33.67655563354492,
|
| 1918 |
+
"loss": 0.7601,
|
| 1919 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1920 |
+
"rewards/chosen": 0.24323053658008575,
|
| 1921 |
+
"rewards/margins": 0.40283313393592834,
|
| 1922 |
+
"rewards/rejected": -0.15960261225700378,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.850182294845581,
|
| 1929 |
+
"logits/rejected": -1.8384950160980225,
|
| 1930 |
+
"logps/chosen": -33.03240203857422,
|
| 1931 |
+
"logps/rejected": -33.570960998535156,
|
| 1932 |
+
"loss": 0.8285,
|
| 1933 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1934 |
+
"rewards/chosen": 0.2706558108329773,
|
| 1935 |
+
"rewards/margins": 0.4060825705528259,
|
| 1936 |
+
"rewards/rejected": -0.13542678952217102,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.94095778465271,
|
| 1943 |
+
"logits/rejected": -1.934069275856018,
|
| 1944 |
+
"logps/chosen": -29.424108505249023,
|
| 1945 |
+
"logps/rejected": -29.53556251525879,
|
| 1946 |
+
"loss": 0.8909,
|
| 1947 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1948 |
+
"rewards/chosen": 0.25522512197494507,
|
| 1949 |
+
"rewards/margins": 0.38882559537887573,
|
| 1950 |
+
"rewards/rejected": -0.13360042870044708,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -1.9732204675674438,
|
| 1957 |
+
"logits/rejected": -1.97970449924469,
|
| 1958 |
+
"logps/chosen": -33.94004440307617,
|
| 1959 |
+
"logps/rejected": -32.08849334716797,
|
| 1960 |
+
"loss": 0.8667,
|
| 1961 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1962 |
+
"rewards/chosen": 0.29106852412223816,
|
| 1963 |
+
"rewards/margins": 0.4051054120063782,
|
| 1964 |
+
"rewards/rejected": -0.11403689533472061,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.9436290264129639,
|
| 1971 |
+
"logits/rejected": -1.944186806678772,
|
| 1972 |
+
"logps/chosen": -31.774776458740234,
|
| 1973 |
+
"logps/rejected": -33.607566833496094,
|
| 1974 |
+
"loss": 0.7227,
|
| 1975 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1976 |
+
"rewards/chosen": 0.25075581669807434,
|
| 1977 |
+
"rewards/margins": 0.40929579734802246,
|
| 1978 |
+
"rewards/rejected": -0.15853998064994812,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.8878374099731445,
|
| 1985 |
+
"logits/rejected": -1.887284517288208,
|
| 1986 |
+
"logps/chosen": -31.73285484313965,
|
| 1987 |
+
"logps/rejected": -34.178916931152344,
|
| 1988 |
+
"loss": 0.8078,
|
| 1989 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1990 |
+
"rewards/chosen": 0.2715964913368225,
|
| 1991 |
+
"rewards/margins": 0.4032578468322754,
|
| 1992 |
+
"rewards/rejected": -0.13166138529777527,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.7711877822875977,
|
| 1999 |
+
"logits/rejected": -1.7664165496826172,
|
| 2000 |
+
"logps/chosen": -33.83073425292969,
|
| 2001 |
+
"logps/rejected": -31.557464599609375,
|
| 2002 |
+
"loss": 0.7326,
|
| 2003 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2004 |
+
"rewards/chosen": 0.28134211897850037,
|
| 2005 |
+
"rewards/margins": 0.407537043094635,
|
| 2006 |
+
"rewards/rejected": -0.12619495391845703,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -1.9766957759857178,
|
| 2013 |
+
"logits/rejected": -1.96097731590271,
|
| 2014 |
+
"logps/chosen": -33.02162551879883,
|
| 2015 |
+
"logps/rejected": -35.638221740722656,
|
| 2016 |
+
"loss": 0.6214,
|
| 2017 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2018 |
+
"rewards/chosen": 0.26518964767456055,
|
| 2019 |
+
"rewards/margins": 0.4397699236869812,
|
| 2020 |
+
