Commit
•
d95588a
1
Parent(s):
5fffe7c
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 +1905 -89
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
| 4 |
-
- alignment-handbook
|
| 5 |
- trl
|
| 6 |
- dpo
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
| 9 |
-
datasets:
|
| 10 |
-
- hugodk-sch/aftonposten_title_prefs
|
| 11 |
model-index:
|
| 12 |
- name: aftonposten-6b-align-scan
|
| 13 |
results: []
|
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
# aftonposten-6b-align-scan
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 0.
|
| 24 |
-
- Rewards/chosen:
|
| 25 |
-
- Rewards/rejected:
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,15 +55,27 @@ The following hyperparameters were used during training:
|
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
| 59 |
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
-
- num_epochs:
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
-
| Training Loss | Epoch | Step |
|
| 65 |
-
|
| 66 |
-
| 0.5903 | 0.26 | 100 |
|
| 67 |
-
| 0.59 | 0.52 | 200 |
|
| 68 |
-
| 0.3871 | 0.78 | 300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
|
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
|
|
|
| 4 |
- trl
|
| 5 |
- dpo
|
| 6 |
+
- alignment-handbook
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
|
|
|
|
|
|
| 9 |
model-index:
|
| 10 |
- name: aftonposten-6b-align-scan
|
| 11 |
results: []
|
|
|
|
| 16 |
|
| 17 |
# aftonposten-6b-align-scan
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on an unknown dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.7352
|
| 22 |
+
- Rewards/chosen: 0.0537
|
| 23 |
+
- Rewards/rejected: 0.0336
|
| 24 |
+
- Rewards/accuracies: 0.5399
|
| 25 |
+
- Rewards/margins: 0.0201
|
| 26 |
+
- Logps/rejected: -37.4606
|
| 27 |
+
- Logps/chosen: -33.9451
|
| 28 |
+
- Logits/rejected: -2.2407
|
| 29 |
+
- Logits/chosen: -2.2456
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 55 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
- lr_scheduler_type: cosine
|
| 57 |
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 4
|
| 59 |
|
| 60 |
### Training results
|
| 61 |
|
| 62 |
+
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
| 63 |
+
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
| 64 |
+
| 0.5903 | 0.26 | 100 | -2.2366 | -2.2318 | -34.0141 | -37.5087 | 0.7450 | 0.5241 | 0.0123 | 0.0075 | 0.0047 |
|
| 65 |
+
| 0.59 | 0.52 | 200 | -2.2332 | -2.2284 | -34.0691 | -37.5752 | 0.7500 | 0.5365 | -0.0208 | 0.0144 | -0.0352 |
|
| 66 |
+
| 0.3871 | 0.78 | 300 | -2.2333 | -2.2285 | -34.0622 | -37.5657 | 0.7655 | 0.5071 | -0.0166 | 0.0129 | -0.0295 |
|
| 67 |
+
| 0.3315 | 1.04 | 400 | 0.7479 | -0.0009 | -0.0296 | 0.5515 | 0.0287 | -37.5660 | -34.0361 | -2.2265 | -2.2314 |
|
| 68 |
+
| 0.3857 | 1.3 | 500 | 0.7601 | -0.0010 | -0.0248 | 0.5577 | 0.0238 | -37.5580 | -34.0362 | -2.2294 | -2.2343 |
|
| 69 |
+
| 0.5531 | 1.56 | 600 | 0.7435 | 0.0040 | -0.0183 | 0.5577 | 0.0223 | -37.5471 | -34.0279 | -2.2416 | -2.2465 |
|
| 70 |
+
| 0.4315 | 1.82 | 700 | 0.8141 | -0.0077 | -0.0028 | 0.4996 | -0.0049 | -37.5213 | -34.0474 | -2.2449 | -2.2498 |
|
| 71 |
+
| 0.3445 | 2.08 | 800 | 0.8104 | 0.0382 | 0.0369 | 0.5282 | 0.0012 | -37.4551 | -33.9710 | -2.2437 | -2.2485 |
|
| 72 |
+
| 0.1902 | 2.34 | 900 | 0.8123 | 0.0329 | 0.0364 | 0.5199 | -0.0035 | -37.4560 | -33.9797 | -2.2426 | -2.2475 |
|
| 73 |
+
| 0.1724 | 2.6 | 1000 | 0.7850 | 0.0479 | 0.0385 | 0.5365 | 0.0094 | -37.4525 | -33.9548 | -2.2433 | -2.2481 |
|
| 74 |
+
| 0.1864 | 2.86 | 1100 | 0.7625 | 0.0435 | 0.0297 | 0.5133 | 0.0137 | -37.4671 | -33.9621 | -2.2392 | -2.2441 |
|
| 75 |
+
| 0.1629 | 3.12 | 1200 | 0.7624 | 0.0479 | 0.0337 | 0.5341 | 0.0142 | -37.4604 | -33.9547 | -2.2387 | -2.2436 |
|
| 76 |
+
| 0.1881 | 3.38 | 1300 | 0.7945 | 0.0454 | 0.0440 | 0.4967 | 0.0014 | -37.4432 | -33.9588 | -2.2391 | -2.2440 |
|
| 77 |
+
| 0.1395 | 3.64 | 1400 | 0.7520 | 0.0496 | 0.0341 | 0.5453 | 0.0156 | -37.4598 | -33.9518 | -2.2408 | -2.2457 |
|
| 78 |
+
| 0.1436 | 3.9 | 1500 | 0.7352 | 0.0537 | 0.0336 | 0.5399 | 0.0201 | -37.4606 | -33.9451 | -2.2407 | -2.2456 |
|
| 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:f6179b75e61780f60641e1691c7393bf4991bc2b020dc7ae011150d91bac1dbb
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.2331926822662354,
|
| 4 |
"eval_logits/rejected": -2.2283315658569336,
|
| 5 |
"eval_logps/chosen": -34.061973571777344,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.357,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
-
"train_loss": 0.
|
| 17 |
-
"train_runtime":
|
| 18 |
"train_samples": 3079,
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
"eval_logits/chosen": -2.2331926822662354,
|
| 4 |
"eval_logits/rejected": -2.2283315658569336,
|
| 5 |
"eval_logps/chosen": -34.061973571777344,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.357,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.2286405209984098,
|
| 17 |
+
"train_runtime": 10803.5829,
|
| 18 |
"train_samples": 3079,
|
| 19 |
+
"train_samples_per_second": 1.14,
|
| 20 |
+
"train_steps_per_second": 0.143
|
| 21 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 3079,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
+
"train_loss": 0.2286405209984098,
|
| 4 |
+
"train_runtime": 10803.5829,
|
| 5 |
"train_samples": 3079,
|
| 6 |
+
"train_samples_per_second": 1.14,
|
| 7 |
+
"train_steps_per_second": 0.143
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
@@ -492,130 +492,1946 @@
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
-
"learning_rate":
|
| 496 |
-
"logits/chosen": -1.
|
| 497 |
-
"logits/rejected": -1.
|
| 498 |
-
"logps/chosen": -31.
|
| 499 |
-
"logps/rejected": -33.
|
| 500 |
-
"loss": 0.
|
| 501 |
-
"rewards/accuracies": 0.
|
| 502 |
-
"rewards/chosen": 0.
|
| 503 |
-
"rewards/margins": 0.
|
| 504 |
-
"rewards/rejected": -0.
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
-
"learning_rate": 4.
|
| 510 |
-
"logits/chosen": -1.
|
| 511 |
-
"logits/rejected": -1.
|
| 512 |
-
"logps/chosen": -34.
|
| 513 |
-
"logps/rejected": -33.
|
| 514 |
-
"loss": 0.
|
| 515 |
-
"rewards/accuracies": 0.
|
| 516 |
-
"rewards/chosen": 0.
|
| 517 |
-
"rewards/margins": 0.
|
| 518 |
-
"rewards/rejected": -0.
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
-
"learning_rate":
|
| 524 |
-
"logits/chosen": -2.
|
| 525 |
-
"logits/rejected": -2.
|
| 526 |
-
"logps/chosen": -33.
|
| 527 |
-
"logps/rejected": -32.
|
| 528 |
-
"loss": 0.
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": 0.
|
| 531 |
-
"rewards/margins": 0.
|
| 532 |
-
"rewards/rejected": -0.
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -33.
|
| 541 |
-
"logps/rejected": -33.
|
| 542 |
-
"loss": 0.
|
| 543 |
-
"rewards/accuracies": 0.
|
| 544 |
-
"rewards/chosen": 0.
|
| 545 |
-
"rewards/margins": 0.
|
| 546 |
-
"rewards/rejected":
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
-
"learning_rate":
|
| 552 |
-
"logits/chosen": -1.
|
| 553 |
-
"logits/rejected": -1.
|
| 554 |
-
"logps/chosen": -32.
|
| 555 |
-
"logps/rejected": -32.
|
| 556 |
-
"loss": 0.
|
| 557 |
-
"rewards/accuracies": 0.
|
| 558 |
-
"rewards/chosen": 0.
|
| 559 |
-
"rewards/margins": 0.
|
| 560 |
-
"rewards/rejected": -0.
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
-
"learning_rate":
|
| 566 |
-
"logits/chosen": -1.
|
| 567 |
-
"logits/rejected": -1.
|
| 568 |
-
"logps/chosen": -31.
|
| 569 |
-
"logps/rejected": -35.
|
| 570 |
-
"loss": 0.
|
| 571 |
-
"rewards/accuracies": 0.
|
| 572 |
-
"rewards/chosen": 0.
|
| 573 |
-
"rewards/margins": 0.
|
| 574 |
-
"rewards/rejected": -0.
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
-
"learning_rate":
|
| 580 |
-
"logits/chosen": -2.
|
| 581 |
-
"logits/rejected": -2.
|
| 582 |
-
"logps/chosen": -33.
|
| 583 |
-
"logps/rejected": -29.
|
| 584 |
-
"loss": 0.
|
| 585 |
-
"rewards/accuracies": 0.
|
| 586 |
-
"rewards/chosen": 0.
|
| 587 |
-
"rewards/margins": 0.
|
| 588 |
-
"rewards/rejected": -0.
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
-
"learning_rate":
|
| 594 |
-
"logits/chosen": -1.
|
| 595 |
-
"logits/rejected": -1.
|
| 596 |
-
"logps/chosen": -33.
|
| 597 |
-
"logps/rejected": -30.
|
| 598 |
-
"loss": 0.
|
| 599 |
"rewards/accuracies": 0.7250000238418579,
|
| 600 |
-
"rewards/chosen": 0.
|
| 601 |
-
"rewards/margins": 0.
|
| 602 |
-
"rewards/rejected": -0.
