Commit
•
ab5fba1
1
Parent(s):
6fe47e2
Model save
Browse files- README.md +32 -22
- adapter_model.safetensors +1 -1
- all_results.json +5 -5
- train_results.json +5 -5
- trainer_state.json +2030 -90
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
tags:
|
| 4 |
-
- alignment-handbook
|
| 5 |
- trl
|
| 6 |
- dpo
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
| 9 |
-
datasets:
|
| 10 |
-
- hugodk-sch/aftonposten_title_prefs
|
| 11 |
model-index:
|
| 12 |
- name: aftonposten-6b-align-scan
|
| 13 |
results: []
|
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
|
| 19 |
# aftonposten-6b-align-scan
|
| 20 |
|
| 21 |
-
This model is a fine-tuned version of [
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 0.
|
| 24 |
-
- Rewards/chosen: 0.
|
| 25 |
-
- Rewards/rejected: 0.
|
| 26 |
-
- Rewards/accuracies: 0.
|
| 27 |
-
- Rewards/margins: 0.
|
| 28 |
-
- Logps/rejected: -37.
|
| 29 |
-
- Logps/chosen: -
|
| 30 |
-
- Logits/rejected: -2.
|
| 31 |
-
- Logits/chosen: -2.
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
@@ -57,21 +55,33 @@ The following hyperparameters were used during training:
|
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
| 59 |
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
-
- num_epochs:
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
-
| Training Loss | Epoch | Step |
|
| 65 |
-
|
| 66 |
-
| 0.3038 | 0.26 | 100 |
|
| 67 |
-
| 0.3157 | 0.52 | 200 |
|
| 68 |
-
| 0.2156 | 0.78 | 300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
| 72 |
|
| 73 |
-
- PEFT 0.
|
| 74 |
-
- Transformers 4.
|
| 75 |
- Pytorch 2.1.2+cu121
|
| 76 |
-
- Datasets 2.
|
| 77 |
- Tokenizers 0.15.1
|
|
|
|
| 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 the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.3543
|
| 22 |
+
- Rewards/chosen: 0.1332
|
| 23 |
+
- Rewards/rejected: 0.1192
|
| 24 |
+
- Rewards/accuracies: 0.5486
|
| 25 |
+
- Rewards/margins: 0.0139
|
| 26 |
+
- Logps/rejected: -37.3842
|
| 27 |
+
- Logps/chosen: -33.8866
|
| 28 |
+
- Logits/rejected: -2.2421
|
| 29 |
+
- Logits/chosen: -2.2469
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 55 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
- lr_scheduler_type: cosine
|
| 57 |
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 4
|
| 59 |
|
| 60 |
### Training results
|
| 61 |
|
| 62 |
+
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
| 63 |
+
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
| 64 |
+
| 0.3038 | 0.26 | 100 | -2.2372 | -2.2324 | -34.0128 | -37.5115 | 0.3512 | 0.5424 | 0.0196 | 0.0150 | 0.0046 |
|
| 65 |
+
| 0.3157 | 0.52 | 200 | -2.2371 | -2.2322 | -34.0181 | -37.5184 | 0.3716 | 0.5245 | 0.0148 | 0.0164 | -0.0016 |
|
| 66 |
+
| 0.2156 | 0.78 | 300 | -2.2364 | -2.2316 | -34.0143 | -37.4970 | 0.3845 | 0.4934 | 0.0182 | 0.0005 | 0.0177 |
|
| 67 |
+
| 0.4084 | 1.04 | 400 | 0.4059 | 0.0705 | 0.0718 | 0.5066 | -0.0013 | -37.4369 | -33.9562 | -2.2400 | -2.2448 |
|
| 68 |
+
| 0.2788 | 1.3 | 500 | 0.3866 | 0.0701 | 0.0576 | 0.5191 | 0.0125 | -37.4526 | -33.9566 | -2.2356 | -2.2405 |
|
| 69 |
+
| 0.3874 | 1.56 | 600 | 0.4265 | 0.0711 | 0.0890 | 0.4726 | -0.0180 | -37.4177 | -33.9556 | -2.2421 | -2.2470 |
|
| 70 |
+
| 0.2695 | 1.82 | 700 | 0.4028 | 0.0816 | 0.0876 | 0.5079 | -0.0060 | -37.4193 | -33.9439 | -2.2429 | -2.2478 |
|
| 71 |
+
| 0.1725 | 2.08 | 800 | 0.4083 | 0.0967 | 0.1077 | 0.4821 | -0.0110 | -37.3970 | -33.9271 | -2.2415 | -2.2463 |
|
| 72 |
+
| 0.2502 | 2.34 | 900 | 0.4099 | 0.1154 | 0.1311 | 0.4900 | -0.0157 | -37.3709 | -33.9064 | -2.2438 | -2.2487 |
|
| 73 |
+
| 0.1529 | 2.6 | 1000 | 0.3879 | 0.1222 | 0.1257 | 0.5216 | -0.0034 | -37.3770 | -33.8988 | -2.2428 | -2.2477 |
|
| 74 |
+
| 0.1583 | 2.86 | 1100 | 0.3968 | 0.1193 | 0.1250 | 0.4875 | -0.0057 | -37.3777 | -33.9020 | -2.2433 | -2.2482 |
|
| 75 |
+
| 0.113 | 3.12 | 1200 | 0.3849 | 0.1137 | 0.1163 | 0.4784 | -0.0025 | -37.3874 | -33.9082 | -2.2421 | -2.2470 |
|
| 76 |
+
| 0.0937 | 3.38 | 1300 | 0.3738 | 0.1235 | 0.1177 | 0.5046 | 0.0058 | -37.3859 | -33.8973 | -2.2423 | -2.2472 |
|
| 77 |
+
| 0.0815 | 3.64 | 1400 | 0.3595 | 0.1338 | 0.1197 | 0.5224 | 0.0141 | -37.3836 | -33.8859 | -2.2427 | -2.2476 |
|
| 78 |
+
| 0.0757 | 3.9 | 1500 | 0.3543 | 0.1332 | 0.1192 | 0.5486 | 0.0139 | -37.3842 | -33.8866 | -2.2421 | -2.2469 |
|
| 79 |
|
| 80 |
|
| 81 |
### Framework versions
|
| 82 |
|
| 83 |
+
- PEFT 0.10.0
|
| 84 |
+
- Transformers 4.39.0.dev0
|
| 85 |
- Pytorch 2.1.2+cu121
|
| 86 |
+
- Datasets 2.14.6
|
| 87 |
- Tokenizers 0.15.1
|
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 sha256:6204b6d9975755eaf1dfc0eeb539db7dfda9602c6eb76efa388fc044fba516b9
|
| 3 |
size 176183216
|
all_results.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"eval_logits/chosen": -2.238292694091797,
|
| 4 |
"eval_logits/rejected": -2.2334160804748535,
|
| 5 |
"eval_logps/chosen": -34.009239196777344,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.355,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
-
"train_loss": 0.