"rewards/rejected": -0.17458033561706543,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.9200379848480225,
|
| 2027 |
+
"logits/rejected": -1.9249954223632812,
|
| 2028 |
+
"logps/chosen": -32.194488525390625,
|
| 2029 |
+
"logps/rejected": -36.9532356262207,
|
| 2030 |
+
"loss": 0.5152,
|
| 2031 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2032 |
+
"rewards/chosen": 0.26462727785110474,
|
| 2033 |
+
"rewards/margins": 0.4346303343772888,
|
| 2034 |
+
"rewards/rejected": -0.17000307142734528,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.2051475048065186,
|
| 2040 |
+
"eval_logits/rejected": -2.200305461883545,
|
| 2041 |
+
"eval_logps/chosen": -34.10041427612305,
|
| 2042 |
+
"eval_logps/rejected": -37.70795822143555,
|
| 2043 |
+
"eval_loss": 6.057835102081299,
|
| 2044 |
+
"eval_rewards/accuracies": 0.5544019937515259,
|
| 2045 |
+
"eval_rewards/chosen": -0.013172022067010403,
|
| 2046 |
+
"eval_rewards/margins": 0.025095123797655106,
|
| 2047 |
+
"eval_rewards/rejected": -0.038267143070697784,
|
| 2048 |
+
"eval_runtime": 145.9616,
|
| 2049 |
+
"eval_samples_per_second": 2.35,
|
| 2050 |
+
"eval_steps_per_second": 0.295,
|
| 2051 |
+
"step": 1300
|
| 2052 |
+
},
|
| 2053 |
+
{
|
| 2054 |
+
"epoch": 3.4,
|
| 2055 |
+
"learning_rate": 3.321087280364757e-07,
|
| 2056 |
+
"logits/chosen": -1.8912603855133057,
|
| 2057 |
+
"logits/rejected": -1.8913657665252686,
|
| 2058 |
+
"logps/chosen": -34.96311950683594,
|
| 2059 |
+
"logps/rejected": -37.694114685058594,
|
| 2060 |
+
"loss": 1.0811,
|
| 2061 |
+
"rewards/accuracies": 0.9375,
|
| 2062 |
+
"rewards/chosen": 0.2561797797679901,
|
| 2063 |
+
"rewards/margins": 0.37817925214767456,
|
| 2064 |
+
"rewards/rejected": -0.12199944257736206,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -1.9785616397857666,
|
| 2071 |
+
"logits/rejected": -1.9770876169204712,
|
| 2072 |
+
"logps/chosen": -30.837026596069336,
|
| 2073 |
+
"logps/rejected": -31.65818214416504,
|
| 2074 |
+
"loss": 0.8959,
|
| 2075 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2076 |
+
"rewards/chosen": 0.25136545300483704,
|
| 2077 |
+
"rewards/margins": 0.39232978224754333,
|
| 2078 |
+
"rewards/rejected": -0.1409643441438675,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.882939100265503,
|
| 2085 |
+
"logits/rejected": -1.884426474571228,
|
| 2086 |
+
"logps/chosen": -33.03701400756836,
|
| 2087 |
+
"logps/rejected": -34.348941802978516,
|
| 2088 |
+
"loss": 0.6463,
|
| 2089 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2090 |
+
"rewards/chosen": 0.26744818687438965,
|
| 2091 |
+
"rewards/margins": 0.4075010418891907,
|
| 2092 |
+
"rewards/rejected": -0.14005282521247864,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.9446265697479248,
|
| 2099 |
+
"logits/rejected": -1.9434131383895874,
|
| 2100 |
+
"logps/chosen": -29.983739852905273,
|
| 2101 |
+
"logps/rejected": -32.796180725097656,
|
| 2102 |
+
"loss": 0.8896,
|
| 2103 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2104 |
+
"rewards/chosen": 0.24298477172851562,
|
| 2105 |
+
"rewards/margins": 0.4038262367248535,
|
| 2106 |
+
"rewards/rejected": -0.1608414649963379,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -1.9611021280288696,
|
| 2113 |
+
"logits/rejected": -1.9463632106781006,
|
| 2114 |
+
"logps/chosen": -33.147804260253906,
|
| 2115 |
+
"logps/rejected": -36.96843338012695,
|
| 2116 |
+
"loss": 0.8124,
|
| 2117 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2118 |
+
"rewards/chosen": 0.23377390205860138,
|
| 2119 |
+
"rewards/margins": 0.37457385659217834,
|
| 2120 |