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
-
"epoch": 1.
|
| 607 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 608 |
"total_flos": 0.0,
|
| 609 |
-
"train_loss": 0.
|
| 610 |
-
"train_runtime":
|
| 611 |
-
"train_samples_per_second":
|
| 612 |
-
"train_steps_per_second": 0.
|
| 613 |
}
|
| 614 |
],
|
| 615 |
"logging_steps": 10,
|
| 616 |
-
"max_steps":
|
| 617 |
"num_input_tokens_seen": 0,
|
| 618 |
-
"num_train_epochs":
|
| 619 |
"save_steps": 100,
|
| 620 |
"total_flos": 0.0,
|
| 621 |
"train_batch_size": 4,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 4.0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 1540,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
+
"learning_rate": 4.84533120650964e-06,
|
| 496 |
+
"logits/chosen": -1.9183489084243774,
|
| 497 |
+
"logits/rejected": -1.9150705337524414,
|
| 498 |
+
"logps/chosen": -31.377817153930664,
|
| 499 |
+
"logps/rejected": -33.78107833862305,
|
| 500 |
+
"loss": 0.573,
|
| 501 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 502 |
+
"rewards/chosen": 0.13532012701034546,
|
| 503 |
+
"rewards/margins": 0.1586771160364151,
|
| 504 |
+
"rewards/rejected": -0.02335697039961815,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.9682114124298096,
|
| 511 |
+
"logits/rejected": -1.9559390544891357,
|
| 512 |
+
"logps/chosen": -34.42866134643555,
|
| 513 |
+
"logps/rejected": -33.64933395385742,
|
| 514 |
+
"loss": 0.47,
|
| 515 |
+
"rewards/accuracies": 0.737500011920929,
|
| 516 |
+
"rewards/chosen": 0.10044805705547333,
|
| 517 |
+
"rewards/margins": 0.16674570739269257,
|
| 518 |
+
"rewards/rejected": -0.06629763543605804,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -2.004417896270752,
|
| 525 |
+
"logits/rejected": -2.0029640197753906,
|
| 526 |
+
"logps/chosen": -33.30437469482422,
|
| 527 |
+
"logps/rejected": -32.512184143066406,
|
| 528 |
+
"loss": 0.5649,
|
| 529 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 530 |
+
"rewards/chosen": 0.10884843021631241,
|
| 531 |
+
"rewards/margins": 0.12841013073921204,
|
| 532 |
+
"rewards/rejected": -0.019561681896448135,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.091501474380493,
|
| 539 |
+
"logits/rejected": -2.075732707977295,
|
| 540 |
+
"logps/chosen": -33.85096740722656,
|
| 541 |
+
"logps/rejected": -33.104610443115234,
|
| 542 |
+
"loss": 0.5076,
|
| 543 |
+
"rewards/accuracies": 0.699999988079071,
|
| 544 |
+
"rewards/chosen": 0.19705387949943542,
|
| 545 |
+
"rewards/margins": 0.18899771571159363,
|
| 546 |
+
"rewards/rejected": 0.008056174032390118,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.964172124862671,
|
| 553 |
+
"logits/rejected": -1.96323561668396,
|
| 554 |
+
"logps/chosen": -32.96615219116211,
|
| 555 |
+
"logps/rejected": -32.529842376708984,
|
| 556 |
+
"loss": 0.4871,
|
| 557 |
+
"rewards/accuracies": 0.737500011920929,
|
| 558 |
+
"rewards/chosen": 0.18956632912158966,
|
| 559 |
+
"rewards/margins": 0.2283860146999359,
|
| 560 |
+
"rewards/rejected": -0.03881971538066864,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.919790267944336,
|
| 567 |
+
"logits/rejected": -1.9300991296768188,
|
| 568 |
+
"logps/chosen": -31.94921875,
|
| 569 |
+
"logps/rejected": -35.33544921875,
|
| 570 |
+
"loss": 0.5992,
|
| 571 |
+
"rewards/accuracies": 0.737500011920929,
|
| 572 |
+
"rewards/chosen": 0.15328440070152283,
|
| 573 |
+
"rewards/margins": 0.18129965662956238,
|
| 574 |
+
"rewards/rejected": -0.028015267103910446,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.0574381351470947,
|
| 581 |
+
"logits/rejected": -2.050945520401001,
|
| 582 |
+
"logps/chosen": -33.3778076171875,
|
| 583 |
+
"logps/rejected": -29.26238441467285,
|
| 584 |
+
"loss": 0.4666,
|
| 585 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 586 |
+
"rewards/chosen": 0.16449405252933502,
|
| 587 |
+
"rewards/margins": 0.186069518327713,
|
| 588 |
+
"rewards/rejected": -0.021575476974248886,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.9166724681854248,
|
| 595 |
+
"logits/rejected": -1.9188873767852783,
|
| 596 |
+
"logps/chosen": -33.89116668701172,
|
| 597 |
+
"logps/rejected": -30.885635375976562,
|
| 598 |
+
"loss": 0.4244,
|
| 599 |
"rewards/accuracies": 0.7250000238418579,
|
| 600 |
+
"rewards/chosen": 0.2087288796901703,
|
| 601 |
+
"rewards/margins": 0.21653230488300323,
|
| 602 |
+
"rewards/rejected": -0.007803438697010279,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9459316730499268,
|
| 609 |
+
"logits/rejected": -1.9447450637817383,
|
| 610 |
+
"logps/chosen": -34.006561279296875,
|
| 611 |
+
"logps/rejected": -35.8227424621582,
|
| 612 |
+
"loss": 0.3945,
|
| 613 |
+
"rewards/accuracies": 0.7958333492279053,
|
| 614 |
+
"rewards/chosen": 0.22693590819835663,
|
| 615 |
+
"rewards/margins": 0.3269794285297394,
|
| 616 |
+
"rewards/rejected": -0.10004355013370514,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.8829723596572876,
|
| 623 |
+
"logits/rejected": -1.8746156692504883,
|
| 624 |
+
"logps/chosen": -31.377273559570312,
|
| 625 |
+
"logps/rejected": -36.17646408081055,
|
| 626 |
+
"loss": 0.3315,
|
| 627 |
+
"rewards/accuracies": 0.875,
|
| 628 |
+
"rewards/chosen": 0.23802152276039124,
|
| 629 |
+
"rewards/margins": 0.3795410692691803,
|
| 630 |
+
"rewards/rejected": -0.14151951670646667,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.231396436691284,
|
| 636 |
+
"eval_logits/rejected": -2.226522207260132,
|
| 637 |
+
"eval_logps/chosen": -34.036094665527344,
|
| 638 |
+
"eval_logps/rejected": -37.56599044799805,
|
| 639 |
+
"eval_loss": 0.7478801608085632,
|
| 640 |
+
"eval_rewards/accuracies": 0.5514950156211853,
|
| 641 |
+
"eval_rewards/chosen": -0.0009239073260687292,
|
| 642 |
+
"eval_rewards/margins": 0.02869875729084015,
|
| 643 |
+
"eval_rewards/rejected": -0.02962266281247139,
|
| 644 |
+
"eval_runtime": 146.4126,
|
| 645 |
+
"eval_samples_per_second": 2.343,
|
| 646 |
+
"eval_steps_per_second": 0.294,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 1.06,
|
| 651 |
+
"learning_rate": 4.590793060955158e-06,
|
| 652 |
+
"logits/chosen": -2.0595316886901855,
|
| 653 |
+
"logits/rejected": -2.062485933303833,
|
| 654 |
+
"logps/chosen": -32.621891021728516,
|
| 655 |
+
"logps/rejected": -35.11821365356445,
|
| 656 |
+
"loss": 0.5251,
|
| 657 |
+
"rewards/accuracies": 0.875,
|
| 658 |
+
"rewards/chosen": 0.2592288553714752,
|
| 659 |
+
"rewards/margins": 0.45444852113723755,
|
| 660 |
+
"rewards/rejected": -0.19521966576576233,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.903078317642212,
|
| 667 |
+
"logits/rejected": -1.9016958475112915,
|
| 668 |
+
"logps/chosen": -28.610631942749023,
|
| 669 |
+
"logps/rejected": -32.66521453857422,
|
| 670 |
+
"loss": 0.3322,
|
| 671 |
+
"rewards/accuracies": 0.949999988079071,
|
| 672 |
+
"rewards/chosen": 0.27415698766708374,
|
| 673 |
+
"rewards/margins": 0.48268675804138184,
|
| 674 |
+
"rewards/rejected": -0.2085297405719757,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.8610217571258545,
|
| 681 |
+
"logits/rejected": -1.854017972946167,
|
| 682 |
+
"logps/chosen": -33.55864334106445,
|
| 683 |
+
"logps/rejected": -34.378604888916016,
|
| 684 |
+
"loss": 0.3646,
|
| 685 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 686 |
+
"rewards/chosen": 0.3092653155326843,
|
| 687 |
+
"rewards/margins": 0.389264315366745,
|
| 688 |
+
"rewards/rejected": -0.07999899238348007,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -2.0365495681762695,
|
| 695 |
+
"logits/rejected": -2.0314345359802246,
|
| 696 |
+
"logps/chosen": -31.39179039001465,
|
| 697 |
+
"logps/rejected": -32.26957702636719,
|
| 698 |
+
"loss": 0.3048,
|
| 699 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 700 |
+
"rewards/chosen": 0.23490269482135773,
|
| 701 |
+
"rewards/margins": 0.41608113050460815,
|
| 702 |
+
"rewards/rejected": -0.18117846548557281,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.92501962184906,
|
| 709 |
+
"logits/rejected": -1.9292068481445312,
|
| 710 |
+
"logps/chosen": -34.05653381347656,
|
| 711 |
+
"logps/rejected": -32.91465759277344,
|
| 712 |
+
"loss": 0.551,
|
| 713 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 714 |
+
"rewards/chosen": 0.27558597922325134,
|
| 715 |
+
"rewards/margins": 0.4711950421333313,
|
| 716 |
+
"rewards/rejected": -0.19560904800891876,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.934442162513733,
|
| 723 |
+
"logits/rejected": -1.9286327362060547,
|
| 724 |
+
"logps/chosen": -34.91238784790039,
|
| 725 |
+
"logps/rejected": -35.231353759765625,
|
| 726 |
+
"loss": 0.3759,
|
| 727 |
+
"rewards/accuracies": 0.925000011920929,
|
| 728 |
+
"rewards/chosen": 0.3042834401130676,
|
| 729 |
+
"rewards/margins": 0.47363799810409546,
|
| 730 |
+
"rewards/rejected": -0.16935452818870544,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.9695565700531006,
|
| 737 |
+
"logits/rejected": -1.9690759181976318,
|
| 738 |
+
"logps/chosen": -33.69108581542969,
|
| 739 |
+
"logps/rejected": -34.42035675048828,
|
| 740 |
+
"loss": 0.2728,
|
| 741 |
+
"rewards/accuracies": 0.949999988079071,
|
| 742 |
+
"rewards/chosen": 0.2863459885120392,
|
| 743 |
+
"rewards/margins": 0.42727118730545044,
|
| 744 |
+
"rewards/rejected": -0.14092518389225006,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -2.0204994678497314,
|
| 751 |
+
"logits/rejected": -2.0016019344329834,
|
| 752 |
+
"logps/chosen": -33.737281799316406,
|
| 753 |
+
"logps/rejected": -33.34347152709961,
|
| 754 |
+
