Commit
•
781235a
1
Parent(s):
9f05d53
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: 2.
|
| 24 |
-
- Rewards/chosen:
|
| 25 |
-
- Rewards/rejected: -0.
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -34.
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,15 +55,27 @@ The following hyperparameters were used during training:
|
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
| 59 |
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
-
- num_epochs:
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
-
| Training Loss | Epoch | Step |
|
| 65 |
-
|
| 66 |
-
| 2.59 | 0.26 | 100 | 2.
|
| 67 |
-
| 2.2795 | 0.52 | 200 | 2.
|
| 68 |
-
| 1.7902 | 0.78 | 300 | 2.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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: 2.7673
|
| 22 |
+
- Rewards/chosen: 0.0035
|
| 23 |
+
- Rewards/rejected: -0.0188
|
| 24 |
+
- Rewards/accuracies: 0.5723
|
| 25 |
+
- Rewards/margins: 0.0223
|
| 26 |
+
- Logps/rejected: -37.5792
|
| 27 |
+
- Logps/chosen: -34.0229
|
| 28 |
+
- Logits/rejected: -2.2269
|
| 29 |
+
- Logits/chosen: -2.2318
|
| 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 |
+
| 2.59 | 0.26 | 100 | -2.2353 | -2.2305 | -34.0250 | -37.4993 | 2.8958 | 0.4730 | 0.0029 | -0.0024 | 0.0052 |
|
| 65 |
+
| 2.2795 | 0.52 | 200 | -2.2339 | -2.2290 | -34.0545 | -37.5651 | 2.8012 | 0.5278 | -0.0060 | 0.0085 | -0.0145 |
|
| 66 |
+
| 1.7902 | 0.78 | 300 | -2.2294 | -2.2245 | -34.0446 | -37.5724 | 2.7585 | 0.5748 | -0.0030 | 0.0137 | -0.0167 |
|
| 67 |
+
| 1.1358 | 1.04 | 400 | 2.7145 | 0.0043 | -0.0169 | 0.5573 | 0.0212 | -37.5730 | -34.0204 | -2.2172 | -2.2221 |
|
| 68 |
+
| 0.842 | 1.3 | 500 | 2.7488 | -0.0094 | -0.0331 | 0.5453 | 0.0237 | -37.6269 | -34.0658 | -2.2052 | -2.2100 |
|
| 69 |
+
| 1.3074 | 1.56 | 600 | 2.7022 | -0.0190 | -0.0450 | 0.5718 | 0.0260 | -37.6668 | -34.0979 | -2.2232 | -2.2281 |
|
| 70 |
+
| 1.2824 | 1.82 | 700 | 2.7508 | -0.0224 | -0.0454 | 0.5627 | 0.0230 | -37.6679 | -34.1092 | -2.2297 | -2.2346 |
|
| 71 |
+
| 0.5946 | 2.08 | 800 | 2.8085 | -0.0013 | -0.0203 | 0.5573 | 0.0190 | -37.5843 | -34.0389 | -2.2324 | -2.2372 |
|
| 72 |
+
| 0.3535 | 2.34 | 900 | 2.7256 | 0.0037 | -0.0203 | 0.5540 | 0.0240 | -37.5843 | -34.0222 | -2.2286 | -2.2334 |
|
| 73 |
+
| 0.4205 | 2.6 | 1000 | 2.8066 | -0.0001 | -0.0187 | 0.5307 | 0.0186 | -37.5790 | -34.0348 | -2.2279 | -2.2328 |
|
| 74 |
+
| 0.5121 | 2.86 | 1100 | 2.7946 | 0.0022 | -0.0186 | 0.5727 | 0.0208 | -37.5786 | -34.0271 | -2.2261 | -2.2310 |
|
| 75 |
+
| 0.4011 | 3.12 | 1200 | 2.7715 | -0.0010 | -0.0223 | 0.5486 | 0.0213 | -37.5909 | -34.0378 | -2.2272 | -2.2321 |
|
| 76 |
+
| 0.2672 | 3.38 | 1300 | 2.8251 | 0.0002 | -0.0172 | 0.5631 | 0.0174 | -37.5741 | -34.0340 | -2.2271 | -2.2320 |
|
| 77 |
+
| 0.2484 | 3.64 | 1400 | 2.7913 | 0.0003 | -0.0200 | 0.5664 | 0.0203 | -37.5833 | -34.0337 | -2.2267 | -2.2316 |
|
| 78 |
+
| 0.3309 | 3.9 | 1500 | 2.7673 | 0.0035 | -0.0188 | 0.5723 | 0.0223 | -37.5792 | -34.0229 | -2.2269 | -2.2318 |
|
| 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 sha256:5be67a20fb9ec9e173f0c34b37f8efff061c1aa14ab63ad38093ea5ef1197b48
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.2295007705688477,
|
| 4 |
"eval_logits/rejected": -2.2246556282043457,
|
| 5 |
"eval_logps/chosen": -34.054466247558594,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.353,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
-
"train_loss":
|
| 17 |
-
"train_runtime":
|
| 18 |
"train_samples": 3079,
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
"eval_logits/chosen": -2.2295007705688477,
|
| 4 |
"eval_logits/rejected": -2.2246556282043457,
|
| 5 |
"eval_logps/chosen": -34.054466247558594,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.353,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.5763176552661053,
|
| 17 |
+
"train_runtime": 10799.143,
|
| 18 |
"train_samples": 3079,
|
| 19 |
+
"train_samples_per_second": 1.14,
|
| 20 |
+
"train_steps_per_second": 0.143
|
| 21 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 3079,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
+
"train_loss": 0.5763176552661053,
|
| 4 |
+
"train_runtime": 10799.143,
|
| 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": 2.
|
| 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":
|
| 515 |
-
"rewards/accuracies": 0.
|
| 516 |
-
"rewards/chosen": 0.
|
| 517 |
-
"rewards/margins": 0.
|
| 518 |
-
"rewards/rejected": -0.
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
-
"learning_rate":
|
| 524 |
-
"logits/chosen": -1.
|
| 525 |
-
"logits/rejected": -1.
|
| 526 |
-
"logps/chosen": -33.
|
| 527 |
-
"logps/rejected": -32.
|
| 528 |
-
"loss":
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": 0.
|
| 531 |
-
"rewards/margins": 0.
|
| 532 |
-
"rewards/rejected": -0.
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -33.
|
| 541 |
-
"logps/rejected": -33.
|
| 542 |
-
"loss":
|
| 543 |
-
"rewards/accuracies": 0.
|
| 544 |
-
"rewards/chosen": 0.
|
| 545 |
-
"rewards/margins": 0.
|
| 546 |
-
"rewards/rejected": 0.
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
-
"learning_rate":
|
| 552 |
-
"logits/chosen": -1.
|
| 553 |
-
"logits/rejected": -1.
|
| 554 |
-
"logps/chosen": -32.
|
| 555 |
-
"logps/rejected": -32.
|
| 556 |
-
"loss": 1.
|
| 557 |
"rewards/accuracies": 0.800000011920929,
|
| 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": 2.
|
| 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":
|
| 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": 1.
|
| 599 |
-
"rewards/accuracies": 0.
|
| 600 |
-
"rewards/chosen": 0.
|
| 601 |
-
"rewards/margins": 0.
|
| 602 |
-
"rewards/rejected": -0.
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
-
"epoch": 1.
|
| 607 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 608 |
"total_flos": 0.0,
|
| 609 |
-
"train_loss":
|
| 610 |
-
"train_runtime":
|
| 611 |
-
"train_samples_per_second":
|
| 612 |
-
"train_steps_per_second": 0.
|
| 613 |
}
|
| 614 |
],
|
| 615 |
"logging_steps": 10,
|
| 616 |
-
"max_steps":
|
| 617 |
"num_input_tokens_seen": 0,
|
| 618 |
-
"num_train_epochs":
|
| 619 |
"save_steps": 100,
|
| 620 |
"total_flos": 0.0,
|
| 621 |
"train_batch_size": 4,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 4.0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 1540,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
+
"learning_rate": 4.84533120650964e-06,
|
| 496 |
+
"logits/chosen": -1.913961410522461,
|
| 497 |
+
"logits/rejected": -1.9106886386871338,
|
| 498 |
+
"logps/chosen": -31.318857192993164,
|
| 499 |
+
"logps/rejected": -33.77204513549805,
|
| 500 |
+
"loss": 2.2088,
|
| 501 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 502 |
+
"rewards/chosen": 0.08534741401672363,
|
| 503 |
+
"rewards/margins": 0.09431613236665726,
|
| 504 |
+
"rewards/rejected": -0.008968725800514221,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
+
"learning_rate": 4.825108134172131e-06,
|
| 510 |
+
"logits/chosen": -1.9629987478256226,
|
| 511 |
+
"logits/rejected": -1.9507677555084229,
|
| 512 |
+
"logps/chosen": -34.35413360595703,
|
| 513 |
+
"logps/rejected": -33.61958312988281,
|
| 514 |
+
"loss": 2.0934,
|
| 515 |
+
"rewards/accuracies": 0.675000011920929,
|
| 516 |
+
"rewards/chosen": 0.07258231937885284,
|
| 517 |
+
"rewards/margins": 0.09680517762899399,
|
| 518 |
+
"rewards/rejected": -0.024222861975431442,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
+
"learning_rate": 4.80369052967602e-06,
|
| 524 |
+
"logits/chosen": -1.998734712600708,
|
| 525 |
+
"logits/rejected": -1.9973121881484985,
|
| 526 |
+
"logps/chosen": -33.19254684448242,
|
| 527 |
+
"logps/rejected": -32.54850387573242,
|
| 528 |
+
"loss": 1.9721,
|
| 529 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 530 |
+
"rewards/chosen": 0.08797496557235718,
|
| 531 |
+
"rewards/margins": 0.10865161567926407,
|
| 532 |
+
"rewards/rejected": -0.020676637068390846,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
+
"learning_rate": 4.781089396387968e-06,
|
| 538 |
+
"logits/chosen": -2.0848472118377686,
|
| 539 |
+
"logits/rejected": -2.0691235065460205,
|
| 540 |
+
"logps/chosen": -33.76462173461914,
|
| 541 |
+
"logps/rejected": -33.10074234008789,
|
| 542 |
+
"loss": 1.9899,
|
| 543 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 544 |
+
"rewards/chosen": 0.12443150579929352,
|
| 545 |
+
"rewards/margins": 0.11924245208501816,
|
| 546 |
+
"rewards/rejected": 0.005189062096178532,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
+
"learning_rate": 4.757316345716554e-06,
|
| 552 |
+
"logits/chosen": -1.9581273794174194,
|
| 553 |
+
"logits/rejected": -1.957232117652893,
|
| 554 |
+
"logps/chosen": -32.85580062866211,
|
| 555 |
+
"logps/rejected": -32.48882293701172,
|
| 556 |
+
"loss": 1.9291,
|
| 557 |
"rewards/accuracies": 0.800000011920929,
|
| 558 |
+
"rewards/chosen": 0.1278878152370453,
|
| 559 |
+
"rewards/margins": 0.13498859107494354,
|
| 560 |
+
"rewards/rejected": -0.007100761868059635,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
+
"learning_rate": 4.73238359114687e-06,
|
| 566 |
+
"logits/chosen": -1.9130855798721313,
|
| 567 |
+
"logits/rejected": -1.9233741760253906,
|
| 568 |
+
"logps/chosen": -31.818429946899414,
|
| 569 |
+
"logps/rejected": -35.345218658447266,
|
| 570 |
+
"loss": 2.0234,
|
| 571 |
+
"rewards/accuracies": 0.737500011920929,
|
| 572 |
+
"rewards/chosen": 0.11587802320718765,
|
| 573 |
+
"rewards/margins": 0.1328156441450119,
|
| 574 |
+
"rewards/rejected": -0.0169376190751791,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
+
"learning_rate": 4.706303941965804e-06,
|
| 580 |
+
"logits/chosen": -2.04890775680542,
|
| 581 |
+
"logits/rejected": -2.042478561401367,
|
| 582 |
+
"logps/chosen": -33.29634094238281,
|
| 583 |
+
"logps/rejected": -29.252614974975586,
|
| 584 |
+
"loss": 1.9656,
|
| 585 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 586 |
+
"rewards/chosen": 0.10668818652629852,
|
| 587 |
+
"rewards/margins": 0.1145448237657547,
|
| 588 |
+
"rewards/rejected": -0.00785664189606905,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
+
"learning_rate": 4.679090796681225e-06,
|
| 594 |
+
"logits/chosen": -1.9075864553451538,
|
| 595 |
+
"logits/rejected": -1.909790277481079,
|
| 596 |
+
"logps/chosen": -33.689239501953125,
|
| 597 |
+
"logps/rejected": -30.878931045532227,
|
| 598 |
+
"loss": 1.7909,
|
| 599 |
+