|
| 17 |
-
"train_runtime":
|
| 18 |
"train_samples": 3079,
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
"eval_logits/chosen": -2.238292694091797,
|
| 4 |
"eval_logits/rejected": -2.2334160804748535,
|
| 5 |
"eval_logps/chosen": -34.009239196777344,
|
|
|
|
| 13 |
"eval_samples": 343,
|
| 14 |
"eval_samples_per_second": 2.355,
|
| 15 |
"eval_steps_per_second": 0.295,
|
| 16 |
+
"train_loss": 0.17404492969636792,
|
| 17 |
+
"train_runtime": 10767.1792,
|
| 18 |
"train_samples": 3079,
|
| 19 |
+
"train_samples_per_second": 1.144,
|
| 20 |
+
"train_steps_per_second": 0.143
|
| 21 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 3079,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 4.0,
|
| 3 |
+
"train_loss": 0.17404492969636792,
|
| 4 |
+
"train_runtime": 10767.1792,
|
| 5 |
"train_samples": 3079,
|
| 6 |
+
"train_samples_per_second": 1.144,
|
| 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 +492,2070 @@
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
-
"
|
| 496 |
-
"
|
| 497 |
-
"logits/
|
| 498 |
-
"
|
| 499 |
-
"logps/
|
| 500 |
-
"
|
| 501 |
-
"
|
| 502 |
-
"rewards/
|
| 503 |
-
"rewards/
|
| 504 |
-
"rewards/
|
|
|
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.83,
|
| 509 |
-
"
|
| 510 |
-
"
|
| 511 |
-
"logits/
|
| 512 |
-
"
|
| 513 |
-
"logps/
|
| 514 |
-
"
|
| 515 |
-
"
|
| 516 |
-
"rewards/
|
| 517 |
-
"rewards/
|
| 518 |
-
"rewards/
|
|
|
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.86,
|
| 523 |
-
"
|
| 524 |
-
"
|
| 525 |
-
"logits/
|
| 526 |
-
"
|
| 527 |
-
"logps/
|
| 528 |
-
"
|
| 529 |
-
"
|
| 530 |
-
"rewards/
|
| 531 |
-
"rewards/
|
| 532 |
-
"rewards/
|
|
|
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.88,
|
| 537 |
-
"
|
| 538 |
-
"
|
| 539 |
-
"logits/
|
| 540 |
-
"
|
| 541 |
-
"logps/
|
| 542 |
-
"
|
| 543 |
-
"
|
| 544 |
-
"rewards/
|
| 545 |
-
"rewards/
|
| 546 |
-
"rewards/
|
|
|
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.91,
|
| 551 |
-
"
|
| 552 |
-
"
|
| 553 |
-
"logits/
|
| 554 |
-
"
|
| 555 |
-
"logps/
|
| 556 |
-
"
|
| 557 |
-
"
|
| 558 |
-
"rewards/
|
| 559 |
-
"rewards/
|
| 560 |
-
"rewards/
|
|
|
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.94,
|
| 565 |
-
"
|
| 566 |
-
"
|
| 567 |
-
"logits/
|
| 568 |
-
"
|
| 569 |
-
"logps/
|
| 570 |
-
"
|
| 571 |
-
"
|
| 572 |
-
"rewards/
|
| 573 |
-
"rewards/
|
| 574 |
-
"rewards/
|
|
|
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.96,
|
| 579 |
-
"
|
| 580 |
-
"
|
| 581 |
-
"logits/
|
| 582 |
-
"
|
| 583 |
-
"logps/
|
| 584 |
-
"
|
| 585 |
-
"
|
| 586 |
-
"rewards/
|
| 587 |
-
"rewards/
|
| 588 |
-
"rewards/
|
|
|
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.99,
|
| 593 |
-
"
|
| 594 |
-
"
|
| 595 |
-
"logits/
|
| 596 |
-
"
|
| 597 |
-
"logps/
|
| 598 |
-
"
|
| 599 |
-
"
|
| 600 |
-
"rewards/
|
| 601 |
-
"rewards/
|
| 602 |
-
"rewards/
|
|
|
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
-
"epoch": 1.