"loss": 0.4172,
|
| 755 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 756 |
+
"rewards/chosen": 0.2540892958641052,
|
| 757 |
+
"rewards/margins": 0.4181840419769287,
|
| 758 |
+
"rewards/rejected": -0.1640947312116623,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.9881887435913086,
|
| 765 |
+
"logits/rejected": -1.9876505136489868,
|
| 766 |
+
"logps/chosen": -31.16998291015625,
|
| 767 |
+
"logps/rejected": -31.431163787841797,
|
| 768 |
+
"loss": 0.315,
|
| 769 |
+
"rewards/accuracies": 0.887499988079071,
|
| 770 |
+
"rewards/chosen": 0.264793336391449,
|
| 771 |
+
"rewards/margins": 0.37586522102355957,
|
| 772 |
+
"rewards/rejected": -0.1110718846321106,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.9539533853530884,
|
| 779 |
+
"logits/rejected": -1.946997880935669,
|
| 780 |
+
"logps/chosen": -32.230567932128906,
|
| 781 |
+
"logps/rejected": -34.859554290771484,
|
| 782 |
+
"loss": 0.3857,
|
| 783 |
+
"rewards/accuracies": 0.887499988079071,
|
| 784 |
+
"rewards/chosen": 0.2637366056442261,
|
| 785 |
+
"rewards/margins": 0.4034242630004883,
|
| 786 |
+
"rewards/rejected": -0.139687642455101,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.2342920303344727,
|
| 792 |
+
"eval_logits/rejected": -2.2294206619262695,
|
| 793 |
+
"eval_logps/chosen": -34.03615951538086,
|
| 794 |
+
"eval_logps/rejected": -37.557979583740234,
|
| 795 |
+
"eval_loss": 0.7601029276847839,
|
| 796 |
+
"eval_rewards/accuracies": 0.5577242374420166,
|
| 797 |
+
"eval_rewards/chosen": -0.0009655931498855352,
|
| 798 |
+
"eval_rewards/margins": 0.023849941790103912,
|
| 799 |
+
"eval_rewards/rejected": -0.02481553703546524,
|
| 800 |
+
"eval_runtime": 145.9216,
|
| 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": -1.9524669647216797,
|
| 809 |
+
"logits/rejected": -1.9562795162200928,
|
| 810 |
+
"logps/chosen": -28.682544708251953,
|
| 811 |
+
"logps/rejected": -33.07499313354492,
|
| 812 |
+
"loss": 0.3645,
|
| 813 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 814 |
+
"rewards/chosen": 0.29536908864974976,
|
| 815 |
+
"rewards/margins": 0.42479196190834045,
|
| 816 |
+
"rewards/rejected": -0.12942291796207428,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.9641717672348022,
|
| 823 |
+
"logits/rejected": -1.9746614694595337,
|
| 824 |
+
"logps/chosen": -32.76649475097656,
|
| 825 |
+
"logps/rejected": -30.801523208618164,
|
| 826 |
+
"loss": 0.2418,
|
| 827 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 828 |
+
"rewards/chosen": 0.2975025177001953,
|
| 829 |
+
"rewards/margins": 0.3874046206474304,
|
| 830 |
+
"rewards/rejected": -0.0899021252989769,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.9106000661849976,
|
| 837 |
+
"logits/rejected": -1.908151626586914,
|
| 838 |
+
"logps/chosen": -31.15958023071289,
|
| 839 |
+
"logps/rejected": -30.377086639404297,
|
| 840 |
+
"loss": 0.419,
|
| 841 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 842 |
+
"rewards/chosen": 0.30450108647346497,
|
| 843 |
+
"rewards/margins": 0.42469558119773865,
|
| 844 |
+
"rewards/rejected": -0.12019453942775726,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.9983298778533936,
|
| 851 |
+
"logits/rejected": -1.9911689758300781,
|
| 852 |
+
"logps/chosen": -33.74740219116211,
|
| 853 |
+
"logps/rejected": -33.24574279785156,
|
| 854 |
+
"loss": 0.3833,
|
| 855 |
+
"rewards/accuracies": 0.9375,
|
| 856 |
+
"rewards/chosen": 0.29125431180000305,
|
| 857 |
+
"rewards/margins": 0.4282079339027405,
|
| 858 |
+
"rewards/rejected": -0.13695362210273743,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.9616893529891968,
|
| 865 |
+
"logits/rejected": -1.9681020975112915,
|
| 866 |
+
"logps/chosen": -31.261194229125977,
|
| 867 |
+
"logps/rejected": -32.988670349121094,
|
| 868 |
+
"loss": 0.29,
|
| 869 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 870 |
+
"rewards/chosen": 0.29519280791282654,
|
| 871 |
+
"rewards/margins": 0.43315425515174866,
|
| 872 |
+
"rewards/rejected": -0.1379614621400833,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -2.0316014289855957,
|
| 879 |
+
"logits/rejected": -2.0083844661712646,
|
| 880 |
+
"logps/chosen": -30.99959945678711,
|
| 881 |
+
"logps/rejected": -33.247528076171875,
|
| 882 |
+
"loss": 0.3983,
|
| 883 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 884 |
+
"rewards/chosen": 0.2627808749675751,
|
| 885 |
+
"rewards/margins": 0.4136640131473541,
|
| 886 |
+
"rewards/rejected": -0.15088310837745667,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.9671332836151123,
|
| 893 |
+
"logits/rejected": -1.9622156620025635,
|
| 894 |
+
"logps/chosen": -31.91083335876465,
|
| 895 |
+
"logps/rejected": -32.353614807128906,
|
| 896 |
+
"loss": 0.4301,
|
| 897 |
+
"rewards/accuracies": 0.875,
|
| 898 |
+
"rewards/chosen": 0.3035394251346588,
|
| 899 |
+
"rewards/margins": 0.4431583881378174,
|
| 900 |
+
"rewards/rejected": -0.13961896300315857,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -2.01167368888855,
|
| 907 |
+
"logits/rejected": -2.0135717391967773,
|
| 908 |
+
"logps/chosen": -34.01561737060547,
|
| 909 |
+
"logps/rejected": -35.08711624145508,
|
| 910 |
+
"loss": 0.36,
|
| 911 |
+
"rewards/accuracies": 0.862500011920929,
|
| 912 |
+
"rewards/chosen": 0.21610423922538757,
|
| 913 |
+
"rewards/margins": 0.38782355189323425,
|
| 914 |
+
"rewards/rejected": -0.17171931266784668,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.9978135824203491,
|
| 921 |
+
"logits/rejected": -1.9970858097076416,
|
| 922 |
+
"logps/chosen": -30.205724716186523,
|
| 923 |
+
"logps/rejected": -31.725692749023438,
|
| 924 |
+
"loss": 0.3897,
|
| 925 |
+
"rewards/accuracies": 0.862500011920929,
|
| 926 |
+
"rewards/chosen": 0.23179996013641357,
|
| 927 |
+
"rewards/margins": 0.3902333974838257,
|
| 928 |
+
"rewards/rejected": -0.1584334373474121,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -2.0145509243011475,
|
| 935 |
+
"logits/rejected": -2.012350082397461,
|
| 936 |
+
"logps/chosen": -32.4493408203125,
|
| 937 |
+
"logps/rejected": -32.55584716796875,
|
| 938 |
+
"loss": 0.5531,
|
| 939 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 940 |
+
"rewards/chosen": 0.3293648362159729,
|
| 941 |
+
"rewards/margins": 0.4828970432281494,
|
| 942 |
+
"rewards/rejected": -0.15353217720985413,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.2464921474456787,
|
| 948 |
+
"eval_logits/rejected": -2.241631031036377,
|
| 949 |
+
"eval_logps/chosen": -34.027915954589844,
|
| 950 |
+
"eval_logps/rejected": -37.547115325927734,
|
| 951 |
+
"eval_loss": 0.7434703707695007,
|
| 952 |
+
"eval_rewards/accuracies": 0.5577242374420166,
|
| 953 |
+
"eval_rewards/chosen": 0.003980191424489021,
|
| 954 |
+
"eval_rewards/margins": 0.022275729104876518,
|
| 955 |
+
"eval_rewards/rejected": -0.018295539543032646,
|
| 956 |
+
"eval_runtime": 145.9008,
|
| 957 |
+
"eval_samples_per_second": 2.351,
|
| 958 |
+
"eval_steps_per_second": 0.295,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 1.58,
|
| 963 |
+
"learning_rate": 3.7793282895240927e-06,
|
| 964 |
+
"logits/chosen": -2.0622594356536865,
|
| 965 |
+
"logits/rejected": -2.0689759254455566,
|
| 966 |
+
"logps/chosen": -31.731964111328125,
|
| 967 |
+
"logps/rejected": -32.2499885559082,
|
| 968 |
+
"loss": 0.2891,
|
| 969 |
+
"rewards/accuracies": 0.887499988079071,
|
| 970 |
+
"rewards/chosen": 0.3032490909099579,
|
| 971 |
+
"rewards/margins": 0.4153287410736084,
|
| 972 |
+
"rewards/rejected": -0.11207965761423111,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -2.0177249908447266,
|
| 979 |
+
"logits/rejected": -2.0152080059051514,
|
| 980 |
+
"logps/chosen": -34.171417236328125,
|
| 981 |
+
"logps/rejected": -31.206134796142578,
|
| 982 |
+
"loss": 0.5415,
|
| 983 |
+
"rewards/accuracies": 0.875,
|
| 984 |
+
"rewards/chosen": 0.35067808628082275,
|
| 985 |
+
"rewards/margins": 0.4830544888973236,
|
| 986 |
+
"rewards/rejected": -0.13237646222114563,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.9650866985321045,
|
| 993 |
+
"logits/rejected": -1.9581254720687866,
|
| 994 |
+
"logps/chosen": -34.88035202026367,
|
| 995 |
+
"logps/rejected": -32.54679489135742,
|
| 996 |
+
"loss": 0.4357,
|
| 997 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 998 |
+
"rewards/chosen": 0.35032278299331665,
|
| 999 |
+
"rewards/margins": 0.4640774130821228,
|
| 1000 |
+
"rewards/rejected": -0.11375461518764496,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.9704965353012085,
|
| 1007 |
+
"logits/rejected": -1.9766871929168701,
|
| 1008 |
+
"logps/chosen": -33.551597595214844,
|
| 1009 |
+
"logps/rejected": -33.46538162231445,
|
| 1010 |
+
"loss": 0.2942,
|
| 1011 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1012 |
+
"rewards/chosen": 0.2566205859184265,
|
| 1013 |
+
"rewards/margins": 0.39169201254844666,
|
| 1014 |
+
"rewards/rejected": -0.13507142663002014,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -2.0592033863067627,
|
| 1021 |
+
"logits/rejected": -2.073496103286743,
|
| 1022 |
+
"logps/chosen": -31.369670867919922,
|
| 1023 |
+
"logps/rejected": -32.21284484863281,
|
| 1024 |
+
"loss": 0.3661,
|
| 1025 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1026 |
+
"rewards/chosen": 0.28967565298080444,
|
| 1027 |
+
"rewards/margins": 0.4187701642513275,
|
| 1028 |
+
"rewards/rejected": -0.12909449636936188,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.9380521774291992,
|
| 1035 |
+
"logits/rejected": -1.9349801540374756,
|
| 1036 |
+