"rewards/accuracies": 0.737500011920929,
|
| 600 |
+
"rewards/chosen": 0.16494306921958923,
|
| 601 |
+
"rewards/margins": 0.16683341562747955,
|
| 602 |
+
"rewards/rejected": -0.0018903283635154366,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 1.01,
|
| 607 |
+
"learning_rate": 4.650758136138454e-06,
|
| 608 |
+
"logits/chosen": -1.9365580081939697,
|
| 609 |
+
"logits/rejected": -1.9353389739990234,
|
| 610 |
+
"logps/chosen": -33.854270935058594,
|
| 611 |
+
"logps/rejected": -35.9100227355957,
|
| 612 |
+
"loss": 1.3576,
|
| 613 |
+
"rewards/accuracies": 0.845833420753479,
|
| 614 |
+
"rewards/chosen": 0.15915581583976746,
|
| 615 |
+
"rewards/margins": 0.23536045849323273,
|
| 616 |
+
"rewards/rejected": -0.07620462775230408,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.04,
|
| 621 |
+
"learning_rate": 4.621320516337559e-06,
|
| 622 |
+
"logits/chosen": -1.8729753494262695,
|
| 623 |
+
"logits/rejected": -1.8645976781845093,
|
| 624 |
+
"logps/chosen": -31.197795867919922,
|
| 625 |
+
"logps/rejected": -36.29020690917969,
|
| 626 |
+
"loss": 1.1358,
|
| 627 |
+
"rewards/accuracies": 0.925000011920929,
|
| 628 |
+
"rewards/chosen": 0.1728539764881134,
|
| 629 |
+
"rewards/margins": 0.2777361571788788,
|
| 630 |
+
"rewards/rejected": -0.10488219559192657,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.04,
|
| 635 |
+
"eval_logits/chosen": -2.2220804691314697,
|
| 636 |
+
"eval_logits/rejected": -2.21720027923584,
|
| 637 |
+
"eval_logps/chosen": -34.020362854003906,
|
| 638 |
+
"eval_logps/rejected": -37.57297897338867,
|
| 639 |
+
"eval_loss": 2.7145347595214844,
|
| 640 |
+
"eval_rewards/accuracies": 0.5573089718818665,
|
| 641 |
+
"eval_rewards/chosen": 0.004255868028849363,
|
| 642 |
+
"eval_rewards/margins": 0.021163523197174072,
|
| 643 |
+
"eval_rewards/rejected": -0.01690765656530857,
|
| 644 |
+
"eval_runtime": 146.3895,
|
| 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": -2.048337936401367,
|
| 653 |
+
"logits/rejected": -2.051287889480591,
|
| 654 |
+
"logps/chosen": -32.42567825317383,
|
| 655 |
+
"logps/rejected": -35.118385314941406,
|
| 656 |
+
"loss": 1.1201,
|
| 657 |
+
"rewards/accuracies": 0.887499988079071,
|
| 658 |
+
"rewards/chosen": 0.18847835063934326,
|
| 659 |
+
"rewards/margins": 0.2861395478248596,
|
| 660 |
+
"rewards/rejected": -0.09766118973493576,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.09,
|
| 665 |
+
"learning_rate": 4.559191453574582e-06,
|
| 666 |
+
"logits/chosen": -1.8899818658828735,
|
| 667 |
+
"logits/rejected": -1.8886104822158813,
|
| 668 |
+
"logps/chosen": -28.469974517822266,
|
| 669 |
+
"logps/rejected": -32.66831588745117,
|
| 670 |
+
"loss": 0.9649,
|
| 671 |
+
"rewards/accuracies": 0.925000011920929,
|
| 672 |
+
"rewards/chosen": 0.17927470803260803,
|
| 673 |
+
"rewards/margins": 0.28447118401527405,
|
| 674 |
+
"rewards/rejected": -0.10519645363092422,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.12,
|
| 679 |
+
"learning_rate": 4.52653192962838e-06,
|
| 680 |
+
"logits/chosen": -1.847696304321289,
|
| 681 |
+
"logits/rejected": -1.840720534324646,
|
| 682 |
+
"logps/chosen": -33.315834045410156,
|
| 683 |
+
"logps/rejected": -34.448524475097656,
|
| 684 |
+
"loss": 1.0039,
|
| 685 |
+
"rewards/accuracies": 0.925000011920929,
|
| 686 |
+
"rewards/chosen": 0.22747480869293213,
|
| 687 |
+
"rewards/margins": 0.28844988346099854,
|
| 688 |
+
"rewards/rejected": -0.0609750859439373,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.14,
|
| 693 |
+
"learning_rate": 4.492831268057307e-06,
|
| 694 |
+
"logits/chosen": -2.0194411277770996,
|
| 695 |
+
"logits/rejected": -2.0143327713012695,
|
| 696 |
+
"logps/chosen": -31.07706069946289,
|
| 697 |
+
"logps/rejected": -32.33830642700195,
|
| 698 |
+
"loss": 1.0755,
|
| 699 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 700 |
+
"rewards/chosen": 0.2118721455335617,
|
| 701 |
+
"rewards/margins": 0.323079913854599,
|
| 702 |
+
"rewards/rejected": -0.1112077385187149,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.17,
|
| 707 |
+
"learning_rate": 4.458106782690094e-06,
|
| 708 |
+
"logits/chosen": -1.9053621292114258,
|
| 709 |
+
"logits/rejected": -1.909564733505249,
|
| 710 |
+
"logps/chosen": -33.75780487060547,
|
| 711 |
+
"logps/rejected": -33.03002166748047,
|
| 712 |
+
"loss": 0.762,
|
| 713 |
+
"rewards/accuracies": 0.925000011920929,
|
| 714 |
+
"rewards/chosen": 0.2274133712053299,
|
| 715 |
+
"rewards/margins": 0.35982832312583923,
|
| 716 |
+
"rewards/rejected": -0.13241496682167053,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.19,
|
| 721 |
+
"learning_rate": 4.422376313348405e-06,
|
| 722 |
+
"logits/chosen": -1.9145190715789795,
|
| 723 |
+
"logits/rejected": -1.9087648391723633,
|
| 724 |
+
"logps/chosen": -34.66414260864258,
|
| 725 |
+
"logps/rejected": -35.435611724853516,
|
| 726 |
+
"loss": 0.7729,
|
| 727 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 728 |
+
"rewards/chosen": 0.2266150265932083,
|
| 729 |
+
"rewards/margins": 0.3725685179233551,
|
| 730 |
+
"rewards/rejected": -0.1459534913301468,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.22,
|
| 735 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 736 |
+
"logits/chosen": -1.9441449642181396,
|
| 737 |
+
"logits/rejected": -1.9437577724456787,
|
| 738 |
+
"logps/chosen": -33.43156814575195,
|
| 739 |
+
"logps/rejected": -34.55840301513672,
|
| 740 |
+
"loss": 0.8983,
|
| 741 |
+
"rewards/accuracies": 0.949999988079071,
|
| 742 |
+
"rewards/chosen": 0.22102901339530945,
|
| 743 |
+
"rewards/margins": 0.3329058289527893,
|
| 744 |
+
"rewards/rejected": -0.11187677085399628,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.25,
|
| 749 |
+
"learning_rate": 4.347971356735789e-06,
|
| 750 |
+
"logits/chosen": -1.9919426441192627,
|
| 751 |
+
"logits/rejected": -1.9730838537216187,
|
| 752 |
+
"logps/chosen": -33.42109298706055,
|
| 753 |
+
"logps/rejected": -33.51959228515625,
|
| 754 |
+
"loss": 0.8516,
|
| 755 |
+
"rewards/accuracies": 0.9375,
|
| 756 |
+
"rewards/chosen": 0.22190213203430176,
|
| 757 |
+
"rewards/margins": 0.3567855954170227,
|
| 758 |
+
"rewards/rejected": -0.13488344848155975,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.27,
|
| 763 |
+
"learning_rate": 4.309335095262675e-06,
|
| 764 |
+
"logits/chosen": -1.9600021839141846,
|
| 765 |
+
"logits/rejected": -1.959425687789917,
|
| 766 |
+
"logps/chosen": -30.914112091064453,
|
| 767 |
+
"logps/rejected": -31.54854393005371,
|
| 768 |
+
"loss": 0.9185,
|
| 769 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 770 |
+
"rewards/chosen": 0.2091566026210785,
|
| 771 |
+
"rewards/margins": 0.29990634322166443,
|
| 772 |
+
"rewards/rejected": -0.09074974805116653,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.3,
|
| 777 |
+
"learning_rate": 4.269769281772082e-06,
|
| 778 |
+
"logits/chosen": -1.925381064414978,
|
| 779 |
+
"logits/rejected": -1.918460488319397,
|
| 780 |
+
"logps/chosen": -32.001102447509766,
|
| 781 |
+
"logps/rejected": -35.07819366455078,
|
| 782 |
+
"loss": 0.842,
|
| 783 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 784 |
+
"rewards/chosen": 0.20070871710777283,
|
| 785 |
+
"rewards/margins": 0.33614423871040344,
|
| 786 |
+
"rewards/rejected": -0.13543552160263062,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.3,
|
| 791 |
+
"eval_logits/chosen": -2.2100203037261963,
|
| 792 |
+
"eval_logits/rejected": -2.2051749229431152,
|
| 793 |
+
"eval_logps/chosen": -34.06578063964844,
|
| 794 |
+
"eval_logps/rejected": -37.626869201660156,
|
| 795 |
+
"eval_loss": 2.748755693435669,
|
| 796 |
+
"eval_rewards/accuracies": 0.545265793800354,
|
| 797 |
+
"eval_rewards/chosen": -0.009368383325636387,
|
| 798 |
+
"eval_rewards/margins": 0.023705704137682915,
|
| 799 |
+
"eval_rewards/rejected": -0.033074092119932175,
|
| 800 |
+
"eval_runtime": 145.9371,
|
| 801 |
+
"eval_samples_per_second": 2.35,
|
| 802 |
+
"eval_steps_per_second": 0.295,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 1.32,
|
| 807 |
+
"learning_rate": 4.22929424333435e-06,
|
| 808 |
+
"logits/chosen": -1.9246925115585327,
|
| 809 |
+
"logits/rejected": -1.9284820556640625,
|
| 810 |
+
"logps/chosen": -28.50702476501465,
|
| 811 |
+
"logps/rejected": -33.27742004394531,
|
| 812 |
+
"loss": 0.9643,
|
| 813 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 814 |
+
"rewards/chosen": 0.20034047961235046,
|
| 815 |
+
"rewards/margins": 0.32577869296073914,
|
| 816 |
+
"rewards/rejected": -0.12543818354606628,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.35,
|
| 821 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 822 |
+
"logits/chosen": -1.9368045330047607,
|
| 823 |
+
"logits/rejected": -1.947389006614685,
|
| 824 |
+
"logps/chosen": -32.59033966064453,
|
| 825 |
+
"logps/rejected": -30.988018035888672,
|
| 826 |
+
"loss": 1.0336,
|
| 827 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 828 |
+
"rewards/chosen": 0.20159777998924255,
|
| 829 |
+
"rewards/margins": 0.30249762535095215,
|
| 830 |
+
"rewards/rejected": -0.100899837911129,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.38,
|
| 835 |
+
"learning_rate": 4.145700124802693e-06,
|
| 836 |
+
"logits/chosen": -1.8843421936035156,
|
| 837 |
+
"logits/rejected": -1.8818985223770142,
|
| 838 |
+
"logps/chosen": -30.96946144104004,
|
| 839 |
+
"logps/rejected": -30.476232528686523,
|
| 840 |
+
"loss": 0.8922,
|
| 841 |
+
"rewards/accuracies": 0.9375,
|
| 842 |
+
"rewards/chosen": 0.20928528904914856,
|
| 843 |
+
"rewards/margins": 0.29912617802619934,
|
| 844 |
+
"rewards/rejected": -0.08984090387821198,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.4,
|
| 849 |
+
"learning_rate": 4.102623991469562e-06,
|
| 850 |
+
"logits/chosen": -1.9689937829971313,
|
| 851 |
+
"logits/rejected": -1.9618867635726929,
|
| 852 |
+
"logps/chosen": -33.4919548034668,
|
| 853 |
+
"logps/rejected": -33.414024353027344,
|
| 854 |
+
"loss": 0.8707,
|
| 855 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 856 |
+
"rewards/chosen": 0.22226205468177795,
|
| 857 |
+
"rewards/margins": 0.34122234582901,
|
| 858 |
+
"rewards/rejected": -0.11896030604839325,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.43,
|
| 863 |
+
"learning_rate": 4.058724504646834e-06,
|
| 864 |
+
"logits/chosen": -1.9358627796173096,
|
| 865 |
+
"logits/rejected": -1.9422714710235596,
|
| 866 |
+
"logps/chosen": -31.05177879333496,