|
| 607 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 608 |
"total_flos": 0.0,
|
| 609 |
-
"train_loss": 0.
|
| 610 |
-
"train_runtime":
|
| 611 |
-
"train_samples_per_second":
|
| 612 |
-
"train_steps_per_second": 0.
|
| 613 |
}
|
| 614 |
],
|
| 615 |
"logging_steps": 10,
|
| 616 |
-
"max_steps":
|
| 617 |
"num_input_tokens_seen": 0,
|
| 618 |
-
"num_train_epochs":
|
| 619 |
"save_steps": 100,
|
| 620 |
"total_flos": 0.0,
|
| 621 |
"train_batch_size": 4,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 4.0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 1540,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.81,
|
| 495 |
+
"grad_norm": 107.0,
|
| 496 |
+
"learning_rate": 4.84533120650964e-06,
|
| 497 |
+
"logits/chosen": -2.0731887817382812,
|
| 498 |
+
"logits/rejected": -2.060344934463501,
|
| 499 |
+
"logps/chosen": -32.24032211303711,
|
| 500 |
+
"logps/rejected": -32.77452087402344,
|
| 501 |
+
"loss": 0.2321,
|
| 502 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 503 |
+
"rewards/chosen": 0.2592628598213196,
|
| 504 |
+
"rewards/margins": 0.3793802857398987,
|
| 505 |
+
"rewards/rejected": -0.12011740356683731,
|
| 506 |
"step": 310
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.83,
|
| 510 |
+
"grad_norm": 37.0,
|
| 511 |
+
"learning_rate": 4.825108134172131e-06,
|
| 512 |
+
"logits/chosen": -1.9875545501708984,
|
| 513 |
+
"logits/rejected": -1.978822112083435,
|
| 514 |
+
"logps/chosen": -32.03451919555664,
|
| 515 |
+
"logps/rejected": -30.3249454498291,
|
| 516 |
+
"loss": 0.2062,
|
| 517 |
+
"rewards/accuracies": 0.824999988079071,
|
| 518 |
+
"rewards/chosen": 0.288422554731369,
|
| 519 |
+
"rewards/margins": 0.3636184334754944,
|
| 520 |
+
"rewards/rejected": -0.07519589364528656,
|
| 521 |
"step": 320
|
| 522 |
},
|
| 523 |
{
|
| 524 |
"epoch": 0.86,
|
| 525 |
+
"grad_norm": 25.75,
|
| 526 |
+
"learning_rate": 4.80369052967602e-06,
|
| 527 |
+
"logits/chosen": -1.9262025356292725,
|
| 528 |
+
"logits/rejected": -1.9383175373077393,
|
| 529 |
+
"logps/chosen": -30.169696807861328,
|
| 530 |
+
"logps/rejected": -33.530517578125,
|
| 531 |
+
"loss": 0.214,
|
| 532 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 533 |
+
"rewards/chosen": 0.2786136865615845,
|
| 534 |
+
"rewards/margins": 0.3940487504005432,
|
| 535 |
+
"rewards/rejected": -0.11543502658605576,
|
| 536 |
"step": 330
|
| 537 |
},
|
| 538 |
{
|
| 539 |
"epoch": 0.88,
|
| 540 |
+
"grad_norm": 52.75,
|
| 541 |
+
"learning_rate": 4.781089396387968e-06,
|
| 542 |
+
"logits/chosen": -1.8931338787078857,
|
| 543 |
+
"logits/rejected": -1.8838812112808228,
|
| 544 |
+
"logps/chosen": -34.24372100830078,
|
| 545 |
+
"logps/rejected": -35.92654037475586,
|
| 546 |
+
"loss": 0.3165,
|
| 547 |
+
"rewards/accuracies": 0.887499988079071,
|
| 548 |
+
"rewards/chosen": 0.40968865156173706,
|
| 549 |
+
"rewards/margins": 0.47252458333969116,
|
| 550 |
+
"rewards/rejected": -0.06283595412969589,
|
| 551 |
"step": 340
|
| 552 |
},
|
| 553 |
{
|
| 554 |
"epoch": 0.91,
|
| 555 |
+
"grad_norm": 38.75,
|
| 556 |
+
"learning_rate": 4.757316345716554e-06,
|
| 557 |
+
"logits/chosen": -1.9516513347625732,
|
| 558 |
+
"logits/rejected": -1.9522939920425415,
|
| 559 |
+
"logps/chosen": -34.0687141418457,
|
| 560 |
+
"logps/rejected": -33.91803741455078,
|
| 561 |
+
"loss": 0.2699,
|
| 562 |
+
"rewards/accuracies": 0.800000011920929,
|
| 563 |
+
"rewards/chosen": 0.30291253328323364,
|
| 564 |
+
"rewards/margins": 0.3650719225406647,
|
| 565 |
+
"rewards/rejected": -0.06215935945510864,
|
| 566 |
"step": 350
|
| 567 |
},
|
| 568 |
{
|
| 569 |
"epoch": 0.94,
|
| 570 |
+
"grad_norm": 44.75,
|
| 571 |
+
"learning_rate": 4.73238359114687e-06,
|
| 572 |
+
"logits/chosen": -2.07902193069458,
|
| 573 |
+
"logits/rejected": -2.085242748260498,
|
| 574 |
+
"logps/chosen": -31.320343017578125,
|
| 575 |
+
"logps/rejected": -32.6987419128418,
|
| 576 |
+
"loss": 0.3034,