"logps/chosen": -33.320587158203125,
|
| 1037 |
+
"logps/rejected": -34.86935806274414,
|
| 1038 |
+
"loss": 0.5616,
|
| 1039 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1040 |
+
"rewards/chosen": 0.3614245057106018,
|
| 1041 |
+
"rewards/margins": 0.46288686990737915,
|
| 1042 |
+
"rewards/rejected": -0.10146232694387436,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -2.0301218032836914,
|
| 1049 |
+
"logits/rejected": -2.0302772521972656,
|
| 1050 |
+
"logps/chosen": -30.811248779296875,
|
| 1051 |
+
"logps/rejected": -34.03461837768555,
|
| 1052 |
+
"loss": 0.2823,
|
| 1053 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1054 |
+
"rewards/chosen": 0.2846068739891052,
|
| 1055 |
+
"rewards/margins": 0.40955057740211487,
|
| 1056 |
+
"rewards/rejected": -0.12494371086359024,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.97750723361969,
|
| 1063 |
+
"logits/rejected": -1.9745209217071533,
|
| 1064 |
+
"logps/chosen": -30.4150333404541,
|
| 1065 |
+
"logps/rejected": -33.014732360839844,
|
| 1066 |
+
"loss": 0.4637,
|
| 1067 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1068 |
+
"rewards/chosen": 0.2743886411190033,
|
| 1069 |
+
"rewards/margins": 0.3925444185733795,
|
| 1070 |
+
"rewards/rejected": -0.11815576255321503,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -2.015723466873169,
|
| 1077 |
+
"logits/rejected": -2.0155797004699707,
|
| 1078 |
+
"logps/chosen": -29.640954971313477,
|
| 1079 |
+
"logps/rejected": -30.86568832397461,
|
| 1080 |
+
"loss": 0.3834,
|
| 1081 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1082 |
+
"rewards/chosen": 0.31328272819519043,
|
| 1083 |
+
"rewards/margins": 0.4303969740867615,
|
| 1084 |
+
"rewards/rejected": -0.11711428314447403,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.9227421283721924,
|
| 1091 |
+
"logits/rejected": -1.9260936975479126,
|
| 1092 |
+
"logps/chosen": -34.0415153503418,
|
| 1093 |
+
"logps/rejected": -31.62136459350586,
|
| 1094 |
+
"loss": 0.4315,
|
| 1095 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1096 |
+
"rewards/chosen": 0.2621740996837616,
|
| 1097 |
+
"rewards/margins": 0.3561774790287018,
|
| 1098 |
+
"rewards/rejected": -0.09400339424610138,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.2498111724853516,
|
| 1104 |
+
"eval_logits/rejected": -2.2449464797973633,
|
| 1105 |
+
"eval_logps/chosen": -34.047359466552734,
|
| 1106 |
+
"eval_logps/rejected": -37.52130126953125,
|
| 1107 |
+
"eval_loss": 0.8140609264373779,
|
| 1108 |
+
"eval_rewards/accuracies": 0.49958473443984985,
|
| 1109 |
+
"eval_rewards/chosen": -0.007684962823987007,
|
| 1110 |
+
"eval_rewards/margins": -0.0048751430585980415,
|
| 1111 |
+
"eval_rewards/rejected": -0.0028098197653889656,
|
| 1112 |
+
"eval_runtime": 145.859,
|
| 1113 |
+
"eval_samples_per_second": 2.352,
|
| 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": -2.0064587593078613,
|
| 1121 |
+
"logits/rejected": -1.9971309900283813,
|
| 1122 |
+
"logps/chosen": -36.031028747558594,
|
| 1123 |
+
"logps/rejected": -32.0997428894043,
|
| 1124 |
+
"loss": 0.2508,
|
| 1125 |
+
"rewards/accuracies": 0.9375,
|
| 1126 |
+
"rewards/chosen": 0.28859394788742065,
|
| 1127 |
+
"rewards/margins": 0.4049535393714905,
|
| 1128 |
+
"rewards/rejected": -0.11635950952768326,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -2.020625352859497,
|
| 1135 |
+
"logits/rejected": -2.0215954780578613,
|
| 1136 |
+
"logps/chosen": -35.916099548339844,
|
| 1137 |
+
"logps/rejected": -33.29255676269531,
|
| 1138 |
+
"loss": 0.416,
|
| 1139 |
+
"rewards/accuracies": 0.875,
|
| 1140 |
+
"rewards/chosen": 0.31157156825065613,
|
| 1141 |
+
"rewards/margins": 0.4414919316768646,
|
| 1142 |
+
"rewards/rejected": -0.1299203783273697,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -2.0350329875946045,
|
| 1149 |
+
"logits/rejected": -2.032480239868164,
|
| 1150 |
+
"logps/chosen": -32.11791229248047,
|
| 1151 |
+
"logps/rejected": -33.21588897705078,
|
| 1152 |
+
"loss": 0.3069,
|
| 1153 |
+
"rewards/accuracies": 0.9375,
|
| 1154 |
+
"rewards/chosen": 0.2672501802444458,
|
| 1155 |
+
"rewards/margins": 0.3986472487449646,
|
| 1156 |
+
"rewards/rejected": -0.1313970983028412,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -2.022247791290283,
|
| 1163 |
+
"logits/rejected": -2.0273849964141846,
|
| 1164 |
+
"logps/chosen": -31.30985450744629,
|
| 1165 |
+
"logps/rejected": -33.424705505371094,
|
| 1166 |
+
"loss": 0.3425,
|
| 1167 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1168 |
+
"rewards/chosen": 0.28552335500717163,
|
| 1169 |
+
"rewards/margins": 0.447455495595932,
|
| 1170 |
+
"rewards/rejected": -0.161932110786438,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.9659734964370728,
|
| 1177 |
+
"logits/rejected": -1.9647700786590576,
|
| 1178 |
+
"logps/chosen": -32.636070251464844,
|
| 1179 |
+
"logps/rejected": -34.969459533691406,
|
| 1180 |
+
"loss": 0.3445,
|
| 1181 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1182 |
+
"rewards/chosen": 0.2933460772037506,
|
| 1183 |
+
"rewards/margins": 0.46344757080078125,
|
| 1184 |
+
"rewards/rejected": -0.17010149359703064,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.8426628112792969,
|
| 1191 |
+
"logits/rejected": -1.837442398071289,
|
| 1192 |
+
"logps/chosen": -35.48710632324219,
|
| 1193 |
+
"logps/rejected": -35.31467056274414,
|
| 1194 |
+
"loss": 0.573,
|
| 1195 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1196 |
+
"rewards/chosen": 0.3068006932735443,
|
| 1197 |
+
"rewards/margins": 0.37055516242980957,
|
| 1198 |
+
"rewards/rejected": -0.06375448405742645,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.983940839767456,
|
| 1205 |
+
"logits/rejected": -1.9875280857086182,
|
| 1206 |
+
"logps/chosen": -34.71583938598633,
|
| 1207 |
+
"logps/rejected": -33.711788177490234,
|
| 1208 |
+
"loss": 0.4316,
|
| 1209 |
+
"rewards/accuracies": 0.8208333849906921,
|
| 1210 |
+
"rewards/chosen": 0.29869017004966736,
|
| 1211 |
+
"rewards/margins": 0.36885932087898254,
|
| 1212 |
+
"rewards/rejected": -0.0701691135764122,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.9259727001190186,
|
| 1219 |
+
"logits/rejected": -1.924078345298767,
|
| 1220 |
+
"logps/chosen": -33.34611129760742,
|
| 1221 |
+
"logps/rejected": -34.635719299316406,
|
| 1222 |
+
"loss": 0.2786,
|
| 1223 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1224 |
+
"rewards/chosen": 0.33328744769096375,
|
| 1225 |
+
"rewards/margins": 0.4719271659851074,
|
| 1226 |
+
"rewards/rejected": -0.1386396586894989,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.973812460899353,
|
| 1233 |
+
"logits/rejected": -1.9723854064941406,
|
| 1234 |
+
"logps/chosen": -32.87940216064453,
|
| 1235 |
+
"logps/rejected": -33.219146728515625,
|
| 1236 |
+
"loss": 0.213,
|
| 1237 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1238 |
+
"rewards/chosen": 0.29780465364456177,
|
| 1239 |
+
"rewards/margins": 0.37734922766685486,
|
| 1240 |
+
"rewards/rejected": -0.0795445665717125,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.9056497812271118,
|
| 1247 |
+
"logits/rejected": -1.899532675743103,
|
| 1248 |
+
"logps/chosen": -33.77119064331055,
|
| 1249 |
+
"logps/rejected": -31.565038681030273,
|
| 1250 |
+
"loss": 0.3445,
|
| 1251 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1252 |
+
"rewards/chosen": 0.3671306371688843,
|
| 1253 |
+
"rewards/margins": 0.4275972247123718,
|
| 1254 |
+
"rewards/rejected": -0.060466617345809937,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.248525857925415,
|
| 1260 |
+
"eval_logits/rejected": -2.2436628341674805,
|
| 1261 |
+
"eval_logps/chosen": -33.9709587097168,
|
| 1262 |
+
"eval_logps/rejected": -37.455101013183594,
|
| 1263 |
+
"eval_loss": 0.8104392886161804,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5282392501831055,
|
| 1265 |
+
"eval_rewards/chosen": 0.038154613226652145,
|
| 1266 |
+
"eval_rewards/margins": 0.0012450924841687083,
|
| 1267 |
+
"eval_rewards/rejected": 0.036909520626068115,
|
| 1268 |
+
"eval_runtime": 145.9117,
|
| 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": -2.0072484016418457,
|
| 1277 |
+
"logits/rejected": -2.014772653579712,
|
| 1278 |
+
"logps/chosen": -31.6469669342041,
|
| 1279 |
+
"logps/rejected": -34.390533447265625,
|
| 1280 |
+
"loss": 0.2774,
|
| 1281 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1282 |
+
"rewards/chosen": 0.2887730598449707,
|
| 1283 |
+
"rewards/margins": 0.36958909034729004,
|
| 1284 |
+
"rewards/rejected": -0.08081604540348053,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.9808098077774048,
|
| 1291 |
+
"logits/rejected": -1.983655571937561,
|
| 1292 |
+
"logps/chosen": -32.01128005981445,
|
| 1293 |
+
"logps/rejected": -33.38429641723633,
|
| 1294 |
+
"loss": 0.2516,
|
| 1295 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1296 |
+
"rewards/chosen": 0.33588114380836487,
|
| 1297 |
+
"rewards/margins": 0.42857733368873596,
|
| 1298 |
+
"rewards/rejected": -0.09269621223211288,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.8852159976959229,
|
| 1305 |
+
"logits/rejected": -1.8893970251083374,
|
| 1306 |
+
"logps/chosen": -32.007896423339844,
|
| 1307 |
+
"logps/rejected": -35.13220977783203,
|
| 1308 |
+
"loss": 0.2545,
|
| 1309 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1310 |
+
"rewards/chosen": 0.3501349091529846,
|
| 1311 |
+
"rewards/margins": 0.4833507537841797,
|
| 1312 |
+
"rewards/rejected": -0.13321581482887268,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.948075294494629,