|
| 867 |
+
"logps/rejected": -33.11052703857422,
|
| 868 |
+
"loss": 0.8711,
|
| 869 |
+
"rewards/accuracies": 0.925000011920929,
|
| 870 |
+
"rewards/chosen": 0.2104228436946869,
|
| 871 |
+
"rewards/margins": 0.3159605860710144,
|
| 872 |
+
"rewards/rejected": -0.10553774982690811,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.45,
|
| 877 |
+
"learning_rate": 4.014024217844167e-06,
|
| 878 |
+
"logits/chosen": -2.007368803024292,
|
| 879 |
+
"logits/rejected": -1.9841699600219727,
|
| 880 |
+
"logps/chosen": -30.822118759155273,
|
| 881 |
+
"logps/rejected": -33.421234130859375,
|
| 882 |
+
"loss": 1.112,
|
| 883 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 884 |
+
"rewards/chosen": 0.1846354603767395,
|
| 885 |
+
"rewards/margins": 0.31218937039375305,
|
| 886 |
+
"rewards/rejected": -0.12755386531352997,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.48,
|
| 891 |
+
"learning_rate": 3.968546095984911e-06,
|
| 892 |
+
"logits/chosen": -1.9438260793685913,
|
| 893 |
+
"logits/rejected": -1.938947319984436,
|
| 894 |
+
"logps/chosen": -31.676071166992188,
|
| 895 |
+
"logps/rejected": -32.48112106323242,
|
| 896 |
+
"loss": 1.1236,
|
| 897 |
+
"rewards/accuracies": 0.887499988079071,
|
| 898 |
+
"rewards/chosen": 0.22219793498516083,
|
| 899 |
+
"rewards/margins": 0.3302595913410187,
|
| 900 |
+
"rewards/rejected": -0.10806162655353546,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.51,
|
| 905 |
+
"learning_rate": 3.922313503607806e-06,
|
| 906 |
+
"logits/chosen": -1.988651990890503,
|
| 907 |
+
"logits/rejected": -1.9905750751495361,
|
| 908 |
+
"logps/chosen": -33.790245056152344,
|
| 909 |
+
"logps/rejected": -35.30215835571289,
|
| 910 |
+
"loss": 0.9454,
|
| 911 |
+
"rewards/accuracies": 0.9375,
|
| 912 |
+
"rewards/chosen": 0.17566607892513275,
|
| 913 |
+
"rewards/margins": 0.3260383605957031,
|
| 914 |
+
"rewards/rejected": -0.15037226676940918,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.53,
|
| 919 |
+
"learning_rate": 3.875350192863368e-06,
|
| 920 |
+
"logits/chosen": -1.9764407873153687,
|
| 921 |
+
"logits/rejected": -1.9757487773895264,
|
| 922 |
+
"logps/chosen": -29.972003936767578,
|
| 923 |
+
"logps/rejected": -31.86748695373535,
|
| 924 |
+
"loss": 1.3098,
|
| 925 |
+
"rewards/accuracies": 0.887499988079071,
|
| 926 |
+
"rewards/chosen": 0.18601693212985992,
|
| 927 |
+
"rewards/margins": 0.3077720105648041,
|
| 928 |
+
"rewards/rejected": -0.12175510078668594,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.56,
|
| 933 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 934 |
+
"logits/chosen": -1.9929969310760498,
|
| 935 |
+
"logits/rejected": -1.9908206462860107,
|
| 936 |
+
"logps/chosen": -32.231727600097656,
|
| 937 |
+
"logps/rejected": -32.70503616333008,
|
| 938 |
+
"loss": 1.3074,
|
| 939 |
+
"rewards/accuracies": 0.925000011920929,
|
| 940 |
+
"rewards/chosen": 0.2299668788909912,
|
| 941 |
+
"rewards/margins": 0.35148885846138,
|
| 942 |
+
"rewards/rejected": -0.12152198702096939,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.56,
|
| 947 |
+
"eval_logits/chosen": -2.2280964851379395,
|
| 948 |
+
"eval_logits/rejected": -2.2232325077056885,
|
| 949 |
+
"eval_logps/chosen": -34.09792709350586,
|
| 950 |
+
"eval_logps/rejected": -37.6667594909668,
|
| 951 |
+
"eval_loss": 2.702193021774292,
|
| 952 |
+
"eval_rewards/accuracies": 0.5718438625335693,
|
| 953 |
+
"eval_rewards/chosen": -0.019013158977031708,
|
| 954 |
+
"eval_rewards/margins": 0.0260298028588295,
|
| 955 |
+
"eval_rewards/rejected": -0.04504295811057091,
|
| 956 |
+
"eval_runtime": 145.8362,
|
| 957 |
+
"eval_samples_per_second": 2.352,
|
| 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": -2.0413756370544434,
|
| 965 |
+
"logits/rejected": -2.0480709075927734,
|
| 966 |
+
"logps/chosen": -31.450891494750977,
|
| 967 |
+
"logps/rejected": -32.45174789428711,
|
| 968 |
+
"loss": 0.8495,
|
| 969 |
+
"rewards/accuracies": 0.875,
|
| 970 |
+
"rewards/chosen": 0.2359478920698166,
|
| 971 |
+
"rewards/margins": 0.3525156080722809,
|
| 972 |
+
"rewards/rejected": -0.1165677085518837,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.61,
|
| 977 |
+
"learning_rate": 3.730319028506478e-06,
|
| 978 |
+
"logits/chosen": -1.9975080490112305,
|
| 979 |
+
"logits/rejected": -1.9950459003448486,
|
| 980 |
+
"logps/chosen": -33.930023193359375,
|
| 981 |
+
"logps/rejected": -31.34659194946289,
|
| 982 |
+
"loss": 0.9151,
|
| 983 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 984 |
+
"rewards/chosen": 0.24775627255439758,
|
| 985 |
+
"rewards/margins": 0.3560810685157776,
|
| 986 |
+
"rewards/rejected": -0.10832476615905762,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.64,
|
| 991 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 992 |
+
"logits/chosen": -1.9439729452133179,
|
| 993 |
+
"logits/rejected": -1.9370648860931396,
|
| 994 |
+
"logps/chosen": -34.60912322998047,
|
| 995 |
+
"logps/rejected": -32.83094024658203,
|
| 996 |
+
"loss": 0.9749,
|
| 997 |
+
"rewards/accuracies": 0.949999988079071,
|
| 998 |
+
"rewards/chosen": 0.25653165578842163,
|
| 999 |
+
"rewards/margins": 0.39865174889564514,
|
| 1000 |
+
"rewards/rejected": -0.14212007820606232,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.66,
|
| 1005 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1006 |
+
"logits/chosen": -1.9510215520858765,
|
| 1007 |
+
"logits/rejected": -1.9572662115097046,
|
| 1008 |
+
"logps/chosen": -33.284202575683594,
|
| 1009 |
+
"logps/rejected": -33.653526306152344,
|
| 1010 |
+
"loss": 0.9691,
|
| 1011 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1012 |
+
"rewards/chosen": 0.20853087306022644,
|
| 1013 |
+
"rewards/margins": 0.33251121640205383,
|
| 1014 |
+
"rewards/rejected": -0.12398035824298859,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.69,
|
| 1019 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1020 |
+
"logits/chosen": -2.0401089191436768,
|
| 1021 |
+
"logits/rejected": -2.0543549060821533,
|
| 1022 |
+
"logps/chosen": -31.178333282470703,
|
| 1023 |
+
"logps/rejected": -32.36495590209961,
|
| 1024 |
+
"loss": 1.0007,
|
| 1025 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1026 |
+
"rewards/chosen": 0.20223848521709442,
|
| 1027 |
+
"rewards/margins": 0.31241875886917114,
|
| 1028 |
+
"rewards/rejected": -0.11018023639917374,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.71,
|
| 1033 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1034 |
+
"logits/chosen": -1.9176908731460571,
|
| 1035 |
+
"logits/rejected": -1.9146522283554077,
|
| 1036 |
+
"logps/chosen": -33.0969352722168,
|
| 1037 |
+
"logps/rejected": -35.121665954589844,
|
| 1038 |
+
"loss": 1.0214,
|
| 1039 |
+
"rewards/accuracies": 0.9375,
|
| 1040 |
+
"rewards/chosen": 0.24780750274658203,
|
| 1041 |
+
"rewards/margins": 0.3742314875125885,
|
| 1042 |
+
"rewards/rejected": -0.12642398476600647,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.74,
|
| 1047 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1048 |
+
"logits/chosen": -2.0097954273223877,
|
| 1049 |
+
"logits/rejected": -2.0099480152130127,
|
| 1050 |
+
"logps/chosen": -30.54343032836914,
|
| 1051 |
+
"logps/rejected": -34.20466232299805,
|
| 1052 |
+
"loss": 0.7447,
|
| 1053 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1054 |
+
"rewards/chosen": 0.2226485311985016,
|
| 1055 |
+
"rewards/margins": 0.33613303303718567,
|
| 1056 |
+
"rewards/rejected": -0.11348447948694229,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.77,
|
| 1061 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1062 |
+
"logits/chosen": -1.9596326351165771,
|
| 1063 |
+
"logits/rejected": -1.9565805196762085,
|
| 1064 |
+
"logps/chosen": -30.143539428710938,
|
| 1065 |
+
"logps/rejected": -33.235843658447266,
|
| 1066 |
+
"loss": 1.0717,
|
| 1067 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1068 |
+
"rewards/chosen": 0.21864235401153564,
|
| 1069 |
+
"rewards/margins": 0.3440553545951843,
|
| 1070 |
+
"rewards/rejected": -0.12541300058364868,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.79,
|
| 1075 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1076 |
+
"logits/chosen": -1.9983398914337158,
|
| 1077 |
+
"logits/rejected": -1.998247742652893,
|
| 1078 |
+
"logps/chosen": -29.290761947631836,
|
| 1079 |
+
"logps/rejected": -31.007495880126953,
|
| 1080 |
+
"loss": 0.8594,
|
| 1081 |
+
"rewards/accuracies": 0.9375,
|
| 1082 |
+
"rewards/chosen": 0.2616986036300659,
|
| 1083 |
+
"rewards/margins": 0.362798273563385,
|
| 1084 |
+
"rewards/rejected": -0.10109970718622208,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.82,
|
| 1089 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1090 |
+
"logits/chosen": -1.9065154790878296,
|
| 1091 |
+
"logits/rejected": -1.9098196029663086,
|
| 1092 |
+
"logps/chosen": -33.76243209838867,
|
| 1093 |
+
"logps/rejected": -31.786779403686523,
|
| 1094 |
+
"loss": 1.2824,
|
| 1095 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1096 |
+
"rewards/chosen": 0.2148117572069168,
|
| 1097 |
+
"rewards/margins": 0.31144005060195923,
|
| 1098 |
+
"rewards/rejected": -0.09662826359272003,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.82,
|
| 1103 |
+
"eval_logits/chosen": -2.234584331512451,
|
| 1104 |
+
"eval_logits/rejected": -2.2297234535217285,
|
| 1105 |
+
"eval_logps/chosen": -34.10921096801758,
|
| 1106 |
+
"eval_logps/rejected": -37.667877197265625,
|
| 1107 |
+
"eval_loss": 2.7507741451263428,
|
| 1108 |
+
"eval_rewards/accuracies": 0.5627076625823975,
|
| 1109 |
+
"eval_rewards/chosen": -0.022397221997380257,
|
| 1110 |
+
"eval_rewards/margins": 0.02297896333038807,
|
| 1111 |
+
"eval_rewards/rejected": -0.045376185327768326,
|
| 1112 |
+
"eval_runtime": 145.9412,
|
| 1113 |
+
"eval_samples_per_second": 2.35,
|
| 1114 |
+
"eval_steps_per_second": 0.295,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 1.84,
|
| 1119 |
+
"learning_rate": 3.2639145321045933e-06,
|
| 1120 |
+
"logits/chosen": -1.9887504577636719,
|
| 1121 |
+
"logits/rejected": -1.9794700145721436,
|
| 1122 |
+
"logps/chosen": -35.74055099487305,
|
| 1123 |
+
"logps/rejected": -32.28141403198242,
|
| 1124 |
+
"loss": 0.8176,
|
| 1125 |
+
"rewards/accuracies": 0.9375,
|
| 1126 |
+
"rewards/chosen": 0.23144133388996124,
|
| 1127 |
+
"rewards/margins": 0.3441217541694641,
|
| 1128 |
+