|
| 577 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 578 |
+
"rewards/chosen": 0.22761543095111847,
|
| 579 |
+
"rewards/margins": 0.2550276219844818,
|
| 580 |
+
"rewards/rejected": -0.027412226423621178,
|
| 581 |
"step": 360
|
| 582 |
},
|
| 583 |
{
|
| 584 |
"epoch": 0.96,
|
| 585 |
+
"grad_norm": 94.0,
|
| 586 |
+
"learning_rate": 4.706303941965804e-06,
|
| 587 |
+
"logits/chosen": -2.0127124786376953,
|
| 588 |
+
"logits/rejected": -2.0121982097625732,
|
| 589 |
+
"logps/chosen": -33.03043746948242,
|
| 590 |
+
"logps/rejected": -36.05421447753906,
|
| 591 |
+
"loss": 0.294,
|
| 592 |
+
"rewards/accuracies": 0.862500011920929,
|
| 593 |
+
"rewards/chosen": 0.3730550706386566,
|
| 594 |
+
"rewards/margins": 0.3770460784435272,
|
| 595 |
+
"rewards/rejected": -0.003990952856838703,
|
| 596 |
"step": 370
|
| 597 |
},
|
| 598 |
{
|
| 599 |
"epoch": 0.99,
|
| 600 |
+
"grad_norm": 27.25,
|
| 601 |
+
"learning_rate": 4.679090796681225e-06,
|
| 602 |
+
"logits/chosen": -2.0482752323150635,
|
| 603 |
+
"logits/rejected": -2.043614625930786,
|
| 604 |
+
"logps/chosen": -30.266733169555664,
|
| 605 |
+
"logps/rejected": -29.273147583007812,
|
| 606 |
+
"loss": 0.3277,
|
| 607 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 608 |
+
"rewards/chosen": 0.3994811177253723,
|
| 609 |
+
"rewards/margins": 0.45281118154525757,
|
| 610 |
+
"rewards/rejected": -0.05333010479807854,
|
| 611 |
"step": 380
|
| 612 |
},
|
| 613 |
{
|
| 614 |
+
"epoch": 1.01,
|
| 615 |
+
"grad_norm": 48.75,
|
| 616 |
+
"learning_rate": 4.650758136138454e-06,
|
| 617 |
+
"logits/chosen": -1.8216171264648438,
|
| 618 |
+
"logits/rejected": -1.8280115127563477,
|
| 619 |
+
"logps/chosen": -32.05220031738281,
|
| 620 |
+
"logps/rejected": -36.293907165527344,
|
| 621 |
+
"loss": 0.4405,
|
| 622 |
+
"rewards/accuracies": 0.8125,
|
| 623 |
+
"rewards/chosen": 0.30016615986824036,
|
| 624 |
+
"rewards/margins": 0.45822396874427795,
|
| 625 |
+
"rewards/rejected": -0.1580578237771988,
|
| 626 |
+
"step": 390
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 1.04,
|
| 630 |
+
"grad_norm": 60.75,
|
| 631 |
+
"learning_rate": 4.621320516337559e-06,
|
| 632 |
+
"logits/chosen": -1.9820178747177124,
|
| 633 |
+
"logits/rejected": -1.9757049083709717,
|
| 634 |
+
"logps/chosen": -33.380653381347656,
|
| 635 |
+
"logps/rejected": -32.34453582763672,
|
| 636 |
+
"loss": 0.4084,
|
| 637 |
+
"rewards/accuracies": 0.800000011920929,
|
| 638 |
+
"rewards/chosen": 0.4068370759487152,
|
| 639 |
+
"rewards/margins": 0.43893057107925415,
|
| 640 |
+
"rewards/rejected": -0.032093439251184464,
|
| 641 |
+
"step": 400
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"epoch": 1.04,
|
| 645 |
+
"eval_logits/chosen": -2.244844436645508,
|
| 646 |
+
"eval_logits/rejected": -2.2399790287017822,
|
| 647 |
+
"eval_logps/chosen": -33.95623016357422,
|
| 648 |
+
"eval_logps/rejected": -37.43685531616211,
|
| 649 |
+
"eval_loss": 0.40588921308517456,
|
| 650 |
+
"eval_rewards/accuracies": 0.5066444873809814,
|
| 651 |
+
"eval_rewards/chosen": 0.07048822194337845,
|
| 652 |
+
"eval_rewards/margins": -0.001299927243962884,
|
| 653 |
+
"eval_rewards/rejected": 0.07178815454244614,
|
| 654 |
+
"eval_runtime": 145.5279,
|
| 655 |
+
"eval_samples_per_second": 2.357,
|
| 656 |
+
"eval_steps_per_second": 0.295,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 1.06,
|
| 661 |
+
"grad_norm": 37.5,
|
| 662 |
+
"learning_rate": 4.590793060955158e-06,
|
| 663 |
+
"logits/chosen": -1.981856107711792,
|
| 664 |
+
"logits/rejected": -1.9890705347061157,
|
| 665 |
+
"logps/chosen": -28.692569732666016,
|
| 666 |
+
"logps/rejected": -29.251338958740234,
|
| 667 |
+
"loss": 0.3376,
|
| 668 |
+
"rewards/accuracies": 0.925000011920929,
|
| 669 |
+
"rewards/chosen": 0.3850264847278595,
|
| 670 |
+
"rewards/margins": 0.5228284597396851,
|
| 671 |
+
"rewards/rejected": -0.13780196011066437,
|
| 672 |
+
"step": 410