|
| 1319 |
+
"logits/rejected": -1.9441587924957275,
|
| 1320 |
+
"logps/chosen": -32.47659683227539,
|
| 1321 |
+
"logps/rejected": -33.118370056152344,
|
| 1322 |
+
"loss": 0.2688,
|
| 1323 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1324 |
+
"rewards/chosen": 0.29523491859436035,
|
| 1325 |
+
"rewards/margins": 0.3694687485694885,
|
| 1326 |
+
"rewards/rejected": -0.07423384487628937,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -2.035362482070923,
|
| 1333 |
+
"logits/rejected": -2.037264585494995,
|
| 1334 |
+
"logps/chosen": -32.23796081542969,
|
| 1335 |
+
"logps/rejected": -35.856971740722656,
|
| 1336 |
+
"loss": 0.2484,
|
| 1337 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1338 |
+
"rewards/chosen": 0.3224264085292816,
|
| 1339 |
+
"rewards/margins": 0.4394490122795105,
|
| 1340 |
+
"rewards/rejected": -0.11702261120080948,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.8381340503692627,
|
| 1347 |
+
"logits/rejected": -1.8319995403289795,
|
| 1348 |
+
"logps/chosen": -35.71126174926758,
|
| 1349 |
+
"logps/rejected": -33.0915641784668,
|
| 1350 |
+
"loss": 0.2965,
|
| 1351 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1352 |
+
"rewards/chosen": 0.34224915504455566,
|
| 1353 |
+
"rewards/margins": 0.46204671263694763,
|
| 1354 |
+
"rewards/rejected": -0.11979756504297256,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.9029734134674072,
|
| 1361 |
+
"logits/rejected": -1.9030624628067017,
|
| 1362 |
+
"logps/chosen": -35.23638916015625,
|
| 1363 |
+
"logps/rejected": -34.9959602355957,
|
| 1364 |
+
"loss": 0.1819,
|
| 1365 |
+
"rewards/accuracies": 0.9375,
|
| 1366 |
+
"rewards/chosen": 0.3638507127761841,
|
| 1367 |
+
"rewards/margins": 0.4536662995815277,
|
| 1368 |
+
"rewards/rejected": -0.08981556445360184,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.9545879364013672,
|
| 1375 |
+
"logits/rejected": -1.9431183338165283,
|
| 1376 |
+
"logps/chosen": -33.588871002197266,
|
| 1377 |
+
"logps/rejected": -33.196922302246094,
|
| 1378 |
+
"loss": 0.2214,
|
| 1379 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1380 |
+
"rewards/chosen": 0.31062597036361694,
|
| 1381 |
+
"rewards/margins": 0.4755612015724182,
|
| 1382 |
+
"rewards/rejected": -0.16493523120880127,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -2.055995464324951,
|
| 1389 |
+
"logits/rejected": -2.0605499744415283,
|
| 1390 |
+
"logps/chosen": -32.802162170410156,
|
| 1391 |
+
"logps/rejected": -32.60199737548828,
|
| 1392 |
+
"loss": 0.3724,
|
| 1393 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1394 |
+
"rewards/chosen": 0.3166634738445282,
|
| 1395 |
+
"rewards/margins": 0.41597333550453186,
|
| 1396 |
+
"rewards/rejected": -0.09930987656116486,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.9720876216888428,
|
| 1403 |
+
"logits/rejected": -1.9648761749267578,
|
| 1404 |
+
"logps/chosen": -34.31425857543945,
|
| 1405 |
+
"logps/rejected": -32.596370697021484,
|
| 1406 |
+
"loss": 0.1902,
|
| 1407 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1408 |
+
"rewards/chosen": 0.3238928020000458,
|
| 1409 |
+
"rewards/margins": 0.44043517112731934,
|
| 1410 |
+
"rewards/rejected": -0.11654232442378998,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.2474730014801025,
|
| 1416 |
+
"eval_logits/rejected": -2.24257755279541,
|
| 1417 |
+
"eval_logps/chosen": -33.979740142822266,
|
| 1418 |
+
"eval_logps/rejected": -37.45598220825195,
|
| 1419 |
+
"eval_loss": 0.8122833371162415,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5199335813522339,
|
| 1421 |
+
"eval_rewards/chosen": 0.03288789838552475,
|
| 1422 |
+
"eval_rewards/margins": -0.003495381912216544,
|
| 1423 |
+
"eval_rewards/rejected": 0.03638327494263649,
|
| 1424 |
+
"eval_runtime": 145.9163,
|
| 1425 |
+
"eval_samples_per_second": 2.351,
|
| 1426 |
+
"eval_steps_per_second": 0.295,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 2.36,
|
| 1431 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 1432 |
+
"logits/chosen": -1.9646570682525635,
|
| 1433 |
+
"logits/rejected": -1.9652878046035767,
|
| 1434 |
+
"logps/chosen": -30.66654396057129,
|
| 1435 |
+
"logps/rejected": -34.830745697021484,
|
| 1436 |
+
"loss": 0.3036,
|
| 1437 |
+
"rewards/accuracies": 0.9375,
|
| 1438 |
+
"rewards/chosen": 0.30036959052085876,
|
| 1439 |
+
"rewards/margins": 0.38500353693962097,
|
| 1440 |
+
"rewards/rejected": -0.08463393151760101,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.9272854328155518,
|
| 1447 |
+
"logits/rejected": -1.9191501140594482,
|
| 1448 |
+
"logps/chosen": -31.269882202148438,
|
| 1449 |
+
"logps/rejected": -33.781036376953125,
|
| 1450 |
+
"loss": 0.2861,
|
| 1451 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1452 |
+
"rewards/chosen": 0.3414386212825775,
|
| 1453 |
+
"rewards/margins": 0.46723994612693787,
|
| 1454 |
+
"rewards/rejected": -0.12580129504203796,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -2.09513521194458,
|
| 1461 |
+
"logits/rejected": -2.095780849456787,
|
| 1462 |
+
"logps/chosen": -29.461566925048828,
|
| 1463 |
+
"logps/rejected": -33.07443618774414,
|
| 1464 |
+
"loss": 0.3863,
|
| 1465 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1466 |
+
"rewards/chosen": 0.3212849497795105,
|
| 1467 |
+
"rewards/margins": 0.4470299780368805,
|
| 1468 |
+
"rewards/rejected": -0.12574505805969238,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.9750986099243164,
|
| 1475 |
+
"logits/rejected": -1.984668493270874,
|
| 1476 |
+
"logps/chosen": -33.88821792602539,
|
| 1477 |
+
"logps/rejected": -33.15973663330078,
|
| 1478 |
+
"loss": 0.229,
|
| 1479 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1480 |
+
"rewards/chosen": 0.3221094012260437,
|
| 1481 |
+
"rewards/margins": 0.4241645336151123,
|
| 1482 |
+
"rewards/rejected": -0.10205511748790741,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -2.016716480255127,
|
| 1489 |
+
"logits/rejected": -2.0206027030944824,
|
| 1490 |
+
"logps/chosen": -30.1220703125,
|
| 1491 |
+
"logps/rejected": -31.91337013244629,
|
| 1492 |
+
"loss": 0.2308,
|
| 1493 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1494 |
+
"rewards/chosen": 0.31539037823677063,
|
| 1495 |
+
"rewards/margins": 0.42176294326782227,
|
| 1496 |
+
"rewards/rejected": -0.10637259483337402,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -2.017374038696289,
|
| 1503 |
+
"logits/rejected": -2.0133793354034424,
|
| 1504 |
+
"logps/chosen": -33.43627166748047,
|
| 1505 |
+
"logps/rejected": -33.95038986206055,
|
| 1506 |
+
"loss": 0.246,
|
| 1507 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1508 |
+
"rewards/chosen": 0.38336849212646484,
|
| 1509 |
+
"rewards/margins": 0.4097708761692047,
|
| 1510 |
+
"rewards/rejected": -0.026402413845062256,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.9316686391830444,
|
| 1517 |
+
"logits/rejected": -1.9342737197875977,
|
| 1518 |
+
"logps/chosen": -32.216575622558594,
|
| 1519 |
+
"logps/rejected": -35.39495849609375,
|
| 1520 |
+
"loss": 0.2262,
|
| 1521 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1522 |
+
"rewards/chosen": 0.35770565271377563,
|
| 1523 |
+
"rewards/margins": 0.4857249855995178,
|
| 1524 |
+
"rewards/rejected": -0.1280193030834198,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.9653689861297607,
|
| 1531 |
+
"logits/rejected": -1.958824872970581,
|
| 1532 |
+
"logps/chosen": -34.86272048950195,
|
| 1533 |
+
"logps/rejected": -36.444908142089844,
|
| 1534 |
+
"loss": 0.4198,
|
| 1535 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1536 |
+
"rewards/chosen": 0.32540959119796753,
|
| 1537 |
+
"rewards/margins": 0.43518051505088806,
|
| 1538 |
+
"rewards/rejected": -0.10977087914943695,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.9367698431015015,
|
| 1545 |
+
"logits/rejected": -1.9314988851547241,
|
| 1546 |
+
"logps/chosen": -30.504741668701172,
|
| 1547 |
+
"logps/rejected": -36.1971549987793,
|
| 1548 |
+
"loss": 0.2745,
|
| 1549 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1550 |
+
"rewards/chosen": 0.29892516136169434,
|
| 1551 |
+
"rewards/margins": 0.4232874810695648,
|
| 1552 |
+
"rewards/rejected": -0.12436232715845108,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.9119179248809814,
|
| 1559 |
+
"logits/rejected": -1.9008598327636719,
|
| 1560 |
+
"logps/chosen": -31.161075592041016,
|
| 1561 |
+
"logps/rejected": -29.76761245727539,
|
| 1562 |
+
"loss": 0.1724,
|
| 1563 |
+
"rewards/accuracies": 0.9375,
|
| 1564 |
+
"rewards/chosen": 0.35256943106651306,
|
| 1565 |
+
"rewards/margins": 0.45329660177230835,
|
| 1566 |
+
"rewards/rejected": -0.10072717815637589,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.2481319904327393,
|
| 1572 |
+
"eval_logits/rejected": -2.243258237838745,
|
| 1573 |
+
"eval_logps/chosen": -33.95476150512695,
|
| 1574 |
+
"eval_logps/rejected": -37.45245361328125,
|
| 1575 |
+
"eval_loss": 0.7849614024162292,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5365448594093323,
|
| 1577 |
+
"eval_rewards/chosen": 0.04787486046552658,
|
| 1578 |
+
"eval_rewards/margins": 0.009373747743666172,
|
| 1579 |
+
"eval_rewards/rejected": 0.03850111365318298,
|
| 1580 |
+
"eval_runtime": 145.98,
|
| 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": -2.025754928588867,
|
| 1589 |
+
"logits/rejected": -2.0250885486602783,
|
| 1590 |
+
"logps/chosen": -31.275997161865234,
|
| 1591 |
+
"logps/rejected": -30.338693618774414,