"rewards/rejected": -0.11268042027950287,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.87,
|
| 1133 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1134 |
+
"logits/chosen": -2.004058837890625,
|
| 1135 |
+
"logits/rejected": -2.005059242248535,
|
| 1136 |
+
"logps/chosen": -35.58434295654297,
|
| 1137 |
+
"logps/rejected": -33.38105010986328,
|
| 1138 |
+
"loss": 0.9979,
|
| 1139 |
+
"rewards/accuracies": 0.9375,
|
| 1140 |
+
"rewards/chosen": 0.25531288981437683,
|
| 1141 |
+
"rewards/margins": 0.34682005643844604,
|
| 1142 |
+
"rewards/rejected": -0.09150713682174683,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.9,
|
| 1147 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1148 |
+
"logits/chosen": -2.019291877746582,
|
| 1149 |
+
"logits/rejected": -2.016794443130493,
|
| 1150 |
+
"logps/chosen": -31.924413681030273,
|
| 1151 |
+
"logps/rejected": -33.40862274169922,
|
| 1152 |
+
"loss": 0.8798,
|
| 1153 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1154 |
+
"rewards/chosen": 0.19167427718639374,
|
| 1155 |
+
"rewards/margins": 0.3151909410953522,
|
| 1156 |
+
"rewards/rejected": -0.12351665645837784,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.92,
|
| 1161 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1162 |
+
"logits/chosen": -2.0047879219055176,
|
| 1163 |
+
"logits/rejected": -2.0099570751190186,
|
| 1164 |
+
"logps/chosen": -30.985260009765625,
|
| 1165 |
+
"logps/rejected": -33.60847091674805,
|
| 1166 |
+
"loss": 0.7509,
|
| 1167 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1168 |
+
"rewards/chosen": 0.24014058709144592,
|
| 1169 |
+
"rewards/margins": 0.37623801827430725,
|
| 1170 |
+
"rewards/rejected": -0.13609740138053894,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.95,
|
| 1175 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1176 |
+
"logits/chosen": -1.949629783630371,
|
| 1177 |
+
"logits/rejected": -1.9484565258026123,
|
| 1178 |
+
"logps/chosen": -32.38895797729492,
|
| 1179 |
+
"logps/rejected": -35.11094665527344,
|
| 1180 |
+
"loss": 0.9358,
|
| 1181 |
+
"rewards/accuracies": 0.9375,
|
| 1182 |
+
"rewards/chosen": 0.22080639004707336,
|
| 1183 |
+
"rewards/margins": 0.34830355644226074,
|
| 1184 |
+
"rewards/rejected": -0.127497136592865,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.97,
|
| 1189 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1190 |
+
"logits/chosen": -1.8270362615585327,
|
| 1191 |
+
"logits/rejected": -1.8219047784805298,
|
| 1192 |
+
"logps/chosen": -35.21699523925781,
|
| 1193 |
+
"logps/rejected": -35.50707244873047,
|
| 1194 |
+
"loss": 1.0518,
|
| 1195 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1196 |
+
"rewards/chosen": 0.2344343364238739,
|
| 1197 |
+
"rewards/margins": 0.3240329623222351,
|
| 1198 |
+
"rewards/rejected": -0.0895986333489418,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.0,
|
| 1203 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1204 |
+
"logits/chosen": -1.969038724899292,
|
| 1205 |
+
"logits/rejected": -1.9727178812026978,
|
| 1206 |
+
"logps/chosen": -34.40656661987305,
|
| 1207 |
+
"logps/rejected": -33.838356018066406,
|
| 1208 |
+
"loss": 1.0179,
|
| 1209 |
+
"rewards/accuracies": 0.8833333849906921,
|
| 1210 |
+
"rewards/chosen": 0.24212701618671417,
|
| 1211 |
+
"rewards/margins": 0.3151811957359314,
|
| 1212 |
+
"rewards/rejected": -0.07305420190095901,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.03,
|
| 1217 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1218 |
+
"logits/chosen": -1.9119853973388672,
|
| 1219 |
+
"logits/rejected": -1.9100983142852783,
|
| 1220 |
+
"logps/chosen": -32.9580192565918,
|
| 1221 |
+
"logps/rejected": -34.962135314941406,
|
| 1222 |
+
"loss": 0.4598,
|
| 1223 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1224 |
+
"rewards/chosen": 0.28307104110717773,
|
| 1225 |
+
"rewards/margins": 0.4503152370452881,
|
| 1226 |
+
"rewards/rejected": -0.16724422574043274,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.05,
|
| 1231 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1232 |
+
"logits/chosen": -1.960938811302185,
|
| 1233 |
+
"logits/rejected": -1.9594917297363281,
|
| 1234 |
+
"logps/chosen": -32.456539154052734,
|
| 1235 |
+
"logps/rejected": -33.508445739746094,
|
| 1236 |
+
"loss": 0.4823,
|
| 1237 |
+
"rewards/accuracies": 0.9375,
|
| 1238 |
+
"rewards/chosen": 0.2757590711116791,
|
| 1239 |
+
"rewards/margins": 0.4023224413394928,
|
| 1240 |
+
"rewards/rejected": -0.12656334042549133,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.08,
|
| 1245 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1246 |
+
"logits/chosen": -1.8920997381210327,
|
| 1247 |
+
"logits/rejected": -1.8860002756118774,
|
| 1248 |
+
"logps/chosen": -33.441009521484375,
|
| 1249 |
+
"logps/rejected": -31.868255615234375,
|
| 1250 |
+
"loss": 0.5946,
|
| 1251 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1252 |
+
"rewards/chosen": 0.2826210558414459,
|
| 1253 |
+
"rewards/margins": 0.40381956100463867,
|
| 1254 |
+
"rewards/rejected": -0.12119851261377335,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.08,
|
| 1259 |
+
"eval_logits/chosen": -2.237243890762329,
|
| 1260 |
+
"eval_logits/rejected": -2.23238205909729,
|
| 1261 |
+
"eval_logps/chosen": -34.038944244384766,
|
| 1262 |
+
"eval_logps/rejected": -37.584251403808594,
|
| 1263 |
+
"eval_loss": 2.8084943294525146,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5573089718818665,
|
| 1265 |
+
"eval_rewards/chosen": -0.0013179187662899494,
|
| 1266 |
+
"eval_rewards/margins": 0.018971215933561325,
|
| 1267 |
+
"eval_rewards/rejected": -0.020289132371544838,
|
| 1268 |
+
"eval_runtime": 145.9108,
|
| 1269 |
+
"eval_samples_per_second": 2.351,
|
| 1270 |
+
"eval_steps_per_second": 0.295,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.1,
|
| 1275 |
+
"learning_rate": 2.7094204786572254e-06,
|
| 1276 |
+
"logits/chosen": -1.9936996698379517,
|
| 1277 |
+
"logits/rejected": -2.0012588500976562,
|
| 1278 |
+
"logps/chosen": -31.18770408630371,
|
| 1279 |
+
"logps/rejected": -34.70355224609375,
|
| 1280 |
+
"loss": 0.5486,
|
| 1281 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1282 |
+
"rewards/chosen": 0.28216588497161865,
|
| 1283 |
+
"rewards/margins": 0.41647869348526,
|
| 1284 |
+
"rewards/rejected": -0.1343127191066742,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.13,
|
| 1289 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1290 |
+
"logits/chosen": -1.9674571752548218,
|
| 1291 |
+
"logits/rejected": -1.9703105688095093,
|
| 1292 |
+
"logps/chosen": -31.648365020751953,
|
| 1293 |
+
"logps/rejected": -33.587764739990234,
|
| 1294 |
+
"loss": 0.5075,
|
| 1295 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1296 |
+
"rewards/chosen": 0.2768138647079468,
|
| 1297 |
+
"rewards/margins": 0.38420218229293823,
|
| 1298 |
+
"rewards/rejected": -0.10738833248615265,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.16,
|
| 1303 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1304 |
+
"logits/chosen": -1.8725074529647827,
|
| 1305 |
+
"logits/rejected": -1.8766599893569946,
|
| 1306 |
+
"logps/chosen": -31.63332748413086,
|
| 1307 |
+
"logps/rejected": -35.44063186645508,
|
| 1308 |
+
"loss": 0.4976,
|
| 1309 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1310 |
+
"rewards/chosen": 0.287439227104187,
|
| 1311 |
+
"rewards/margins": 0.4465748369693756,
|
| 1312 |
+
"rewards/rejected": -0.1591355949640274,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.18,
|
| 1317 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1318 |
+
"logits/chosen": -1.9352226257324219,
|
| 1319 |
+
"logits/rejected": -1.9313485622406006,
|
| 1320 |
+
"logps/chosen": -32.06031799316406,
|
| 1321 |
+
"logps/rejected": -33.345420837402344,
|
| 1322 |
+
"loss": 0.8222,
|
| 1323 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1324 |
+
"rewards/chosen": 0.27250009775161743,
|
| 1325 |
+
"rewards/margins": 0.37773144245147705,
|
| 1326 |
+
"rewards/rejected": -0.10523136705160141,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.21,
|
| 1331 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1332 |
+
"logits/chosen": -2.0206172466278076,
|
| 1333 |
+
"logits/rejected": -2.0226147174835205,
|
| 1334 |
+
"logps/chosen": -31.859445571899414,
|
| 1335 |
+
"logps/rejected": -36.106101989746094,
|
| 1336 |
+
"loss": 0.4523,
|
| 1337 |
+
"rewards/accuracies": 1.0,
|
| 1338 |
+
"rewards/chosen": 0.2747666537761688,
|
| 1339 |
+
"rewards/margins": 0.4080173075199127,
|
| 1340 |
+
"rewards/rejected": -0.1332506239414215,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.23,
|
| 1345 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1346 |
+
"logits/chosen": -1.8238776922225952,
|
| 1347 |
+
"logits/rejected": -1.8177616596221924,
|
| 1348 |
+
"logps/chosen": -35.31499481201172,
|
| 1349 |
+
"logps/rejected": -33.369178771972656,
|
| 1350 |
+
"loss": 0.5286,
|
| 1351 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1352 |
+
"rewards/chosen": 0.2900038957595825,
|
| 1353 |
+
"rewards/margins": 0.4331858158111572,
|
| 1354 |
+
"rewards/rejected": -0.1431819200515747,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.26,
|
| 1359 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1360 |
+
"logits/chosen": -1.8872458934783936,
|
| 1361 |
+
"logits/rejected": -1.8873565196990967,
|
| 1362 |
+
"logps/chosen": -34.84423065185547,
|
| 1363 |
+
"logps/rejected": -35.240962982177734,
|
| 1364 |
+
"loss": 0.5031,
|
| 1365 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1366 |
+
"rewards/chosen": 0.29957348108291626,
|
| 1367 |
+
"rewards/margins": 0.41798338294029236,
|
| 1368 |
+
"rewards/rejected": -0.11840987205505371,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.29,
|
| 1373 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1374 |
+
"logits/chosen": -1.9393589496612549,
|
| 1375 |
+
"logits/rejected": -1.9278790950775146,
|
| 1376 |
+
"logps/chosen": -33.1755256652832,
|
| 1377 |
+
"logps/rejected": -33.432777404785156,
|
| 1378 |
+
"loss": 0.4086,
|
| 1379 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1380 |
+
"rewards/chosen": 0.2793174684047699,
|
| 1381 |
+
"rewards/margins": 0.43254247307777405,
|
| 1382 |
+
"rewards/rejected": -0.15322497487068176,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.31,
|
| 1387 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1388 |
+
"logits/chosen": -2.039930820465088,
|