+
"rewards/rejected": -0.14079990983009338,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -1.9647136926651,
|
| 2127 |
+
"logits/rejected": -1.9772529602050781,
|
| 2128 |
+
"logps/chosen": -30.748397827148438,
|
| 2129 |
+
"logps/rejected": -32.809547424316406,
|
| 2130 |
+
"loss": 1.0166,
|
| 2131 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2132 |
+
"rewards/chosen": 0.2640145421028137,
|
| 2133 |
+
"rewards/margins": 0.37797611951828003,
|
| 2134 |
+
"rewards/rejected": -0.1139616146683693,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -2.0538225173950195,
|
| 2141 |
+
"logits/rejected": -2.0530457496643066,
|
| 2142 |
+
"logps/chosen": -30.740360260009766,
|
| 2143 |
+
"logps/rejected": -36.674530029296875,
|
| 2144 |
+
"loss": 0.5522,
|
| 2145 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2146 |
+
"rewards/chosen": 0.25801384449005127,
|
| 2147 |
+
"rewards/margins": 0.4267433285713196,
|
| 2148 |
+
"rewards/rejected": -0.16872946918010712,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.9202001094818115,
|
| 2155 |
+
"logits/rejected": -1.928109884262085,
|
| 2156 |
+
"logps/chosen": -32.956661224365234,
|
| 2157 |
+
"logps/rejected": -38.160308837890625,
|
| 2158 |
+
"loss": 1.16,
|
| 2159 |
+
"rewards/accuracies": 0.9375,
|
| 2160 |
+
"rewards/chosen": 0.24476349353790283,
|
| 2161 |
+
"rewards/margins": 0.3868965208530426,
|
| 2162 |
+
"rewards/rejected": -0.14213302731513977,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.8742387294769287,
|
| 2169 |
+
"logits/rejected": -1.8700668811798096,
|
| 2170 |
+
"logps/chosen": -33.33854675292969,
|
| 2171 |
+
"logps/rejected": -32.12687301635742,
|
| 2172 |
+
"loss": 0.9939,
|
| 2173 |
+
"rewards/accuracies": 0.9375,
|
| 2174 |
+
"rewards/chosen": 0.28573352098464966,
|
| 2175 |
+
"rewards/margins": 0.39162954688072205,
|
| 2176 |
+
"rewards/rejected": -0.10589603334665298,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -2.037750720977783,
|
| 2183 |
+
"logits/rejected": -2.0352249145507812,
|
| 2184 |
+
"logps/chosen": -30.064016342163086,
|
| 2185 |
+
"logps/rejected": -32.856353759765625,
|
| 2186 |
+
"loss": 0.8378,
|
| 2187 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2188 |
+
"rewards/chosen": 0.24650637805461884,
|
| 2189 |
+
"rewards/margins": 0.38888365030288696,
|
| 2190 |
+
"rewards/rejected": -0.14237727224826813,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.2052032947540283,
|
| 2196 |
+
"eval_logits/rejected": -2.2003655433654785,
|
| 2197 |
+
"eval_logps/chosen": -34.10353088378906,
|
| 2198 |
+
"eval_logps/rejected": -37.7112922668457,
|
| 2199 |
+
"eval_loss": 6.057151794433594,
|
| 2200 |
+
"eval_rewards/accuracies": 0.5747508406639099,
|
| 2201 |
+
"eval_rewards/chosen": -0.013795554637908936,
|
| 2202 |
+
"eval_rewards/margins": 0.025138981640338898,
|
| 2203 |
+
"eval_rewards/rejected": -0.03893453627824783,
|
| 2204 |
+
"eval_runtime": 145.9658,
|
| 2205 |
+
"eval_samples_per_second": 2.35,
|
| 2206 |
+
"eval_steps_per_second": 0.295,
|
| 2207 |
+
"step": 1400
|
| 2208 |
+
},
|
| 2209 |
+
{
|
| 2210 |
+
"epoch": 3.66,
|
| 2211 |
+
"learning_rate": 1.0775199359171346e-07,
|
| 2212 |
+
"logits/chosen": -1.9774906635284424,
|
| 2213 |
+
"logits/rejected": -1.9731305837631226,
|
| 2214 |
+
"logps/chosen": -32.331031799316406,
|
| 2215 |
+
"logps/rejected": -30.20021629333496,
|
| 2216 |
+
"loss": 0.8015,
|
| 2217 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2218 |
+
"rewards/chosen": 0.26229986548423767,
|
| 2219 |
+
"rewards/margins": 0.39145129919052124,