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 1.09,
|
| 676 |
+
"grad_norm": 46.5,
|
| 677 |
+
"learning_rate": 4.559191453574582e-06,
|
| 678 |
+
"logits/chosen": -2.0013296604156494,
|
| 679 |
+
"logits/rejected": -2.000218152999878,
|
| 680 |
+
"logps/chosen": -33.88956832885742,
|
| 681 |
+
"logps/rejected": -30.98602294921875,
|
| 682 |
+
"loss": 0.244,
|
| 683 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 684 |
+
"rewards/chosen": 0.32502850890159607,
|
| 685 |
+
"rewards/margins": 0.36370736360549927,
|
| 686 |
+
"rewards/rejected": -0.038678836077451706,
|
| 687 |
+
"step": 420
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 1.12,
|
| 691 |
+
"grad_norm": 74.5,
|
| 692 |
+
"learning_rate": 4.52653192962838e-06,
|
| 693 |
+
"logits/chosen": -1.9944326877593994,
|
| 694 |
+
"logits/rejected": -1.9771124124526978,
|
| 695 |
+
"logps/chosen": -30.677692413330078,
|
| 696 |
+
"logps/rejected": -33.02484893798828,
|
| 697 |
+
"loss": 0.2988,
|
| 698 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 699 |
+
"rewards/chosen": 0.269562691450119,
|
| 700 |
+
"rewards/margins": 0.35826626420021057,
|
| 701 |
+
"rewards/rejected": -0.08870354294776917,
|
| 702 |
+
"step": 430
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 1.14,
|
| 706 |
+
"grad_norm": 38.5,
|
| 707 |
+
"learning_rate": 4.492831268057307e-06,
|
| 708 |
+
"logits/chosen": -2.0252749919891357,
|
| 709 |
+
"logits/rejected": -2.0271475315093994,
|
| 710 |
+
"logps/chosen": -36.258460998535156,
|
| 711 |
+
"logps/rejected": -34.926544189453125,
|
| 712 |
+
"loss": 0.292,
|
| 713 |
+
"rewards/accuracies": 0.800000011920929,
|
| 714 |
+
"rewards/chosen": 0.31851616501808167,
|
| 715 |
+
"rewards/margins": 0.3784351646900177,
|
| 716 |
+
"rewards/rejected": -0.059918999671936035,
|
| 717 |
+
"step": 440
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.17,
|
| 721 |
+
"grad_norm": 37.5,
|
| 722 |
+
"learning_rate": 4.458106782690094e-06,
|
| 723 |
+
"logits/chosen": -2.1034398078918457,
|
| 724 |
+
"logits/rejected": -2.1033284664154053,
|
| 725 |
+
"logps/chosen": -32.266780853271484,
|
| 726 |
+
"logps/rejected": -33.54703140258789,
|
| 727 |
+
"loss": 0.5497,
|
| 728 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 729 |
+
"rewards/chosen": 0.3299091160297394,
|
| 730 |
+
"rewards/margins": 0.31813210248947144,
|
| 731 |
+
"rewards/rejected": 0.011777004227042198,
|
| 732 |
+
"step": 450
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"epoch": 1.19,
|
| 736 |
+
"grad_norm": 33.75,
|
| 737 |
+
"learning_rate": 4.422376313348405e-06,
|
| 738 |
+
"logits/chosen": -2.0468077659606934,
|
| 739 |
+
"logits/rejected": -2.0393521785736084,
|
| 740 |
+
"logps/chosen": -32.022674560546875,
|
| 741 |
+
"logps/rejected": -36.174320220947266,
|
| 742 |
+
"loss": 0.4905,
|
| 743 |
+
"rewards/accuracies": 0.800000011920929,
|
| 744 |
+
"rewards/chosen": 0.34495481848716736,
|
| 745 |
+
"rewards/margins": 0.4532076418399811,
|
| 746 |
+
"rewards/rejected": -0.10825282335281372,
|
| 747 |
+
"step": 460
|
| 748 |
+
},
|
| 749 |
+
{
|
| 750 |
+
"epoch": 1.22,
|
| 751 |
+
"grad_norm": 59.25,
|
| 752 |
+
"learning_rate": 4.3856582166815696e-06,
|
| 753 |
+
"logits/chosen": -1.9485485553741455,
|
| 754 |
+
"logits/rejected": -1.944986343383789,
|
| 755 |
+
"logps/chosen": -33.546749114990234,
|
| 756 |
+
"logps/rejected": -33.010704040527344,
|
| 757 |
+
"loss": 0.283,
|
| 758 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 759 |
+
"rewards/chosen": 0.34699487686157227,
|
| 760 |
+
"rewards/margins": 0.4251338541507721,
|
| 761 |
+
"rewards/rejected": -0.07813899219036102,
|
| 762 |
+
"step": 470
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 1.25,
|
| 766 |
+
"grad_norm": 61.25,
|
| 767 |
+
"learning_rate": 4.347971356735789e-06,
|
| 768 |
+
"logits/chosen": -2.071899652481079,
|
| 769 |
+
"logits/rejected": -2.064889907836914,
|
| 770 |
+