| 1389 |
+
"logits/rejected": -2.0445265769958496,
|
| 1390 |
+
"logps/chosen": -32.41832733154297,
|
| 1391 |
+
"logps/rejected": -32.86511993408203,
|
| 1392 |
+
"loss": 0.5315,
|
| 1393 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1394 |
+
"rewards/chosen": 0.2734832167625427,
|
| 1395 |
+
"rewards/margins": 0.40207424759864807,
|
| 1396 |
+
"rewards/rejected": -0.12859100103378296,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.34,
|
| 1401 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1402 |
+
"logits/chosen": -1.9552028179168701,
|
| 1403 |
+
"logits/rejected": -1.9479535818099976,
|
| 1404 |
+
"logps/chosen": -33.88296890258789,
|
| 1405 |
+
"logps/rejected": -32.84563064575195,
|
| 1406 |
+
"loss": 0.3535,
|
| 1407 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1408 |
+
"rewards/chosen": 0.29133278131484985,
|
| 1409 |
+
"rewards/margins": 0.42438235878944397,
|
| 1410 |
+
"rewards/rejected": -0.1330495923757553,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.34,
|
| 1415 |
+
"eval_logits/chosen": -2.2334346771240234,
|
| 1416 |
+
"eval_logits/rejected": -2.2285592555999756,
|
| 1417 |
+
"eval_logps/chosen": -34.02215576171875,
|
| 1418 |
+
"eval_logps/rejected": -37.584251403808594,
|
| 1419 |
+
"eval_loss": 2.725637435913086,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5539867281913757,
|
| 1421 |
+
"eval_rewards/chosen": 0.003719477914273739,
|
| 1422 |
+
"eval_rewards/margins": 0.024009360000491142,
|
| 1423 |
+
"eval_rewards/rejected": -0.020289884880185127,
|
| 1424 |
+
"eval_runtime": 145.9216,
|
| 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": -1.947854995727539,
|
| 1433 |
+
"logits/rejected": -1.9484474658966064,
|
| 1434 |
+
"logps/chosen": -30.2750244140625,
|
| 1435 |
+
"logps/rejected": -35.1546630859375,
|
| 1436 |
+
"loss": 0.6375,
|
| 1437 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1438 |
+
"rewards/chosen": 0.26764118671417236,
|
| 1439 |
+
"rewards/margins": 0.40713411569595337,
|
| 1440 |
+
"rewards/rejected": -0.1394929587841034,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.39,
|
| 1445 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1446 |
+
"logits/chosen": -1.9090734720230103,
|
| 1447 |
+
"logits/rejected": -1.9009778499603271,
|
| 1448 |
+
"logps/chosen": -30.872310638427734,
|
| 1449 |
+
"logps/rejected": -34.03510284423828,
|
| 1450 |
+
"loss": 0.6116,
|
| 1451 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1452 |
+
"rewards/chosen": 0.2899911403656006,
|
| 1453 |
+
"rewards/margins": 0.4291112422943115,
|
| 1454 |
+
"rewards/rejected": -0.13912010192871094,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.42,
|
| 1459 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1460 |
+
"logits/chosen": -2.075429916381836,
|
| 1461 |
+
"logits/rejected": -2.075995922088623,
|
| 1462 |
+
"logps/chosen": -29.147113800048828,
|
| 1463 |
+
"logps/rejected": -33.25724411010742,
|
| 1464 |
+
"loss": 0.888,
|
| 1465 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1466 |
+
"rewards/chosen": 0.25497666001319885,
|
| 1467 |
+
"rewards/margins": 0.372692346572876,
|
| 1468 |
+
"rewards/rejected": -0.1177157312631607,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.44,
|
| 1473 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1474 |
+
"logits/chosen": -1.9561622142791748,
|
| 1475 |
+
"logits/rejected": -1.9657951593399048,
|
| 1476 |
+
"logps/chosen": -33.404273986816406,
|
| 1477 |
+
"logps/rejected": -33.500064849853516,
|
| 1478 |
+
"loss": 0.6931,
|
| 1479 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1480 |
+
"rewards/chosen": 0.3062388300895691,
|
| 1481 |
+
"rewards/margins": 0.45936495065689087,
|
| 1482 |
+
"rewards/rejected": -0.153126060962677,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.47,
|
| 1487 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1488 |
+
"logits/chosen": -1.998156189918518,
|
| 1489 |
+
"logits/rejected": -2.0020549297332764,
|
| 1490 |
+
"logps/chosen": -29.782520294189453,
|
| 1491 |
+
"logps/rejected": -32.168800354003906,
|
| 1492 |
+
"loss": 0.511,
|
| 1493 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1494 |
+
"rewards/chosen": 0.25956016778945923,
|
| 1495 |
+
"rewards/margins": 0.38937538862228394,
|
| 1496 |
+
"rewards/rejected": -0.1298152506351471,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.49,
|
| 1501 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1502 |
+
"logits/chosen": -1.9971129894256592,
|
| 1503 |
+
"logits/rejected": -1.9930740594863892,
|
| 1504 |
+
"logps/chosen": -33.038108825683594,
|
| 1505 |
+
"logps/rejected": -34.222476959228516,
|
| 1506 |
+
"loss": 0.5584,
|
| 1507 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1508 |
+
"rewards/chosen": 0.31113213300704956,
|
| 1509 |
+
"rewards/margins": 0.40595847368240356,
|
| 1510 |
+
"rewards/rejected": -0.09482637047767639,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.52,
|
| 1515 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1516 |
+
"logits/chosen": -1.9126231670379639,
|
| 1517 |
+
"logits/rejected": -1.9152530431747437,
|
| 1518 |
+
"logps/chosen": -31.796194076538086,
|
| 1519 |
+
"logps/rejected": -35.5867805480957,
|
| 1520 |
+
"loss": 0.3983,
|
| 1521 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1522 |
+
"rewards/chosen": 0.3049660921096802,
|
| 1523 |
+
"rewards/margins": 0.42652225494384766,
|
| 1524 |
+
"rewards/rejected": -0.12155614793300629,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.55,
|
| 1529 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1530 |
+
"logits/chosen": -1.9468822479248047,
|
| 1531 |
+
"logits/rejected": -1.9403536319732666,
|
| 1532 |
+
"logps/chosen": -34.52521896362305,
|
| 1533 |
+
"logps/rejected": -36.773746490478516,
|
| 1534 |
+
"loss": 0.7554,
|
| 1535 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1536 |
+
"rewards/chosen": 0.26395466923713684,
|
| 1537 |
+
"rewards/margins": 0.4174926280975342,
|
| 1538 |
+
"rewards/rejected": -0.15353801846504211,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.57,
|
| 1543 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1544 |
+
"logits/chosen": -1.9193499088287354,
|
| 1545 |
+
"logits/rejected": -1.9140777587890625,
|
| 1546 |
+
"logps/chosen": -30.143762588500977,
|
| 1547 |
+
"logps/rejected": -36.39270782470703,
|
| 1548 |
+
"loss": 0.5759,
|
| 1549 |
+
"rewards/accuracies": 0.9375,
|
| 1550 |
+
"rewards/chosen": 0.2577562928199768,
|
| 1551 |
+
"rewards/margins": 0.37860342860221863,
|
| 1552 |
+
"rewards/rejected": -0.12084714323282242,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.6,
|
| 1557 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1558 |
+
"logits/chosen": -1.8965826034545898,
|
| 1559 |
+
"logits/rejected": -1.885602593421936,
|
| 1560 |
+
"logps/chosen": -30.77829933166504,
|
| 1561 |
+
"logps/rejected": -30.00567626953125,
|
| 1562 |
+
"loss": 0.4205,
|
| 1563 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1564 |
+
"rewards/chosen": 0.2911164164543152,
|
| 1565 |
+
"rewards/margins": 0.41289979219436646,
|
| 1566 |
+
"rewards/rejected": -0.12178337574005127,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.6,
|
| 1571 |
+
"eval_logits/chosen": -2.232767343521118,
|
| 1572 |
+
"eval_logits/rejected": -2.2278943061828613,
|
| 1573 |
+
"eval_logps/chosen": -34.03477478027344,
|
| 1574 |
+
"eval_logps/rejected": -37.57896423339844,
|
| 1575 |
+
"eval_loss": 2.8065927028656006,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5307309031486511,
|
| 1577 |
+
"eval_rewards/chosen": -6.714276969432831e-05,
|
| 1578 |
+
"eval_rewards/margins": 0.01863543689250946,
|
| 1579 |
+
"eval_rewards/rejected": -0.018702581524848938,
|
| 1580 |
+
"eval_runtime": 145.8056,
|
| 1581 |
+
"eval_samples_per_second": 2.352,
|
| 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": -2.00972843170166,
|
| 1589 |
+
"logits/rejected": -2.009073495864868,
|
| 1590 |
+
"logps/chosen": -30.8530330657959,
|
| 1591 |
+
"logps/rejected": -30.636966705322266,
|
| 1592 |
+
"loss": 0.521,
|
| 1593 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1594 |
+
"rewards/chosen": 0.29134875535964966,
|
| 1595 |
+
"rewards/margins": 0.4048939645290375,
|
| 1596 |
+
"rewards/rejected": -0.11354520171880722,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 2.65,
|
| 1601 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1602 |
+
"logits/chosen": -2.044478178024292,
|
| 1603 |
+
"logits/rejected": -2.045170783996582,
|
| 1604 |
+
"logps/chosen": -30.969898223876953,
|
| 1605 |
+
"logps/rejected": -31.586406707763672,
|
| 1606 |
+
"loss": 0.4631,
|
| 1607 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1608 |
+
"rewards/chosen": 0.29210424423217773,
|
| 1609 |
+
"rewards/margins": 0.4182845950126648,
|
| 1610 |
+
"rewards/rejected": -0.12618036568164825,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 2.68,
|
| 1615 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1616 |
+
"logits/chosen": -2.0170657634735107,
|
| 1617 |
+
"logits/rejected": -2.0201754570007324,
|
| 1618 |
+
"logps/chosen": -31.390050888061523,
|
| 1619 |
+
"logps/rejected": -28.406635284423828,
|
| 1620 |
+
"loss": 0.4956,
|
| 1621 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1622 |
+
"rewards/chosen": 0.26700589060783386,
|
| 1623 |
+
"rewards/margins": 0.39388108253479004,
|
| 1624 |
+
"rewards/rejected": -0.12687517702579498,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 2.7,
|
| 1629 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1630 |
+
"logits/chosen": -2.031101703643799,
|
| 1631 |
+
"logits/rejected": -2.0232481956481934,
|
| 1632 |
+
"logps/chosen": -30.17348861694336,
|
| 1633 |
+
"logps/rejected": -33.96366500854492,
|
| 1634 |
+
"loss": 0.3554,
|
| 1635 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1636 |
+
"rewards/chosen": 0.3013220429420471,
|
| 1637 |
+
"rewards/margins": 0.44035378098487854,
|
| 1638 |
+
"rewards/rejected": -0.139031782746315,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 2.73,
|
| 1643 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1644 |
+
"logits/chosen": -1.959812879562378,
|
| 1645 |
+
"logits/rejected": -1.9708335399627686,
|
| 1646 |
+
"logps/chosen": -32.60597610473633,
|
| 1647 |
+
"logps/rejected": -32.627235412597656,
|
| 1648 |
+
"loss": 0.446,
|
| 1649 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1650 |
+
"rewards/chosen": 0.28413859009742737,