|
| 2220 |
+
"rewards/rejected": -0.1291513890028,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.935434341430664,
|
| 2227 |
+
"logits/rejected": -1.936626672744751,
|
| 2228 |
+
"logps/chosen": -32.0463981628418,
|
| 2229 |
+
"logps/rejected": -35.533714294433594,
|
| 2230 |
+
"loss": 1.0098,
|
| 2231 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2232 |
+
"rewards/chosen": 0.2636954188346863,
|
| 2233 |
+
"rewards/margins": 0.38563698530197144,
|
| 2234 |
+
"rewards/rejected": -0.12194149196147919,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.8892452716827393,
|
| 2241 |
+
"logits/rejected": -1.8824679851531982,
|
| 2242 |
+
"logps/chosen": -32.84967803955078,
|
| 2243 |
+
"logps/rejected": -37.3257942199707,
|
| 2244 |
+
"loss": 0.883,
|
| 2245 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2246 |
+
"rewards/chosen": 0.2584185004234314,
|
| 2247 |
+
"rewards/margins": 0.3856961131095886,
|
| 2248 |
+
"rewards/rejected": -0.12727761268615723,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.9004735946655273,
|
| 2255 |
+
"logits/rejected": -1.8949673175811768,
|
| 2256 |
+
"logps/chosen": -28.052719116210938,
|
| 2257 |
+
"logps/rejected": -32.38301467895508,
|
| 2258 |
+
"loss": 1.1273,
|
| 2259 |
+
"rewards/accuracies": 0.9375,
|
| 2260 |
+
"rewards/chosen": 0.25210484862327576,
|
| 2261 |
+
"rewards/margins": 0.38257673382759094,
|
| 2262 |
+
"rewards/rejected": -0.13047190010547638,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.9209520816802979,
|
| 2269 |
+
"logits/rejected": -1.910509705543518,
|
| 2270 |
+
"logps/chosen": -31.346166610717773,
|
| 2271 |
+
"logps/rejected": -36.17350387573242,
|
| 2272 |
+
"loss": 0.819,
|
| 2273 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2274 |
+
"rewards/chosen": 0.22707252204418182,
|
| 2275 |
+
"rewards/margins": 0.4047010540962219,
|
| 2276 |
+
"rewards/rejected": -0.17762848734855652,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.8970378637313843,
|
| 2283 |
+
"logits/rejected": -1.8941428661346436,
|
| 2284 |
+
"logps/chosen": -32.03502655029297,
|
| 2285 |
+
"logps/rejected": -31.428430557250977,
|
| 2286 |
+
"loss": 1.1237,
|
| 2287 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2288 |
+
"rewards/chosen": 0.25494998693466187,
|
| 2289 |
+
"rewards/margins": 0.36247119307518005,
|
| 2290 |
+
"rewards/rejected": -0.1075211763381958,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -1.9586317539215088,
|
| 2297 |
+
"logits/rejected": -1.9652128219604492,
|
| 2298 |
+
"logps/chosen": -30.69277000427246,
|
| 2299 |
+
"logps/rejected": -33.610328674316406,
|
| 2300 |
+
"loss": 0.8233,
|
| 2301 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2302 |
+
"rewards/chosen": 0.2752807140350342,
|
| 2303 |
+
"rewards/margins": 0.4048624038696289,
|
| 2304 |
+
"rewards/rejected": -0.12958170473575592,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -1.9582951068878174,
|
| 2311 |
+
"logits/rejected": -1.9687860012054443,
|
| 2312 |
+
"logps/chosen": -30.41074562072754,
|
| 2313 |
+
"logps/rejected": -30.278295516967773,
|
| 2314 |
+
"loss": 0.8239,
|
| 2315 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2316 |
+
"rewards/chosen": 0.25690194964408875,
|
| 2317 |
+
"rewards/margins": 0.38684824109077454,
|
| 2318 |
+
"rewards/rejected": -0.1299462616443634,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.8908532857894897,
|
| 2325 |
+
"logits/rejected": -1.8837330341339111,
|
| 2326 |
+
"logps/chosen": -32.43671417236328,
|
| 2327 |
+