|
| 1592 |
+
"loss": 0.1764,
|
| 1593 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1594 |
+
"rewards/chosen": 0.3289177417755127,
|
| 1595 |
+
"rewards/margins": 0.3770465552806854,
|
| 1596 |
+
"rewards/rejected": -0.048128847032785416,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -2.061776876449585,
|
| 1603 |
+
"logits/rejected": -2.0624749660491943,
|
| 1604 |
+
"logps/chosen": -31.365009307861328,
|
| 1605 |
+
"logps/rejected": -31.31036949157715,
|
| 1606 |
+
"loss": 0.2275,
|
| 1607 |
+
"rewards/accuracies": 0.9375,
|
| 1608 |
+
"rewards/chosen": 0.34714099764823914,
|
| 1609 |
+
"rewards/margins": 0.43387675285339355,
|
| 1610 |
+
"rewards/rejected": -0.08673573285341263,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -2.035461664199829,
|
| 1617 |
+
"logits/rejected": -2.038501024246216,
|
| 1618 |
+
"logps/chosen": -31.733938217163086,
|
| 1619 |
+
"logps/rejected": -28.1260929107666,
|
| 1620 |
+
"loss": 0.2038,
|
| 1621 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1622 |
+
"rewards/chosen": 0.327678382396698,
|
| 1623 |
+
"rewards/margins": 0.4131060540676117,
|
| 1624 |
+
"rewards/rejected": -0.08542772382497787,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -2.0482096672058105,
|
| 1631 |
+
"logits/rejected": -2.0403339862823486,
|
| 1632 |
+
"logps/chosen": -30.584192276000977,
|
| 1633 |
+
"logps/rejected": -33.65731430053711,
|
| 1634 |
+
"loss": 0.2815,
|
| 1635 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1636 |
+
"rewards/chosen": 0.3562210202217102,
|
| 1637 |
+
"rewards/margins": 0.4504729211330414,
|
| 1638 |
+
"rewards/rejected": -0.09425191581249237,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.975842833518982,
|
| 1645 |
+
"logits/rejected": -1.986856460571289,
|
| 1646 |
+
"logps/chosen": -33.01598358154297,
|
| 1647 |
+
"logps/rejected": -32.32152557373047,
|
| 1648 |
+
"loss": 0.2538,
|
| 1649 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1650 |
+
"rewards/chosen": 0.32227379083633423,
|
| 1651 |
+
"rewards/margins": 0.3681793212890625,
|
| 1652 |
+
"rewards/rejected": -0.04590549319982529,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -2.056307077407837,
|
| 1659 |
+
"logits/rejected": -2.0468807220458984,
|
| 1660 |
+
"logps/chosen": -32.9362907409668,
|
| 1661 |
+
"logps/rejected": -36.47593688964844,
|
| 1662 |
+
"loss": 0.4119,
|
| 1663 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1664 |
+
"rewards/chosen": 0.3386926054954529,
|
| 1665 |
+
"rewards/margins": 0.4254216253757477,
|
| 1666 |
+
"rewards/rejected": -0.0867290273308754,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.9632923603057861,
|
| 1673 |
+
"logits/rejected": -1.9650466442108154,
|
| 1674 |
+
"logps/chosen": -35.4192008972168,
|
| 1675 |
+
"logps/rejected": -32.78428268432617,
|
| 1676 |
+
"loss": 0.2636,
|
| 1677 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1678 |
+
"rewards/chosen": 0.3420516550540924,
|
| 1679 |
+
"rewards/margins": 0.3938991129398346,
|
| 1680 |
+
"rewards/rejected": -0.05184745788574219,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -2.100684404373169,
|
| 1687 |
+
"logits/rejected": -2.101898193359375,
|
| 1688 |
+
"logps/chosen": -32.51726150512695,
|
| 1689 |
+
"logps/rejected": -34.33898162841797,
|
| 1690 |
+
"loss": 0.24,
|
| 1691 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1692 |
+
"rewards/chosen": 0.32422804832458496,
|
| 1693 |
+
"rewards/margins": 0.4582032561302185,
|
| 1694 |
+
"rewards/rejected": -0.1339752972126007,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -2.06551194190979,
|
| 1701 |
+
"logits/rejected": -2.0696327686309814,
|
| 1702 |
+
"logps/chosen": -30.335834503173828,
|
| 1703 |
+
"logps/rejected": -33.68609619140625,
|
| 1704 |
+
"loss": 0.2635,
|
| 1705 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1706 |
+
"rewards/chosen": 0.3344890773296356,
|
| 1707 |
+
"rewards/margins": 0.41744667291641235,
|
| 1708 |
+
"rewards/rejected": -0.08295761048793793,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -2.05393385887146,
|
| 1715 |
+
"logits/rejected": -2.059272289276123,
|
| 1716 |
+
"logps/chosen": -33.025550842285156,
|
| 1717 |
+
"logps/rejected": -30.527883529663086,
|
| 1718 |
+
"loss": 0.1864,
|
| 1719 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1720 |
+
"rewards/chosen": 0.3260694444179535,
|
| 1721 |
+
"rewards/margins": 0.4017953872680664,
|
| 1722 |
+
"rewards/rejected": -0.07572595030069351,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.244070529937744,
|
| 1728 |
+
"eval_logits/rejected": -2.2392044067382812,
|
| 1729 |
+
"eval_logps/chosen": -33.96210861206055,
|
| 1730 |
+
"eval_logps/rejected": -37.46709060668945,
|
| 1731 |
+
"eval_loss": 0.7625167369842529,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5132890343666077,
|
| 1733 |
+
"eval_rewards/chosen": 0.04346639662981033,
|
| 1734 |
+
"eval_rewards/margins": 0.013746440410614014,
|
| 1735 |
+
"eval_rewards/rejected": 0.02971995621919632,
|
| 1736 |
+
"eval_runtime": 145.9267,
|
| 1737 |
+
"eval_samples_per_second": 2.35,
|
| 1738 |
+
"eval_steps_per_second": 0.295,
|
| 1739 |
+
"step": 1100
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 2.88,
|
| 1743 |
+
"learning_rate": 1.0963854773524548e-06,
|
| 1744 |
+
"logits/chosen": -2.046633243560791,
|
| 1745 |
+
"logits/rejected": -2.046231269836426,
|
| 1746 |
+
"logps/chosen": -32.30084991455078,
|
| 1747 |
+
"logps/rejected": -31.1953125,
|
| 1748 |
+
"loss": 0.3011,
|
| 1749 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1750 |
+
"rewards/chosen": 0.3835814893245697,
|
| 1751 |
+
"rewards/margins": 0.491219699382782,
|
| 1752 |
+
"rewards/rejected": -0.10763820260763168,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.8906314373016357,
|
| 1759 |
+
"logits/rejected": -1.8921263217926025,
|
| 1760 |
+
"logps/chosen": -34.062103271484375,
|
| 1761 |
+
"logps/rejected": -33.113182067871094,
|
| 1762 |
+
"loss": 0.3593,
|
| 1763 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1764 |
+
"rewards/chosen": 0.30172425508499146,
|
| 1765 |
+
"rewards/margins": 0.34487384557724,
|
| 1766 |
+
"rewards/rejected": -0.043149612843990326,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.9306659698486328,
|
| 1773 |
+
"logits/rejected": -1.92781662940979,
|
| 1774 |
+
"logps/chosen": -30.51358985900879,
|
| 1775 |
+
"logps/rejected": -31.30172348022461,
|
| 1776 |
+
"loss": 0.3529,
|
| 1777 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1778 |
+
"rewards/chosen": 0.29887938499450684,
|
| 1779 |
+
"rewards/margins": 0.3804691433906555,
|
| 1780 |
+
"rewards/rejected": -0.08158973604440689,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -2.0672850608825684,
|
| 1787 |
+
"logits/rejected": -2.061563730239868,
|
| 1788 |
+
"logps/chosen": -32.45098876953125,
|
| 1789 |
+
"logps/rejected": -33.04710006713867,
|
| 1790 |
+
"loss": 0.2299,
|
| 1791 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1792 |
+
"rewards/chosen": 0.3532842993736267,
|
| 1793 |
+
"rewards/margins": 0.40913909673690796,
|
| 1794 |
+
"rewards/rejected": -0.05585480481386185,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.9143331050872803,
|
| 1801 |
+
"logits/rejected": -1.9041531085968018,
|
| 1802 |
+
"logps/chosen": -32.86902618408203,
|
| 1803 |
+
"logps/rejected": -31.468719482421875,
|
| 1804 |
+
"loss": 0.2081,
|
| 1805 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1806 |
+
"rewards/chosen": 0.349628746509552,
|
| 1807 |
+
"rewards/margins": 0.455420583486557,
|
| 1808 |
+
"rewards/rejected": -0.1057918444275856,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.9975311756134033,
|
| 1815 |
+
"logits/rejected": -1.9972455501556396,
|
| 1816 |
+
"logps/chosen": -33.09718704223633,
|
| 1817 |
+
"logps/rejected": -31.396026611328125,
|
| 1818 |
+
"loss": 0.1634,
|
| 1819 |
+
"rewards/accuracies": 0.9375,
|
| 1820 |
+
"rewards/chosen": 0.3507378101348877,
|
| 1821 |
+
"rewards/margins": 0.4674840569496155,
|
| 1822 |
+
"rewards/rejected": -0.11674626171588898,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -2.0674796104431152,
|
| 1829 |
+
"logits/rejected": -2.0618603229522705,
|
| 1830 |
+
"logps/chosen": -32.99332809448242,
|
| 1831 |
+
"logps/rejected": -33.066471099853516,
|
| 1832 |
+
"loss": 0.1455,
|
| 1833 |
+
"rewards/accuracies": 0.9375,
|
| 1834 |
+
"rewards/chosen": 0.3195871412754059,
|
| 1835 |
+
"rewards/margins": 0.3947373330593109,
|
| 1836 |
+
"rewards/rejected": -0.07515018433332443,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.9187389612197876,
|
| 1843 |
+
"logits/rejected": -1.9265291690826416,
|
| 1844 |
+
"logps/chosen": -32.774940490722656,
|
| 1845 |
+
"logps/rejected": -32.578086853027344,
|
| 1846 |
+
"loss": 0.1261,
|
| 1847 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1848 |
+
"rewards/chosen": 0.334833025932312,
|
| 1849 |
+
"rewards/margins": 0.46482500433921814,
|
| 1850 |
+
"rewards/rejected": -0.12999196350574493,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -2.0002593994140625,
|
| 1857 |
+
"logits/rejected": -1.9944689273834229,
|
| 1858 |
+
"logps/chosen": -31.51430892944336,
|
| 1859 |
+
"logps/rejected": -33.402320861816406,
|
| 1860 |
+
"loss": 0.1302,
|
| 1861 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1862 |
+
"rewards/chosen": 0.3955826163291931,
|
| 1863 |
+
"rewards/margins": 0.46690741181373596,
|
| 1864 |
+
"rewards/rejected": -0.07132478058338165,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.9393714666366577,