"logps/chosen": -30.4763240814209,
|
| 771 |
+
"logps/rejected": -32.066490173339844,
|
| 772 |
+
"loss": 0.2764,
|
| 773 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 774 |
+
"rewards/chosen": 0.316992849111557,
|
| 775 |
+
"rewards/margins": 0.3826924264431,
|
| 776 |
+
"rewards/rejected": -0.06569959223270416,
|
| 777 |
+
"step": 480
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 1.27,
|
| 781 |
+
"grad_norm": 69.0,
|
| 782 |
+
"learning_rate": 4.309335095262675e-06,
|
| 783 |
+
"logits/chosen": -2.01814341545105,
|
| 784 |
+
"logits/rejected": -2.01973557472229,
|
| 785 |
+
"logps/chosen": -34.93284225463867,
|
| 786 |
+
"logps/rejected": -33.683841705322266,
|
| 787 |
+
"loss": 0.2944,
|
| 788 |
+
"rewards/accuracies": 0.8125,
|
| 789 |
+
"rewards/chosen": 0.37830987572669983,
|
| 790 |
+
"rewards/margins": 0.4164048731327057,
|
| 791 |
+
"rewards/rejected": -0.038094960153102875,
|
| 792 |
+
"step": 490
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 1.3,
|
| 796 |
+
"grad_norm": 44.0,
|
| 797 |
+
"learning_rate": 4.269769281772082e-06,
|
| 798 |
+
"logits/chosen": -1.9071893692016602,
|
| 799 |
+
"logits/rejected": -1.9048168659210205,
|
| 800 |
+
"logps/chosen": -32.796241760253906,
|
| 801 |
+
"logps/rejected": -36.574031829833984,
|
| 802 |
+
"loss": 0.2788,
|
| 803 |
+
"rewards/accuracies": 0.875,
|
| 804 |
+
"rewards/chosen": 0.33844107389450073,
|
| 805 |
+
"rewards/margins": 0.4371793270111084,
|
| 806 |
+
"rewards/rejected": -0.09873829782009125,
|
| 807 |
+
"step": 500
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 1.3,
|
| 811 |
+
"eval_logits/chosen": -2.2405266761779785,
|
| 812 |
+
"eval_logits/rejected": -2.2356326580047607,
|
| 813 |
+
"eval_logps/chosen": -33.95664978027344,
|
| 814 |
+
"eval_logps/rejected": -37.45263671875,
|
| 815 |
+
"eval_loss": 0.38657647371292114,
|
| 816 |
+
"eval_rewards/accuracies": 0.5191029906272888,
|
| 817 |
+
"eval_rewards/chosen": 0.07010918855667114,
|
| 818 |
+
"eval_rewards/margins": 0.012523897923529148,
|
| 819 |
+
"eval_rewards/rejected": 0.05758529528975487,
|
| 820 |
+
"eval_runtime": 145.2533,
|
| 821 |
+
"eval_samples_per_second": 2.361,
|
| 822 |
+
"eval_steps_per_second": 0.296,
|
| 823 |
+
"step": 500
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 1.32,
|
| 827 |
+
"grad_norm": 37.5,
|
| 828 |
+
"learning_rate": 4.22929424333435e-06,
|
| 829 |
+
"logits/chosen": -2.013305187225342,
|
| 830 |
+
"logits/rejected": -2.018036127090454,
|
| 831 |
+
"logps/chosen": -33.06889724731445,
|
| 832 |
+
"logps/rejected": -31.70903968811035,
|
| 833 |
+
"loss": 0.3894,
|
| 834 |
+
"rewards/accuracies": 0.800000011920929,
|
| 835 |
+
"rewards/chosen": 0.34961196780204773,
|
| 836 |
+
"rewards/margins": 0.40624913573265076,
|
| 837 |
+
"rewards/rejected": -0.05663713067770004,
|
| 838 |
+
"step": 510
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 1.35,
|
| 842 |
+
"grad_norm": 43.5,
|
| 843 |
+
"learning_rate": 4.1879307741372085e-06,
|
| 844 |
+
"logits/chosen": -2.0473179817199707,
|
| 845 |
+
"logits/rejected": -2.0582363605499268,
|
| 846 |
+
"logps/chosen": -30.8958797454834,
|
| 847 |
+
"logps/rejected": -31.604564666748047,
|
| 848 |
+
"loss": 0.359,
|
| 849 |
+
"rewards/accuracies": 0.875,
|
| 850 |
+
"rewards/chosen": 0.44507941603660583,
|
| 851 |
+
"rewards/margins": 0.5322669744491577,
|
| 852 |
+
"rewards/rejected": -0.08718754351139069,
|
| 853 |
+
"step": 520
|
| 854 |
+
},
|
| 855 |
+
{
|
| 856 |
+
"epoch": 1.38,
|
| 857 |
+
"grad_norm": 49.5,
|
| 858 |
+
"learning_rate": 4.145700124802693e-06,
|
| 859 |
+
"logits/chosen": -1.974761962890625,
|
| 860 |
+
"logits/rejected": -1.9713010787963867,
|
| 861 |
+
"logps/chosen": -32.03337097167969,
|
| 862 |
+
"logps/rejected": -32.46263885498047,
|
| 863 |
+
"loss": 0.2845,
|
| 864 |
+
"rewards/accuracies": 0.875,
|
| 865 |
+
"rewards/chosen": 0.34842735528945923,
|
| 866 |
+
"rewards/margins": 0.4383048117160797,