|
| 1651 |
+
"rewards/margins": 0.3988059163093567,
|
| 1652 |
+
"rewards/rejected": -0.11466735601425171,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.75,
|
| 1657 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1658 |
+
"logits/chosen": -2.039461135864258,
|
| 1659 |
+
"logits/rejected": -2.0299925804138184,
|
| 1660 |
+
"logps/chosen": -32.57851028442383,
|
| 1661 |
+
"logps/rejected": -36.76939392089844,
|
| 1662 |
+
"loss": 0.7536,
|
| 1663 |
+
"rewards/accuracies": 0.9375,
|
| 1664 |
+
"rewards/chosen": 0.27668142318725586,
|
| 1665 |
+
"rewards/margins": 0.4080823063850403,
|
| 1666 |
+
"rewards/rejected": -0.13140086829662323,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 2.78,
|
| 1671 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1672 |
+
"logits/chosen": -1.9462810754776,
|
| 1673 |
+
"logits/rejected": -1.9480304718017578,
|
| 1674 |
+
"logps/chosen": -34.97504425048828,
|
| 1675 |
+
"logps/rejected": -33.058712005615234,
|
| 1676 |
+
"loss": 0.4457,
|
| 1677 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1678 |
+
"rewards/chosen": 0.3042721450328827,
|
| 1679 |
+
"rewards/margins": 0.41252559423446655,
|
| 1680 |
+
"rewards/rejected": -0.10825344175100327,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 2.81,
|
| 1685 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1686 |
+
"logits/chosen": -2.0834527015686035,
|
| 1687 |
+
"logits/rejected": -2.084644079208374,
|
| 1688 |
+
"logps/chosen": -32.06096649169922,
|
| 1689 |
+
"logps/rejected": -34.56989288330078,
|
| 1690 |
+
"loss": 0.3651,
|
| 1691 |
+
"rewards/accuracies": 1.0,
|
| 1692 |
+
"rewards/chosen": 0.29900139570236206,
|
| 1693 |
+
"rewards/margins": 0.43526363372802734,
|
| 1694 |
+
"rewards/rejected": -0.1362622082233429,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 2.83,
|
| 1699 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1700 |
+
"logits/chosen": -2.0493340492248535,
|
| 1701 |
+
"logits/rejected": -2.0534369945526123,
|
| 1702 |
+
"logps/chosen": -29.894922256469727,
|
| 1703 |
+
"logps/rejected": -33.94621658325195,
|
| 1704 |
+
"loss": 0.6786,
|
| 1705 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1706 |
+
"rewards/chosen": 0.2995173931121826,
|
| 1707 |
+
"rewards/margins": 0.41903191804885864,
|
| 1708 |
+
"rewards/rejected": -0.11951451003551483,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 2.86,
|
| 1713 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1714 |
+
"logits/chosen": -2.038386821746826,
|
| 1715 |
+
"logits/rejected": -2.0437779426574707,
|
| 1716 |
+
"logps/chosen": -32.619266510009766,
|
| 1717 |
+
"logps/rejected": -30.790191650390625,
|
| 1718 |
+
"loss": 0.5121,
|
| 1719 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1720 |
+
"rewards/chosen": 0.284921795129776,
|
| 1721 |
+
"rewards/margins": 0.401477575302124,
|
| 1722 |
+
"rewards/rejected": -0.1165558248758316,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 2.86,
|
| 1727 |
+
"eval_logits/chosen": -2.2309653759002686,
|
| 1728 |
+
"eval_logits/rejected": -2.226088523864746,
|
| 1729 |
+
"eval_logps/chosen": -34.027095794677734,
|
| 1730 |
+
"eval_logps/rejected": -37.578590393066406,
|
| 1731 |
+
"eval_loss": 2.7946248054504395,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5726743936538696,
|
| 1733 |
+
"eval_rewards/chosen": 0.0022365888580679893,
|
| 1734 |
+
"eval_rewards/margins": 0.020828478038311005,
|
| 1735 |
+
"eval_rewards/rejected": -0.01859188824892044,
|
| 1736 |
+
"eval_runtime": 145.9533,
|
| 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": -2.0310146808624268,
|
| 1745 |
+
"logits/rejected": -2.0306754112243652,
|
| 1746 |
+
"logps/chosen": -31.951345443725586,
|
| 1747 |
+
"logps/rejected": -31.370046615600586,
|
| 1748 |
+
"loss": 0.5264,
|
| 1749 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1750 |
+
"rewards/chosen": 0.29664164781570435,
|
| 1751 |
+
"rewards/margins": 0.40288034081459045,
|
| 1752 |
+
"rewards/rejected": -0.1062387004494667,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 2.91,
|
| 1757 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1758 |
+
"logits/chosen": -1.8752002716064453,
|
| 1759 |
+
"logits/rejected": -1.8767414093017578,
|
| 1760 |
+
"logps/chosen": -33.654136657714844,
|
| 1761 |
+
"logps/rejected": -33.44560623168945,
|
| 1762 |
+
"loss": 0.7694,
|
| 1763 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1764 |
+
"rewards/chosen": 0.27325233817100525,
|
| 1765 |
+
"rewards/margins": 0.3945540189743042,
|
| 1766 |
+
"rewards/rejected": -0.12130165100097656,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 2.94,
|
| 1771 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1772 |
+
"logits/chosen": -1.9149949550628662,
|
| 1773 |
+
"logits/rejected": -1.912196397781372,
|
| 1774 |
+
"logps/chosen": -30.21531105041504,
|
| 1775 |
+
"logps/rejected": -31.496496200561523,
|
| 1776 |
+
"loss": 0.8975,
|
| 1777 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1778 |
+
"rewards/chosen": 0.23892304301261902,
|
| 1779 |
+
"rewards/margins": 0.33815068006515503,
|
| 1780 |
+
"rewards/rejected": -0.09922759234905243,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 2.96,
|
| 1785 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1786 |
+
"logits/chosen": -2.0512359142303467,
|
| 1787 |
+
"logits/rejected": -2.045522928237915,
|
| 1788 |
+
"logps/chosen": -31.9990291595459,
|
| 1789 |
+
"logps/rejected": -33.32525634765625,
|
| 1790 |
+
"loss": 0.3862,
|
| 1791 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1792 |
+
"rewards/chosen": 0.31222957372665405,
|
| 1793 |
+
"rewards/margins": 0.42360392212867737,
|
| 1794 |
+
"rewards/rejected": -0.11137431859970093,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 2.99,
|
| 1799 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1800 |
+
"logits/chosen": -1.898686170578003,
|
| 1801 |
+
"logits/rejected": -1.8885414600372314,
|
| 1802 |
+
"logps/chosen": -32.57102584838867,
|
| 1803 |
+
"logps/rejected": -31.68793296813965,
|
| 1804 |
+
"loss": 0.5573,
|
| 1805 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1806 |
+
"rewards/chosen": 0.2642134130001068,
|
| 1807 |
+
"rewards/margins": 0.3828732967376709,
|
| 1808 |
+
"rewards/rejected": -0.11865979433059692,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.01,
|
| 1813 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1814 |
+
"logits/chosen": -1.9817955493927002,
|
| 1815 |
+
"logits/rejected": -1.981529951095581,
|
| 1816 |
+
"logps/chosen": -32.62514877319336,
|
| 1817 |
+
"logps/rejected": -31.584686279296875,
|
| 1818 |
+
"loss": 0.309,
|
| 1819 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1820 |
+
"rewards/chosen": 0.3169806897640228,
|
| 1821 |
+
"rewards/margins": 0.43195247650146484,
|
| 1822 |
+
"rewards/rejected": -0.11497180163860321,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.04,
|
| 1827 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1828 |
+
"logits/chosen": -2.052461624145508,
|
| 1829 |
+
"logits/rejected": -2.0468575954437256,
|
| 1830 |
+
"logps/chosen": -32.544010162353516,
|
| 1831 |
+
"logps/rejected": -33.3515739440918,
|
| 1832 |
+
"loss": 0.3583,
|
| 1833 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1834 |
+
"rewards/chosen": 0.29458826780319214,
|
| 1835 |
+
"rewards/margins": 0.4176958203315735,
|
| 1836 |
+
"rewards/rejected": -0.12310749292373657,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.06,
|
| 1841 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1842 |
+
"logits/chosen": -1.9057941436767578,
|
| 1843 |
+
"logits/rejected": -1.913578987121582,
|
| 1844 |
+
"logps/chosen": -32.369415283203125,
|
| 1845 |
+
"logps/rejected": -32.830265045166016,
|
| 1846 |
+
"loss": 0.293,
|
| 1847 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1848 |
+
"rewards/chosen": 0.2890740931034088,
|
| 1849 |
+
"rewards/margins": 0.42972269654273987,
|
| 1850 |
+
"rewards/rejected": -0.14064857363700867,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.09,
|
| 1855 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1856 |
+
"logits/chosen": -1.9869006872177124,
|
| 1857 |
+
"logits/rejected": -1.9810779094696045,
|
| 1858 |
+
"logps/chosen": -31.156478881835938,
|
| 1859 |
+
"logps/rejected": -33.652259826660156,
|
| 1860 |
+
"loss": 0.3685,
|
| 1861 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1862 |
+
"rewards/chosen": 0.30514153838157654,
|
| 1863 |
+
"rewards/margins": 0.4157847464084625,
|
| 1864 |
+
"rewards/rejected": -0.11064320802688599,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.12,
|
| 1869 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1870 |
+
"logits/chosen": -1.9249032735824585,
|
| 1871 |
+
"logits/rejected": -1.9270884990692139,
|
| 1872 |
+
"logps/chosen": -32.53572082519531,
|
| 1873 |
+
"logps/rejected": -33.25934600830078,
|
| 1874 |
+
"loss": 0.4011,
|
| 1875 |
+
"rewards/accuracies": 0.9375,
|
| 1876 |
+
"rewards/chosen": 0.2864174544811249,
|
| 1877 |
+
"rewards/margins": 0.43620601296424866,
|
| 1878 |
+
"rewards/rejected": -0.14978858828544617,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.12,
|
| 1883 |
+
"eval_logits/chosen": -2.2320587635040283,
|
| 1884 |
+
"eval_logits/rejected": -2.22719407081604,
|
| 1885 |
+
"eval_logps/chosen": -34.03776931762695,
|
| 1886 |
+
"eval_logps/rejected": -37.59090042114258,
|
| 1887 |
+
"eval_loss": 2.771493434906006,
|
| 1888 |
+
"eval_rewards/accuracies": 0.5485880374908447,
|
| 1889 |
+
"eval_rewards/chosen": -0.0009655026951804757,
|
| 1890 |
+
"eval_rewards/margins": 0.021318301558494568,
|
| 1891 |
+
"eval_rewards/rejected": -0.022283805534243584,
|
| 1892 |
+
"eval_runtime": 145.9177,
|
| 1893 |
+
"eval_samples_per_second": 2.351,
|
| 1894 |
+
"eval_steps_per_second": 0.295,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 3.14,
|
| 1899 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1900 |
+
"logits/chosen": -1.8565280437469482,
|
| 1901 |
+
"logits/rejected": -1.8553797006607056,
|
| 1902 |
+
"logps/chosen": -35.000247955322266,
|
| 1903 |
+
"logps/rejected": -33.142433166503906,
|
| 1904 |
+
"loss": 0.3182,
|
| 1905 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1906 |
+
"rewards/chosen": 0.3156726062297821,
|
| 1907 |
+
"rewards/margins": 0.44559246301651,
|
| 1908 |
+
"rewards/rejected": -0.1299198716878891,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.17,