|
| 1871 |
+
"logits/rejected": -1.9415557384490967,
|
| 1872 |
+
"logps/chosen": -32.937740325927734,
|
| 1873 |
+
"logps/rejected": -32.9244499206543,
|
| 1874 |
+
"loss": 0.1629,
|
| 1875 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1876 |
+
"rewards/chosen": 0.3316211700439453,
|
| 1877 |
+
"rewards/margins": 0.4302603602409363,
|
| 1878 |
+
"rewards/rejected": -0.09863914549350739,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.2436180114746094,
|
| 1884 |
+
"eval_logits/rejected": -2.238736629486084,
|
| 1885 |
+
"eval_logps/chosen": -33.95472717285156,
|
| 1886 |
+
"eval_logps/rejected": -37.46039581298828,
|
| 1887 |
+
"eval_loss": 0.7623819708824158,
|
| 1888 |
+
"eval_rewards/accuracies": 0.5340532064437866,
|
| 1889 |
+
"eval_rewards/chosen": 0.04789462685585022,
|
| 1890 |
+
"eval_rewards/margins": 0.01415975671261549,
|
| 1891 |
+
"eval_rewards/rejected": 0.033734869211912155,
|
| 1892 |
+
"eval_runtime": 145.9396,
|
| 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": -1.8697309494018555,
|
| 1901 |
+
"logits/rejected": -1.8685888051986694,
|
| 1902 |
+
"logps/chosen": -35.43067169189453,
|
| 1903 |
+
"logps/rejected": -32.846221923828125,
|
| 1904 |
+
"loss": 0.186,
|
| 1905 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1906 |
+
"rewards/chosen": 0.3730871081352234,
|
| 1907 |
+
"rewards/margins": 0.4552043080329895,
|
| 1908 |
+
"rewards/rejected": -0.0821172297000885,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -2.0492749214172363,
|
| 1915 |
+
"logits/rejected": -2.043208122253418,
|
| 1916 |
+
"logps/chosen": -31.149316787719727,
|
| 1917 |
+
"logps/rejected": -33.047523498535156,
|
| 1918 |
+
"loss": 0.147,
|
| 1919 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1920 |
+
"rewards/chosen": 0.33262285590171814,
|
| 1921 |
+
"rewards/margins": 0.43401265144348145,
|
| 1922 |
+
"rewards/rejected": -0.10138984769582748,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.8950700759887695,
|
| 1929 |
+
"logits/rejected": -1.8832353353500366,
|
| 1930 |
+
"logps/chosen": -33.81637191772461,
|
| 1931 |
+
"logps/rejected": -33.06397247314453,
|
| 1932 |
+
"loss": 0.1532,
|
| 1933 |
+
"rewards/accuracies": 0.9375,
|
| 1934 |
+
"rewards/chosen": 0.3415863811969757,
|
| 1935 |
+
"rewards/margins": 0.4436774253845215,
|
| 1936 |
+
"rewards/rejected": -0.10209109634160995,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.9864425659179688,
|
| 1943 |
+
"logits/rejected": -1.979435682296753,
|
| 1944 |
+
"logps/chosen": -30.0882625579834,
|
| 1945 |
+
"logps/rejected": -29.048751831054688,
|
| 1946 |
+
"loss": 0.1166,
|
| 1947 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1948 |
+
"rewards/chosen": 0.3671828806400299,
|
| 1949 |
+
"rewards/margins": 0.4758971631526947,
|
| 1950 |
+
"rewards/rejected": -0.10871431976556778,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -2.020458221435547,
|
| 1957 |
+
"logits/rejected": -2.026707172393799,
|
| 1958 |
+
"logps/chosen": -34.78377914428711,
|
| 1959 |
+
"logps/rejected": -31.65704917907715,
|
| 1960 |
+
"loss": 0.1432,
|
| 1961 |
+
"rewards/accuracies": 0.9375,
|
| 1962 |
+
"rewards/chosen": 0.36696386337280273,
|
| 1963 |
+
"rewards/margins": 0.4502052664756775,
|
| 1964 |
+
"rewards/rejected": -0.08324144035577774,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.9884742498397827,
|
| 1971 |
+
"logits/rejected": -1.9890835285186768,
|
| 1972 |
+
"logps/chosen": -32.45567321777344,
|
| 1973 |
+
"logps/rejected": -32.96289825439453,
|
| 1974 |
+
"loss": 0.1291,
|
| 1975 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1976 |
+
"rewards/chosen": 0.3437303900718689,
|
| 1977 |
+
"rewards/margins": 0.43254727125167847,
|
| 1978 |
+
"rewards/rejected": -0.08881691098213196,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.9342743158340454,
|
| 1985 |
+
"logits/rejected": -1.9337249994277954,
|
| 1986 |
+
"logps/chosen": -32.51142501831055,
|
| 1987 |
+
"logps/rejected": -33.648887634277344,
|
| 1988 |
+
"loss": 0.1225,
|
| 1989 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1990 |
+
"rewards/chosen": 0.34764808416366577,
|
| 1991 |
+
"rewards/margins": 0.42461419105529785,
|
| 1992 |
+
"rewards/rejected": -0.07696612179279327,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.8161872625350952,
|
| 1999 |
+
"logits/rejected": -1.8113391399383545,
|
| 2000 |
+
"logps/chosen": -34.59672164916992,
|
| 2001 |
+
"logps/rejected": -31.089731216430664,
|
| 2002 |
+
"loss": 0.1114,
|
| 2003 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2004 |
+
"rewards/chosen": 0.3844338059425354,
|
| 2005 |
+
"rewards/margins": 0.48237746953964233,
|
| 2006 |
+
"rewards/rejected": -0.09794367849826813,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -2.023128032684326,
|
| 2013 |
+
"logits/rejected": -2.007215976715088,
|
| 2014 |
+
"logps/chosen": -33.77648162841797,
|
| 2015 |
+
"logps/rejected": -35.00053787231445,
|
| 2016 |
+
"loss": 0.0953,
|
| 2017 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2018 |
+
"rewards/chosen": 0.34265583753585815,
|
| 2019 |
+
"rewards/margins": 0.4837895929813385,
|
| 2020 |
+
"rewards/rejected": -0.14113374054431915,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.9649696350097656,
|
| 2027 |
+
"logits/rejected": -1.969999074935913,
|
| 2028 |
+
"logps/chosen": -32.89286804199219,
|
| 2029 |
+
"logps/rejected": -36.32848358154297,
|
| 2030 |
+
"loss": 0.1881,
|
| 2031 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2032 |
+
"rewards/chosen": 0.37485235929489136,
|
| 2033 |
+
"rewards/margins": 0.5100094676017761,
|
| 2034 |
+
"rewards/rejected": -0.13515713810920715,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.243983507156372,
|
| 2040 |
+
"eval_logits/rejected": -2.2391045093536377,
|
| 2041 |
+
"eval_logps/chosen": -33.958831787109375,
|
| 2042 |
+
"eval_logps/rejected": -37.44320297241211,
|
| 2043 |
+
"eval_loss": 0.7945417165756226,
|
| 2044 |
+
"eval_rewards/accuracies": 0.4966777563095093,
|
| 2045 |
+
"eval_rewards/chosen": 0.045431531965732574,
|
| 2046 |
+
"eval_rewards/margins": 0.0013837195001542568,
|
| 2047 |
+
"eval_rewards/rejected": 0.04404781758785248,
|
| 2048 |
+
"eval_runtime": 145.9089,
|
| 2049 |
+
"eval_samples_per_second": 2.351,
|
| 2050 |
+
"eval_steps_per_second": 0.295,
|
| 2051 |
+
"step": 1300
|
| 2052 |
+
},
|
| 2053 |
+
{
|
| 2054 |
+
"epoch": 3.4,
|
| 2055 |
+
"learning_rate": 3.321087280364757e-07,
|
| 2056 |
+
"logits/chosen": -1.9365456104278564,
|
| 2057 |
+
"logits/rejected": -1.9366706609725952,
|
| 2058 |
+
"logps/chosen": -35.67008590698242,
|
| 2059 |
+
"logps/rejected": -37.238311767578125,
|
| 2060 |
+
"loss": 0.1675,
|
| 2061 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2062 |
+
"rewards/chosen": 0.3443610668182373,
|
| 2063 |
+
"rewards/margins": 0.4368794560432434,
|
| 2064 |
+
"rewards/rejected": -0.09251832962036133,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -2.0239405632019043,
|
| 2071 |
+
"logits/rejected": -2.022326946258545,
|
| 2072 |
+
"logps/chosen": -31.496356964111328,
|
| 2073 |
+
"logps/rejected": -31.08811378479004,
|
| 2074 |
+
"loss": 0.1558,
|
| 2075 |
+
"rewards/accuracies": 0.9375,
|
| 2076 |
+
"rewards/chosen": 0.35850051045417786,
|
| 2077 |
+
"rewards/margins": 0.43935075402259827,
|
| 2078 |
+
"rewards/rejected": -0.08085022866725922,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.9284435510635376,
|
| 2085 |
+
"logits/rejected": -1.9298274517059326,
|
| 2086 |
+
"logps/chosen": -33.75816345214844,
|
| 2087 |
+
"logps/rejected": -33.829715728759766,
|
| 2088 |
+
"loss": 0.127,
|
| 2089 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2090 |
+
"rewards/chosen": 0.36965471506118774,
|
| 2091 |
+
"rewards/margins": 0.47827833890914917,
|
| 2092 |
+
"rewards/rejected": -0.10862363874912262,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.9887325763702393,
|
| 2099 |
+
"logits/rejected": -1.9875112771987915,
|
| 2100 |
+
"logps/chosen": -30.637014389038086,
|
| 2101 |
+
"logps/rejected": -32.18603515625,
|
| 2102 |
+
"loss": 0.1273,
|
| 2103 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2104 |
+
"rewards/chosen": 0.33699196577072144,
|
| 2105 |
+
"rewards/margins": 0.4534294009208679,
|
| 2106 |
+
"rewards/rejected": -0.11643745750188828,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -2.005549907684326,
|
| 2113 |
+
"logits/rejected": -1.9907426834106445,
|
| 2114 |
+
"logps/chosen": -33.7475700378418,
|
| 2115 |
+
"logps/rejected": -36.4091796875,
|
| 2116 |
+
"loss": 0.1417,
|
| 2117 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2118 |
+
"rewards/chosen": 0.3414615988731384,
|
| 2119 |
+
"rewards/margins": 0.4283091127872467,
|
| 2120 |
+
"rewards/rejected": -0.08684752136468887,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -2.0126748085021973,
|
| 2127 |
+
"logits/rejected": -2.0252599716186523,
|
| 2128 |
+
"logps/chosen": -31.497417449951172,
|
| 2129 |
+
"logps/rejected": -32.39682388305664,
|
| 2130 |
+
"loss": 0.1287,
|
| 2131 |
+
"rewards/accuracies": 0.9375,
|
| 2132 |
+
"rewards/chosen": 0.3426300883293152,
|
| 2133 |
+
"rewards/margins": 0.4368777275085449,
|
| 2134 |
+
"rewards/rejected": -0.09424759447574615,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -2.0992069244384766,
|
| 2141 |
+
"logits/rejected": -2.0984139442443848,
|
| 2142 |
+
"logps/chosen": -31.42146873474121,
|
| 2143 |
+
"logps/rejected": -36.053802490234375,
|
| 2144 |
+
"loss": 0.146,
|