|
| 867 |
+
"rewards/rejected": -0.08987746387720108,
|
| 868 |
+
"step": 530
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"epoch": 1.4,
|
| 872 |
+
"grad_norm": 56.0,
|
| 873 |
+
"learning_rate": 4.102623991469562e-06,
|
| 874 |
+
"logits/chosen": -1.8440120220184326,
|
| 875 |
+
"logits/rejected": -1.8531720638275146,
|
| 876 |
+
"logps/chosen": -32.23934555053711,
|
| 877 |
+
"logps/rejected": -31.933460235595703,
|
| 878 |
+
"loss": 0.2635,
|
| 879 |
+
"rewards/accuracies": 0.824999988079071,
|
| 880 |
+
"rewards/chosen": 0.29704010486602783,
|
| 881 |
+
"rewards/margins": 0.36676424741744995,
|
| 882 |
+
"rewards/rejected": -0.06972410529851913,
|
| 883 |
+
"step": 540
|
| 884 |
+
},
|
| 885 |
+
{
|
| 886 |
+
"epoch": 1.43,
|
| 887 |
+
"grad_norm": 41.75,
|
| 888 |
+
"learning_rate": 4.058724504646834e-06,
|
| 889 |
+
"logits/chosen": -1.9423996210098267,
|
| 890 |
+
"logits/rejected": -1.9359699487686157,
|
| 891 |
+
"logps/chosen": -33.36035919189453,
|
| 892 |
+
"logps/rejected": -31.000152587890625,
|
| 893 |
+
"loss": 0.3772,
|
| 894 |
+
"rewards/accuracies": 0.8125,
|
| 895 |
+
"rewards/chosen": 0.3131468892097473,
|
| 896 |
+
"rewards/margins": 0.32107049226760864,
|
| 897 |
+
"rewards/rejected": -0.007923613302409649,
|
| 898 |
+
"step": 550
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 1.45,
|
| 902 |
+
"grad_norm": 147.0,
|
| 903 |
+
"learning_rate": 4.014024217844167e-06,
|
| 904 |
+
"logits/chosen": -2.0321927070617676,
|
| 905 |
+
"logits/rejected": -2.030066967010498,
|
| 906 |
+
"logps/chosen": -34.12587356567383,
|
| 907 |
+
"logps/rejected": -31.518299102783203,
|
| 908 |
+
"loss": 0.5612,
|
| 909 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 910 |
+
"rewards/chosen": 0.40079349279403687,
|
| 911 |
+
"rewards/margins": 0.41961637139320374,
|
| 912 |
+
"rewards/rejected": -0.01882282830774784,
|
| 913 |
+
"step": 560
|
| 914 |
+
},
|
| 915 |
+
{
|
| 916 |
+
"epoch": 1.48,
|
| 917 |
+
"grad_norm": 28.125,
|
| 918 |
+
"learning_rate": 3.968546095984911e-06,
|
| 919 |
+
"logits/chosen": -1.8660882711410522,
|
| 920 |
+
"logits/rejected": -1.863891363143921,
|
| 921 |
+
"logps/chosen": -32.46380615234375,
|
| 922 |
+
"logps/rejected": -31.035659790039062,
|
| 923 |
+
"loss": 0.3042,
|
| 924 |
+
"rewards/accuracies": 0.824999988079071,
|
| 925 |
+
"rewards/chosen": 0.2993435263633728,
|
| 926 |
+
"rewards/margins": 0.3293627202510834,
|
| 927 |
+
"rewards/rejected": -0.030019233003258705,
|
| 928 |
+
"step": 570
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"epoch": 1.51,
|
| 932 |
+
"grad_norm": 126.0,
|
| 933 |
+
"learning_rate": 3.922313503607806e-06,
|
| 934 |
+
"logits/chosen": -2.0015358924865723,
|
| 935 |
+
"logits/rejected": -1.998284935951233,
|
| 936 |
+
"logps/chosen": -30.789913177490234,
|
| 937 |
+
"logps/rejected": -34.707244873046875,
|
| 938 |
+
"loss": 0.3323,
|
| 939 |
+
"rewards/accuracies": 0.824999988079071,
|
| 940 |
+
"rewards/chosen": 0.2832598090171814,
|
| 941 |
+
"rewards/margins": 0.3562392592430115,
|
| 942 |
+
"rewards/rejected": -0.07297946512699127,
|
| 943 |
+
"step": 580
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.53,
|
| 947 |
+
"grad_norm": 26.125,
|
| 948 |
+
"learning_rate": 3.875350192863368e-06,
|
| 949 |
+
"logits/chosen": -1.946082353591919,
|
| 950 |
+
"logits/rejected": -1.9497697353363037,
|
| 951 |
+
"logps/chosen": -29.2518310546875,
|
| 952 |
+
"logps/rejected": -30.736743927001953,
|
| 953 |
+
"loss": 0.3063,
|
| 954 |
+
"rewards/accuracies": 0.762499988079071,
|
| 955 |
+
"rewards/chosen": 0.3268837332725525,
|
| 956 |
+
"rewards/margins": 0.3969722092151642,
|
| 957 |
+
"rewards/rejected": -0.07008843123912811,
|
| 958 |
+
"step": 590
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 1.56,
|
| 962 |
+
"grad_norm": 46.25,
|
| 963 |
+
"learning_rate": 3.8276802913111436e-06,
|
| 964 |
+
"logits/chosen": -1.9876987934112549,