|
| 1913 |
+
"learning_rate": 6.293013489185315e-07,
|
| 1914 |
+
"logits/chosen": -2.0334620475769043,
|
| 1915 |
+
"logits/rejected": -2.027405261993408,
|
| 1916 |
+
"logps/chosen": -30.731725692749023,
|
| 1917 |
+
"logps/rejected": -33.369632720947266,
|
| 1918 |
+
"loss": 0.3014,
|
| 1919 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1920 |
+
"rewards/chosen": 0.2915892004966736,
|
| 1921 |
+
"rewards/margins": 0.4389171004295349,
|
| 1922 |
+
"rewards/rejected": -0.14732789993286133,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.19,
|
| 1927 |
+
"learning_rate": 5.921934551632086e-07,
|
| 1928 |
+
"logits/chosen": -1.880632996559143,
|
| 1929 |
+
"logits/rejected": -1.8688300848007202,
|
| 1930 |
+
"logps/chosen": -33.38051986694336,
|
| 1931 |
+
"logps/rejected": -33.32735061645508,
|
| 1932 |
+
"loss": 0.35,
|
| 1933 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1934 |
+
"rewards/chosen": 0.30154940485954285,
|
| 1935 |
+
"rewards/margins": 0.4316067695617676,
|
| 1936 |
+
"rewards/rejected": -0.13005733489990234,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.22,
|
| 1941 |
+
"learning_rate": 5.560657034652405e-07,
|
| 1942 |
+
"logits/chosen": -1.9715722799301147,
|
| 1943 |
+
"logits/rejected": -1.9645836353302002,
|
| 1944 |
+
"logps/chosen": -29.689788818359375,
|
| 1945 |
+
"logps/rejected": -29.28457260131836,
|
| 1946 |
+
"loss": 0.3088,
|
| 1947 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1948 |
+
"rewards/chosen": 0.3031342923641205,
|
| 1949 |
+
"rewards/margins": 0.4282376170158386,
|
| 1950 |
+
"rewards/rejected": -0.12510332465171814,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.25,
|
| 1955 |
+
"learning_rate": 5.2093665457911e-07,
|
| 1956 |
+
"logits/chosen": -2.005354881286621,
|
| 1957 |
+
"logits/rejected": -2.011704206466675,
|
| 1958 |
+
"logps/chosen": -34.295631408691406,
|
| 1959 |
+
"logps/rejected": -31.874095916748047,
|
| 1960 |
+
"loss": 0.3386,
|
| 1961 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1962 |
+
"rewards/chosen": 0.3299277424812317,
|
| 1963 |
+
"rewards/margins": 0.4366631507873535,
|
| 1964 |
+
"rewards/rejected": -0.10673542320728302,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 3.27,
|
| 1969 |
+
"learning_rate": 4.868243561723535e-07,
|
| 1970 |
+
"logits/chosen": -1.9740073680877686,
|
| 1971 |
+
"logits/rejected": -1.9745842218399048,
|
| 1972 |
+
"logps/chosen": -32.05846405029297,
|
| 1973 |
+
"logps/rejected": -33.33855438232422,
|
| 1974 |
+
"loss": 0.2856,
|
| 1975 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1976 |
+
"rewards/chosen": 0.29102879762649536,
|
| 1977 |
+
"rewards/margins": 0.44813448190689087,
|
| 1978 |
+
"rewards/rejected": -0.1571057140827179,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 3.3,
|
| 1983 |
+
"learning_rate": 4.537463335535161e-07,
|
| 1984 |
+
"logits/chosen": -1.9194390773773193,
|
| 1985 |
+
"logits/rejected": -1.9188978672027588,
|
| 1986 |
+
"logps/chosen": -32.075077056884766,
|
| 1987 |
+
"logps/rejected": -33.93248748779297,
|
| 1988 |
+
"loss": 0.3179,
|
| 1989 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1990 |
+
"rewards/chosen": 0.3047292232513428,
|
| 1991 |
+
"rewards/margins": 0.4282926917076111,
|
| 1992 |
+
"rewards/rejected": -0.12356343120336533,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 3.32,
|
| 1997 |
+
"learning_rate": 4.217195806684629e-07,
|
| 1998 |
+
"logits/chosen": -1.8014293909072876,
|
| 1999 |
+
"logits/rejected": -1.7966254949569702,
|
| 2000 |
+
"logps/chosen": -34.1694221496582,
|
| 2001 |
+
"logps/rejected": -31.367359161376953,
|
| 2002 |
+
"loss": 0.2987,
|
| 2003 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2004 |
+
"rewards/chosen": 0.32040566205978394,
|
| 2005 |
+
"rewards/margins": 0.45266634225845337,
|
| 2006 |
+
"rewards/rejected": -0.13226068019866943,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 3.35,
|
| 2011 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2012 |
+
"logits/chosen": -2.008068561553955,
|
| 2013 |
+
"logits/rejected": -1.9922069311141968,
|
| 2014 |
+
"logps/chosen": -33.34940719604492,
|
| 2015 |
+
"logps/rejected": -35.33498001098633,
|
| 2016 |
+
"loss": 0.2358,
|
| 2017 |
+
"rewards/accuracies": 1.0,
|
| 2018 |
+
"rewards/chosen": 0.29945045709609985,
|
| 2019 |
+
"rewards/margins": 0.47034788131713867,
|
| 2020 |
+
"rewards/rejected": -0.1708974540233612,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 3.38,
|
| 2025 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2026 |
+
"logits/chosen": -1.950233817100525,
|
| 2027 |
+
"logits/rejected": -1.9552303552627563,
|
| 2028 |
+
"logps/chosen": -32.515785217285156,
|
| 2029 |
+
"logps/rejected": -36.63795852661133,
|
| 2030 |
+
"loss": 0.2672,
|
| 2031 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2032 |
+
"rewards/chosen": 0.30055129528045654,
|
| 2033 |
+
"rewards/margins": 0.46097177267074585,
|
| 2034 |
+
"rewards/rejected": -0.1604204922914505,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 3.38,
|
| 2039 |
+
"eval_logits/chosen": -2.2319700717926025,
|
| 2040 |
+
"eval_logits/rejected": -2.227097749710083,
|
| 2041 |
+
"eval_logps/chosen": -34.03395080566406,
|
| 2042 |
+
"eval_logps/rejected": -37.574100494384766,
|
| 2043 |
+
"eval_loss": 2.8251254558563232,
|
| 2044 |
+
"eval_rewards/accuracies": 0.5631229281425476,
|
| 2045 |
+
"eval_rewards/chosen": 0.00018060104048345238,
|
| 2046 |
+
"eval_rewards/margins": 0.01742532290518284,
|
| 2047 |
+
"eval_rewards/rejected": -0.017244720831513405,
|
| 2048 |
+
"eval_runtime": 145.8462,
|
| 2049 |
+
"eval_samples_per_second": 2.352,
|
| 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": -1.921814203262329,
|
| 2057 |
+
"logits/rejected": -1.921927809715271,
|
| 2058 |
+
"logps/chosen": -35.261505126953125,
|
| 2059 |
+
"logps/rejected": -37.443077087402344,
|
| 2060 |
+
"loss": 0.4227,
|
| 2061 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2062 |
+
"rewards/chosen": 0.2947538495063782,
|
| 2063 |
+
"rewards/margins": 0.4024432599544525,
|
| 2064 |
+
"rewards/rejected": -0.10768942534923553,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 3.43,
|
| 2069 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2070 |
+
"logits/chosen": -2.0089898109436035,
|
| 2071 |
+
"logits/rejected": -2.007404327392578,
|
| 2072 |
+
"logps/chosen": -31.107452392578125,
|
| 2073 |
+
"logps/rejected": -31.33074378967285,
|
| 2074 |
+
"loss": 0.3393,
|
| 2075 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2076 |
+
"rewards/chosen": 0.2959212362766266,
|
| 2077 |
+
"rewards/margins": 0.4091377258300781,
|
| 2078 |
+
"rewards/rejected": -0.11321648210287094,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 3.45,
|
| 2083 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2084 |
+
"logits/chosen": -1.913175344467163,
|
| 2085 |
+
"logits/rejected": -1.9146407842636108,
|
| 2086 |
+
"logps/chosen": -33.32870101928711,
|
| 2087 |
+
"logps/rejected": -34.072715759277344,
|
| 2088 |
+
"loss": 0.23,
|
| 2089 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2090 |
+
"rewards/chosen": 0.31366580724716187,
|
| 2091 |
+
"rewards/margins": 0.44087910652160645,
|
| 2092 |
+
"rewards/rejected": -0.12721329927444458,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 3.48,
|
| 2097 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2098 |
+
"logits/chosen": -1.9741986989974976,
|
| 2099 |
+
"logits/rejected": -1.9729766845703125,
|
| 2100 |
+
"logps/chosen": -30.287155151367188,
|
| 2101 |
+
"logps/rejected": -32.518028259277344,
|
| 2102 |
+
"loss": 0.3725,
|
| 2103 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2104 |
+
"rewards/chosen": 0.27345365285873413,
|
| 2105 |
+
"rewards/margins": 0.43126893043518066,
|
| 2106 |
+
"rewards/rejected": -0.15781526267528534,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 3.51,
|
| 2111 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2112 |
+
"logits/chosen": -1.9914159774780273,
|
| 2113 |
+
"logits/rejected": -1.9765985012054443,
|
| 2114 |
+
"logps/chosen": -33.417686462402344,
|
| 2115 |
+
"logps/rejected": -36.712074279785156,
|
| 2116 |
+
"loss": 0.3691,
|
| 2117 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2118 |
+
"rewards/chosen": 0.26969602704048157,
|
| 2119 |
+
"rewards/margins": 0.403988778591156,
|
| 2120 |
+
"rewards/rejected": -0.1342928111553192,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 3.53,
|
| 2125 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2126 |
+
"logits/chosen": -1.9975001811981201,
|
| 2127 |
+
"logits/rejected": -2.010082960128784,
|
| 2128 |
+
"logps/chosen": -31.060623168945312,
|
| 2129 |
+
"logps/rejected": -32.55187225341797,
|
| 2130 |
+
"loss": 0.3585,
|
| 2131 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2132 |
+
"rewards/chosen": 0.3023538291454315,
|
| 2133 |
+
"rewards/margins": 0.3959945738315582,
|
| 2134 |
+
"rewards/rejected": -0.0936407595872879,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.56,
|
| 2139 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2140 |
+
"logits/chosen": -2.0844500064849854,
|
| 2141 |
+
"logits/rejected": -2.083676815032959,
|
| 2142 |
+
"logps/chosen": -31.054336547851562,
|
| 2143 |
+
"logps/rejected": -36.317108154296875,
|
| 2144 |
+
"loss": 0.2726,
|
| 2145 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2146 |
+
"rewards/chosen": 0.29282861948013306,
|
| 2147 |
+
"rewards/margins": 0.43869519233703613,
|
| 2148 |
+
"rewards/rejected": -0.14586657285690308,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 3.58,
|
| 2153 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2154 |
+
"logits/chosen": -1.9504215717315674,
|
| 2155 |
+
"logits/rejected": -1.9583799839019775,
|
| 2156 |
+
"logps/chosen": -33.26082229614258,
|
| 2157 |
+
"logps/rejected": -37.89590072631836,
|
| 2158 |
+
"loss": 0.4793,
|
| 2159 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2160 |
+
"rewards/chosen": 0.2758983075618744,
|
| 2161 |
+
"rewards/margins": 0.4097749590873718,
|
| 2162 |
+
"rewards/rejected": -0.13387663662433624,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 3.61,
|
| 2167 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2168 |
+
"logits/chosen": -1.9037997722625732,
|
| 2169 |
+
"logits/rejected": -1.899585485458374,
|
| 2170 |
+
"logps/chosen": -33.69878005981445,
|
| 2171 |
+
"logps/rejected": -31.925487518310547,