| 2145 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2146 |
+
"rewards/chosen": 0.36537685990333557,
|
| 2147 |
+
"rewards/margins": 0.49912720918655396,
|
| 2148 |
+
"rewards/rejected": -0.13375037908554077,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.9654130935668945,
|
| 2155 |
+
"logits/rejected": -1.9733936786651611,
|
| 2156 |
+
"logps/chosen": -33.638160705566406,
|
| 2157 |
+
"logps/rejected": -37.61304473876953,
|
| 2158 |
+
"loss": 0.1702,
|
| 2159 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2160 |
+
"rewards/chosen": 0.325391948223114,
|
| 2161 |
+
"rewards/margins": 0.423430860042572,
|
| 2162 |
+
"rewards/rejected": -0.0980389267206192,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.9179338216781616,
|
| 2169 |
+
"logits/rejected": -1.913713812828064,
|
| 2170 |
+
"logps/chosen": -34.152076721191406,
|
| 2171 |
+
"logps/rejected": -31.679922103881836,
|
| 2172 |
+
"loss": 0.1542,
|
| 2173 |
+
"rewards/accuracies": 0.9375,
|
| 2174 |
+
"rewards/chosen": 0.3690822720527649,
|
| 2175 |
+
"rewards/margins": 0.4186008870601654,
|
| 2176 |
+
"rewards/rejected": -0.049518562853336334,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -2.0843544006347656,
|
| 2183 |
+
"logits/rejected": -2.0817577838897705,
|
| 2184 |
+
"logps/chosen": -30.708471298217773,
|
| 2185 |
+
"logps/rejected": -32.34890365600586,
|
| 2186 |
+
"loss": 0.1395,
|
| 2187 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2188 |
+
"rewards/chosen": 0.3528476655483246,
|
| 2189 |
+
"rewards/margins": 0.47550955414772034,
|
| 2190 |
+
"rewards/rejected": -0.12266188859939575,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.2457053661346436,
|
| 2196 |
+
"eval_logits/rejected": -2.240827798843384,
|
| 2197 |
+
"eval_logps/chosen": -33.95180130004883,
|
| 2198 |
+
"eval_logps/rejected": -37.45983123779297,
|
| 2199 |
+
"eval_loss": 0.7520378232002258,
|
| 2200 |
+
"eval_rewards/accuracies": 0.545265793800354,
|
| 2201 |
+
"eval_rewards/chosen": 0.04964962601661682,
|
| 2202 |
+
"eval_rewards/margins": 0.015574713237583637,
|
| 2203 |
+
"eval_rewards/rejected": 0.03407491371035576,
|
| 2204 |
+
"eval_runtime": 145.8805,
|
| 2205 |
+
"eval_samples_per_second": 2.351,
|
| 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": -2.0233004093170166,
|
| 2213 |
+
"logits/rejected": -2.0188450813293457,
|
| 2214 |
+
"logps/chosen": -33.010955810546875,
|
| 2215 |
+
"logps/rejected": -29.703121185302734,
|
| 2216 |
+
"loss": 0.152,
|
| 2217 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2218 |
+
"rewards/chosen": 0.3789446949958801,
|
| 2219 |
+
"rewards/margins": 0.46814146637916565,
|
| 2220 |
+
"rewards/rejected": -0.08919680118560791,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.9822219610214233,
|
| 2227 |
+
"logits/rejected": -1.9834082126617432,
|
| 2228 |
+
"logps/chosen": -32.7589111328125,
|
| 2229 |
+
"logps/rejected": -35.04393768310547,
|
| 2230 |
+
"loss": 0.1561,
|
| 2231 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2232 |
+
"rewards/chosen": 0.36357951164245605,
|
| 2233 |
+
"rewards/margins": 0.4355399012565613,
|
| 2234 |
+
"rewards/rejected": -0.07196036726236343,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.9348289966583252,
|
| 2241 |
+
"logits/rejected": -1.9279505014419556,
|
| 2242 |
+
"logps/chosen": -33.61431121826172,
|
| 2243 |
+
"logps/rejected": -36.830039978027344,
|
| 2244 |
+
"loss": 0.1341,
|
| 2245 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2246 |
+
"rewards/chosen": 0.3164740204811096,
|
| 2247 |
+
"rewards/margins": 0.40085238218307495,
|
| 2248 |
+
"rewards/rejected": -0.08437839895486832,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.9475971460342407,
|
| 2255 |
+
"logits/rejected": -1.9420528411865234,
|
| 2256 |
+
"logps/chosen": -28.74324607849121,
|
| 2257 |
+
"logps/rejected": -31.90035057067871,
|
| 2258 |
+
"loss": 0.1889,
|
| 2259 |
+
"rewards/accuracies": 0.9375,
|
| 2260 |
+
"rewards/chosen": 0.3419957458972931,
|
| 2261 |
+
"rewards/margins": 0.44381189346313477,
|
| 2262 |
+
"rewards/rejected": -0.10181611776351929,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.9640169143676758,
|
| 2269 |
+
"logits/rejected": -1.9535976648330688,
|
| 2270 |
+
"logps/chosen": -31.90164566040039,
|
| 2271 |
+
"logps/rejected": -35.4890251159668,
|
| 2272 |
+
"loss": 0.1843,
|
| 2273 |
+
"rewards/accuracies": 0.9375,
|
| 2274 |
+
"rewards/chosen": 0.34793075919151306,
|
| 2275 |
+
"rewards/margins": 0.47012752294540405,
|
| 2276 |
+
"rewards/rejected": -0.12219683080911636,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.9425592422485352,
|
| 2283 |
+
"logits/rejected": -1.9395831823349,
|
| 2284 |
+
"logps/chosen": -32.72911834716797,
|
| 2285 |
+
"logps/rejected": -31.02676773071289,
|
| 2286 |
+
"loss": 0.1479,
|
| 2287 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2288 |
+
"rewards/chosen": 0.3483937978744507,
|
| 2289 |
+
"rewards/margins": 0.4299589991569519,
|
| 2290 |
+
"rewards/rejected": -0.08156517893075943,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -2.0044729709625244,
|
| 2297 |
+
"logits/rejected": -2.011080741882324,
|
| 2298 |
+
"logps/chosen": -31.44146156311035,
|
| 2299 |
+
"logps/rejected": -33.138694763183594,
|
| 2300 |
+
"loss": 0.1786,
|
| 2301 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2302 |
+
"rewards/chosen": 0.37662890553474426,
|
| 2303 |
+
"rewards/margins": 0.48239096999168396,
|
| 2304 |
+
"rewards/rejected": -0.10576207935810089,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -2.0032641887664795,
|
| 2311 |
+
"logits/rejected": -2.0138297080993652,
|
| 2312 |
+
"logps/chosen": -31.077510833740234,
|
| 2313 |
+
"logps/rejected": -29.825885772705078,
|
| 2314 |
+
"loss": 0.1168,
|
| 2315 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2316 |
+
"rewards/chosen": 0.3706461489200592,
|
| 2317 |
+
"rewards/margins": 0.48903828859329224,
|
| 2318 |
+
"rewards/rejected": -0.11839208751916885,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.9371792078018188,
|
| 2325 |
+
"logits/rejected": -1.9299945831298828,
|
| 2326 |
+
"logps/chosen": -33.08808898925781,
|
| 2327 |
+
"logps/rejected": -32.423980712890625,
|
| 2328 |
+
"loss": 0.1489,
|
| 2329 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2330 |
+
"rewards/chosen": 0.3410363495349884,
|
| 2331 |
+
"rewards/margins": 0.4691395163536072,
|
| 2332 |
+
"rewards/rejected": -0.12810321152210236,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.987765908241272,
|
| 2339 |
+
"logits/rejected": -1.9844974279403687,
|
| 2340 |
+
"logps/chosen": -33.07574462890625,
|
| 2341 |
+
"logps/rejected": -33.45602035522461,
|
| 2342 |
+
"loss": 0.1436,
|
| 2343 |
+
"rewards/accuracies": 0.9375,
|
| 2344 |
+
"rewards/chosen": 0.32989656925201416,
|
| 2345 |
+
"rewards/margins": 0.44737356901168823,
|
| 2346 |
+
"rewards/rejected": -0.11747703701257706,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.245612859725952,
|
| 2352 |
+
"eval_logits/rejected": -2.2407407760620117,
|
| 2353 |
+
"eval_logps/chosen": -33.94506072998047,
|
| 2354 |
+
"eval_logps/rejected": -37.460636138916016,
|
| 2355 |
+
"eval_loss": 0.7351647019386292,
|
| 2356 |
+
"eval_rewards/accuracies": 0.5398671627044678,
|
| 2357 |
+
"eval_rewards/chosen": 0.05369636043906212,
|
| 2358 |
+
"eval_rewards/margins": 0.02010684460401535,
|
| 2359 |
+
"eval_rewards/rejected": 0.03358951583504677,
|
| 2360 |
+
"eval_runtime": 145.9788,
|
| 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": -2.067953586578369,
|
| 2369 |
+
"logits/rejected": -2.068563461303711,
|
| 2370 |
+
"logps/chosen": -29.285715103149414,
|
| 2371 |
+
"logps/rejected": -32.92698669433594,
|
| 2372 |
+
"loss": 0.2375,
|
| 2373 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2374 |
+
"rewards/chosen": 0.3183347284793854,
|
| 2375 |
+
"rewards/margins": 0.43025675415992737,
|
| 2376 |
+
"rewards/rejected": -0.11192206293344498,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.9213430881500244,
|
| 2383 |
+
"logits/rejected": -1.9212009906768799,
|
| 2384 |
+
"logps/chosen": -32.53323745727539,
|
| 2385 |
+
"logps/rejected": -34.081382751464844,
|
| 2386 |
+
"loss": 0.1987,
|
| 2387 |
+
"rewards/accuracies": 0.9375,
|
| 2388 |
+
"rewards/chosen": 0.3051798939704895,
|
| 2389 |
+
"rewards/margins": 0.42977064847946167,
|
| 2390 |
+
"rewards/rejected": -0.12459073960781097,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -2.0738425254821777,
|
| 2397 |
+
"logits/rejected": -2.071150541305542,
|
| 2398 |
+
"logps/chosen": -30.976924896240234,
|
| 2399 |
+
"logps/rejected": -31.667194366455078,
|
| 2400 |
+
"loss": 0.1555,
|
| 2401 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2402 |
+
"rewards/chosen": 0.2982819676399231,
|
| 2403 |
+
"rewards/margins": 0.4084067940711975,
|
| 2404 |
+
"rewards/rejected": -0.11012481153011322,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -2.056631326675415,
|
| 2411 |
+
"logits/rejected": -2.0577590465545654,
|
| 2412 |
+
"logps/chosen": -30.05573081970215,
|
| 2413 |
+
"logps/rejected": -30.0135498046875,
|
| 2414 |
+
"loss": 0.1485,
|
| 2415 |
+
"rewards/accuracies": 0.970833420753479,
|
| 2416 |
+
"rewards/chosen": 0.33852633833885193,
|
| 2417 |
+
"rewards/margins": 0.47254690527915955,
|
| 2418 |
+
"rewards/rejected": -0.13402055203914642,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 0.2286405209984098,
|
| 2426 |
+
"train_runtime": 10803.5829,
|
| 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,
|