"logps/rejected": -32.93724822998047,
|
| 2328 |
+
"loss": 0.9657,
|
| 2329 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2330 |
+
"rewards/chosen": 0.24395498633384705,
|
| 2331 |
+
"rewards/margins": 0.38930970430374146,
|
| 2332 |
+
"rewards/rejected": -0.14535477757453918,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.943483591079712,
|
| 2339 |
+
"logits/rejected": -1.940323829650879,
|
| 2340 |
+
"logps/chosen": -32.4309196472168,
|
| 2341 |
+
"logps/rejected": -34.01084518432617,
|
| 2342 |
+
"loss": 0.9599,
|
| 2343 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2344 |
+
"rewards/chosen": 0.2389308512210846,
|
| 2345 |
+
"rewards/margins": 0.38905516266822815,
|
| 2346 |
+
"rewards/rejected": -0.15012428164482117,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.2052109241485596,
|
| 2352 |
+
"eval_logits/rejected": -2.2003753185272217,
|
| 2353 |
+
"eval_logps/chosen": -34.09724044799805,
|
| 2354 |
+
"eval_logps/rejected": -37.70912551879883,
|
| 2355 |
+
"eval_loss": 6.034818649291992,
|
| 2356 |
+
"eval_rewards/accuracies": 0.5834717750549316,
|
| 2357 |
+
"eval_rewards/chosen": -0.012537354603409767,
|
| 2358 |
+
"eval_rewards/margins": 0.025963816791772842,
|
| 2359 |
+
"eval_rewards/rejected": -0.03850117325782776,
|
| 2360 |
+
"eval_runtime": 145.968,
|
| 2361 |
+
"eval_samples_per_second": 2.35,
|
| 2362 |
+
"eval_steps_per_second": 0.295,
|
| 2363 |
+
"step": 1500
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 3.92,
|
| 2367 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2368 |
+
"logits/chosen": -2.0217642784118652,
|
| 2369 |
+
"logits/rejected": -2.022365093231201,
|
| 2370 |
+
"logps/chosen": -28.69793701171875,
|
| 2371 |
+
"logps/rejected": -33.473472595214844,
|
| 2372 |
+
"loss": 1.3374,
|
| 2373 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2374 |
+
"rewards/chosen": 0.22366802394390106,
|
| 2375 |
+
"rewards/margins": 0.37027162313461304,
|
| 2376 |
+
"rewards/rejected": -0.14660362899303436,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.8759753704071045,
|
| 2383 |
+
"logits/rejected": -1.8758065700531006,
|
| 2384 |
+
"logps/chosen": -31.9100341796875,
|
| 2385 |
+
"logps/rejected": -34.565345764160156,
|
| 2386 |
+
"loss": 1.196,
|
| 2387 |
+
"rewards/accuracies": 0.9375,
|
| 2388 |
+
"rewards/chosen": 0.22636690735816956,
|
| 2389 |
+
"rewards/margins": 0.3646892309188843,
|
| 2390 |
+
"rewards/rejected": -0.13832230865955353,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -2.0270986557006836,
|
| 2397 |
+
"logits/rejected": -2.024442195892334,
|
| 2398 |
+
"logps/chosen": -30.32328224182129,
|
| 2399 |
+
"logps/rejected": -32.253662109375,
|
| 2400 |
+
"loss": 0.9082,
|
| 2401 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2402 |
+
"rewards/chosen": 0.23015546798706055,
|
| 2403 |
+
"rewards/margins": 0.384157657623291,
|
| 2404 |
+
"rewards/rejected": -0.15400215983390808,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -2.0110628604888916,
|
| 2411 |
+
"logits/rejected": -2.0121970176696777,
|
| 2412 |
+
"logps/chosen": -29.453231811523438,
|
| 2413 |
+
"logps/rejected": -30.480737686157227,
|
| 2414 |
+
"loss": 0.9153,
|
| 2415 |
+
"rewards/accuracies": 0.98333340883255,
|
| 2416 |
+
"rewards/chosen": 0.23334148526191711,
|
| 2417 |
+
"rewards/margins": 0.37145254015922546,
|
| 2418 |
+
"rewards/rejected": -0.13811106979846954,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 1.3887401007986688,
|
| 2426 |
+
"train_runtime": 10806.9485,
|
| 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,
|