|
| 2172 |
+
"loss": 0.4337,
|
| 2173 |
+
"rewards/accuracies": 0.9375,
|
| 2174 |
+
"rewards/chosen": 0.3205293118953705,
|
| 2175 |
+
"rewards/margins": 0.4189586043357849,
|
| 2176 |
+
"rewards/rejected": -0.09842932224273682,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 3.64,
|
| 2181 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2182 |
+
"logits/chosen": -2.0691137313842773,
|
| 2183 |
+
"logits/rejected": -2.066561460494995,
|
| 2184 |
+
"logps/chosen": -30.29781150817871,
|
| 2185 |
+
"logps/rejected": -32.60316848754883,
|
| 2186 |
+
"loss": 0.2484,
|
| 2187 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2188 |
+
"rewards/chosen": 0.2996208071708679,
|
| 2189 |
+
"rewards/margins": 0.43723049759864807,
|
| 2190 |
+
"rewards/rejected": -0.13760972023010254,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 3.64,
|
| 2195 |
+
"eval_logits/chosen": -2.2316062450408936,
|
| 2196 |
+
"eval_logits/rejected": -2.2267301082611084,
|
| 2197 |
+
"eval_logps/chosen": -34.033653259277344,
|
| 2198 |
+
"eval_logps/rejected": -37.583335876464844,
|
| 2199 |
+
"eval_loss": 2.791294813156128,
|
| 2200 |
+
"eval_rewards/accuracies": 0.5664451718330383,
|
| 2201 |
+
"eval_rewards/chosen": 0.00026977824745699763,
|
| 2202 |
+
"eval_rewards/margins": 0.02028520777821541,
|
| 2203 |
+
"eval_rewards/rejected": -0.020015425980091095,
|
| 2204 |
+
"eval_runtime": 145.9074,
|
| 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": -2.0081191062927246,
|
| 2213 |
+
"logits/rejected": -2.003680944442749,
|
| 2214 |
+
"logps/chosen": -32.60419464111328,
|
| 2215 |
+
"logps/rejected": -29.99764060974121,
|
| 2216 |
+
"loss": 0.3034,
|
| 2217 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2218 |
+
"rewards/chosen": 0.3115007281303406,
|
| 2219 |
+
"rewards/margins": 0.4444548487663269,
|
| 2220 |
+
"rewards/rejected": -0.1329541653394699,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 3.69,
|
| 2225 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2226 |
+
"logits/chosen": -1.9665706157684326,
|
| 2227 |
+
"logits/rejected": -1.9677642583847046,
|
| 2228 |
+
"logps/chosen": -32.38178253173828,
|
| 2229 |
+
"logps/rejected": -35.331478118896484,
|
| 2230 |
+
"loss": 0.4047,
|
| 2231 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2232 |
+
"rewards/chosen": 0.2949289083480835,
|
| 2233 |
+
"rewards/margins": 0.41717034578323364,
|
| 2234 |
+
"rewards/rejected": -0.12224143743515015,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 3.71,
|
| 2239 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2240 |
+
"logits/chosen": -1.9199825525283813,
|
| 2241 |
+
"logits/rejected": -1.9131149053573608,
|
| 2242 |
+
"logps/chosen": -33.19044876098633,
|
| 2243 |
+
"logps/rejected": -37.124359130859375,
|
| 2244 |
+
"loss": 0.3234,
|
| 2245 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2246 |
+
"rewards/chosen": 0.28539571166038513,
|
| 2247 |
+
"rewards/margins": 0.4158816337585449,
|
| 2248 |
+
"rewards/rejected": -0.13048596680164337,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 3.74,
|
| 2253 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2254 |
+
"logits/chosen": -1.931983232498169,
|
| 2255 |
+
"logits/rejected": -1.9264585971832275,
|
| 2256 |
+
"logps/chosen": -28.360937118530273,
|
| 2257 |
+
"logps/rejected": -32.15396499633789,
|
| 2258 |
+
"loss": 0.5022,
|
| 2259 |
+
"rewards/accuracies": 0.9375,
|
| 2260 |
+
"rewards/chosen": 0.285691499710083,
|
| 2261 |
+
"rewards/margins": 0.4126841127872467,
|
| 2262 |
+
"rewards/rejected": -0.12699264287948608,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 3.77,
|
| 2267 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2268 |
+
"logits/chosen": -1.9492918252944946,
|
| 2269 |
+
"logits/rejected": -1.9388021230697632,
|
| 2270 |
+
"logps/chosen": -31.5377140045166,
|
| 2271 |
+
"logps/rejected": -35.85572052001953,
|
| 2272 |
+
"loss": 0.3368,
|
| 2273 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2274 |
+
"rewards/chosen": 0.2831452488899231,
|
| 2275 |
+
"rewards/margins": 0.4542534351348877,
|
| 2276 |
+
"rewards/rejected": -0.1711081564426422,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 3.79,
|
| 2281 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2282 |
+
"logits/chosen": -1.9281123876571655,
|
| 2283 |
+
"logits/rejected": -1.9251676797866821,
|
| 2284 |
+
"logps/chosen": -32.314857482910156,
|
| 2285 |
+
"logps/rejected": -31.260883331298828,
|
| 2286 |
+
"loss": 0.4229,
|
| 2287 |
+
"rewards/accuracies": 0.9375,
|
| 2288 |
+
"rewards/chosen": 0.29847532510757446,
|
| 2289 |
+
"rewards/margins": 0.40949302911758423,
|
| 2290 |
+
"rewards/rejected": -0.11101765930652618,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 3.82,
|
| 2295 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2296 |
+
"logits/chosen": -1.989189863204956,
|
| 2297 |
+
"logits/rejected": -1.9958044290542603,
|
| 2298 |
+
"logps/chosen": -31.037267684936523,
|
| 2299 |
+
"logps/rejected": -33.406105041503906,
|
| 2300 |
+
"loss": 0.3242,
|
| 2301 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2302 |
+
"rewards/chosen": 0.3095727562904358,
|
| 2303 |
+
"rewards/margins": 0.44267648458480835,
|
| 2304 |
+
"rewards/rejected": -0.13310377299785614,
|
| 2305 |
+
"step": 1470
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 3.84,
|
| 2309 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2310 |
+
"logits/chosen": -1.9884917736053467,
|
| 2311 |
+
"logits/rejected": -1.9990545511245728,
|
| 2312 |
+
"logps/chosen": -30.706201553344727,
|
| 2313 |
+
"logps/rejected": -30.090484619140625,
|
| 2314 |
+
"loss": 0.339,
|
| 2315 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2316 |
+
"rewards/chosen": 0.29671651124954224,
|
| 2317 |
+
"rewards/margins": 0.4352938234806061,
|
| 2318 |
+
"rewards/rejected": -0.13857729732990265,
|
| 2319 |
+
"step": 1480
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 3.87,
|
| 2323 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2324 |
+
"logits/chosen": -1.9214226007461548,
|
| 2325 |
+
"logits/rejected": -1.9142558574676514,
|
| 2326 |
+
"logps/chosen": -32.70989227294922,
|
| 2327 |
+
"logps/rejected": -32.67903518676758,
|
| 2328 |
+
"loss": 0.3616,
|
| 2329 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2330 |
+
"rewards/chosen": 0.2839769423007965,
|
| 2331 |
+
"rewards/margins": 0.4245426058769226,
|
| 2332 |
+
"rewards/rejected": -0.1405656784772873,
|
| 2333 |
+
"step": 1490
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 3.9,
|
| 2337 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2338 |
+
"logits/chosen": -1.9736502170562744,
|
| 2339 |
+
"logits/rejected": -1.9704258441925049,
|
| 2340 |
+
"logps/chosen": -32.689056396484375,
|
| 2341 |
+
"logps/rejected": -33.725563049316406,
|
| 2342 |
+
"loss": 0.3309,
|
| 2343 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2344 |
+
"rewards/chosen": 0.28095489740371704,
|
| 2345 |
+
"rewards/margins": 0.42055588960647583,
|
| 2346 |
+
"rewards/rejected": -0.1396009922027588,
|
| 2347 |
+
"step": 1500
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.9,
|
| 2351 |
+
"eval_logits/chosen": -2.231811761856079,
|
| 2352 |
+
"eval_logits/rejected": -2.226945400238037,
|
| 2353 |
+
"eval_logps/chosen": -34.02287292480469,
|
| 2354 |
+
"eval_logps/rejected": -37.57923889160156,
|
| 2355 |
+
"eval_loss": 2.76727557182312,
|
| 2356 |
+
"eval_rewards/accuracies": 0.5722591280937195,
|
| 2357 |
+
"eval_rewards/chosen": 0.003503018757328391,
|
| 2358 |
+
"eval_rewards/margins": 0.022289730608463287,
|
| 2359 |
+
"eval_rewards/rejected": -0.018786713480949402,
|
| 2360 |
+
"eval_runtime": 145.8047,
|
| 2361 |
+
"eval_samples_per_second": 2.352,
|
| 2362 |
+
"eval_steps_per_second": 0.295,
|
| 2363 |
+
"step": 1500
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 3.92,
|
| 2367 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2368 |
+
"logits/chosen": -2.0533382892608643,
|
| 2369 |
+
"logits/rejected": -2.053973436355591,
|
| 2370 |
+
"logps/chosen": -28.94424819946289,
|
| 2371 |
+
"logps/rejected": -33.25762176513672,
|
| 2372 |
+
"loss": 0.5224,
|
| 2373 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2374 |
+
"rewards/chosen": 0.26160866022109985,
|
| 2375 |
+
"rewards/margins": 0.4167584478855133,
|
| 2376 |
+
"rewards/rejected": -0.15514974296092987,
|
| 2377 |
+
"step": 1510
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.95,
|
| 2381 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2382 |
+
"logits/chosen": -1.9069092273712158,
|
| 2383 |
+
"logits/rejected": -1.9067671298980713,
|
| 2384 |
+
"logps/chosen": -32.138919830322266,
|
| 2385 |
+
"logps/rejected": -34.360389709472656,
|
| 2386 |
+
"loss": 0.4383,
|
| 2387 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2388 |
+
"rewards/chosen": 0.2708854377269745,
|
| 2389 |
+
"rewards/margins": 0.41688376665115356,
|
| 2390 |
+
"rewards/rejected": -0.14599832892417908,
|
| 2391 |
+
"step": 1520
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 3.97,
|
| 2395 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2396 |
+
"logits/chosen": -2.0585055351257324,
|
| 2397 |
+
"logits/rejected": -2.055807590484619,
|
| 2398 |
+
"logps/chosen": -30.587963104248047,
|
| 2399 |
+
"logps/rejected": -31.973073959350586,
|
| 2400 |
+
"loss": 0.4131,
|
| 2401 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2402 |
+
"rewards/chosen": 0.2658289968967438,
|
| 2403 |
+
"rewards/margins": 0.41265568137168884,
|
| 2404 |
+
"rewards/rejected": -0.14682674407958984,
|
| 2405 |
+
"step": 1530
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 4.0,
|
| 2409 |
+
"learning_rate": 0.0,
|
| 2410 |
+
"logits/chosen": -2.041874885559082,
|
| 2411 |
+
"logits/rejected": -2.042999744415283,
|
| 2412 |
+
"logps/chosen": -29.7122859954834,
|
| 2413 |
+
"logps/rejected": -30.280029296875,
|
| 2414 |
+
"loss": 0.2918,
|
| 2415 |
+
"rewards/accuracies": 0.98333340883255,
|
| 2416 |
+
"rewards/chosen": 0.27229610085487366,
|
| 2417 |
+
"rewards/margins": 0.4192503094673157,
|
| 2418 |
+
"rewards/rejected": -0.14695420861244202,
|
| 2419 |
+
"step": 1540
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 4.0,
|
| 2423 |
+
"step": 1540,
|
| 2424 |
"total_flos": 0.0,
|
| 2425 |
+
"train_loss": 0.5763176552661053,
|
| 2426 |
+
"train_runtime": 10799.143,
|
| 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,
|