|
| 965 |
+
"logits/rejected": -1.9874299764633179,
|
| 966 |
+
"logps/chosen": -31.799388885498047,
|
| 967 |
+
"logps/rejected": -31.247623443603516,
|
| 968 |
+
"loss": 0.3874,
|
| 969 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 970 |
+
"rewards/chosen": 0.3799284100532532,
|
| 971 |
+
"rewards/margins": 0.4290773868560791,
|
| 972 |
+
"rewards/rejected": -0.049148980528116226,
|
| 973 |
+
"step": 600
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.56,
|
| 977 |
+
"eval_logits/chosen": -2.2469940185546875,
|
| 978 |
+
"eval_logits/rejected": -2.2421164512634277,
|
| 979 |
+
"eval_logps/chosen": -33.955604553222656,
|
| 980 |
+
"eval_logps/rejected": -37.417686462402344,
|
| 981 |
+
"eval_loss": 0.426466703414917,
|
| 982 |
+
"eval_rewards/accuracies": 0.472591370344162,
|
| 983 |
+
"eval_rewards/chosen": 0.07105044275522232,
|
| 984 |
+
"eval_rewards/margins": -0.017991825938224792,
|
| 985 |
+
"eval_rewards/rejected": 0.08904226124286652,
|
| 986 |
+
"eval_runtime": 145.3878,
|
| 987 |
+
"eval_samples_per_second": 2.359,
|
| 988 |
+
"eval_steps_per_second": 0.296,
|
| 989 |
+
"step": 600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 1.58,
|
| 993 |
+
"grad_norm": 23.5,
|
| 994 |
+
"learning_rate": 3.7793282895240927e-06,
|
| 995 |
+
"logits/chosen": -2.0506019592285156,
|
| 996 |
+
"logits/rejected": -2.051326274871826,
|
| 997 |
+
"logps/chosen": -34.585548400878906,
|
| 998 |
+
"logps/rejected": -33.11695098876953,
|
| 999 |
+
"loss": 0.2877,
|
| 1000 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1001 |
+
"rewards/chosen": 0.4001394212245941,
|
| 1002 |
+
"rewards/margins": 0.4296254515647888,
|
| 1003 |
+
"rewards/rejected": -0.02948596142232418,
|
| 1004 |
+
"step": 610
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 1.61,
|
| 1008 |
+
"grad_norm": 69.5,
|
| 1009 |
+
"learning_rate": 3.730319028506478e-06,
|
| 1010 |
+
"logits/chosen": -2.013078212738037,
|
| 1011 |
+
"logits/rejected": -2.010453701019287,
|
| 1012 |
+
"logps/chosen": -32.696006774902344,
|
| 1013 |
+
"logps/rejected": -32.14519500732422,
|
| 1014 |
+
"loss": 0.3336,
|
| 1015 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1016 |
+
"rewards/chosen": 0.3864508867263794,
|
| 1017 |
+
"rewards/margins": 0.4542017877101898,
|
| 1018 |
+
"rewards/rejected": -0.06775090843439102,
|
| 1019 |
+
"step": 620
|
| 1020 |
+
},
|
| 1021 |
+
{
|
| 1022 |
+
"epoch": 1.64,
|
| 1023 |
+
"grad_norm": 31.875,
|
| 1024 |
+
"learning_rate": 3.6806776869317074e-06,
|
| 1025 |
+
"logits/chosen": -2.02937650680542,
|
| 1026 |
+
"logits/rejected": -2.0202343463897705,
|
| 1027 |
+
"logps/chosen": -32.354148864746094,
|
| 1028 |
+
"logps/rejected": -30.961654663085938,
|
| 1029 |
+
"loss": 0.3879,
|
| 1030 |
+
"rewards/accuracies": 0.8125,
|
| 1031 |
+
"rewards/chosen": 0.3797711730003357,
|
| 1032 |
+
"rewards/margins": 0.40343213081359863,
|
| 1033 |
+
"rewards/rejected": -0.023660968989133835,
|
| 1034 |
+
"step": 630
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 1.66,
|
| 1038 |
+
"grad_norm": 119.0,
|
| 1039 |
+
"learning_rate": 3.6304297682067146e-06,
|
| 1040 |
+
"logits/chosen": -2.035832405090332,
|
| 1041 |
+
"logits/rejected": -2.032696485519409,
|
| 1042 |
+
"logps/chosen": -31.714279174804688,
|
| 1043 |
+
"logps/rejected": -32.49519348144531,
|
| 1044 |
+
"loss": 0.3205,
|
| 1045 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1046 |
+
"rewards/chosen": 0.43572768568992615,
|
| 1047 |
+
"rewards/margins": 0.5039243102073669,
|
| 1048 |
+
"rewards/rejected": -0.06819657981395721,
|
| 1049 |
+
"step": 640
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 1.69,
|
| 1053 |
+
"grad_norm": 24.375,
|
| 1054 |
+
"learning_rate": 3.579601087369492e-06,
|
| 1055 |
+
"logits/chosen": -2.0391454696655273,
|
| 1056 |
+
"logits/rejected": -2.041524887084961,
|
| 1057 |
+
"logps/chosen": -33.29151916503906,
|
| 1058 |
+
"logps/rejected": -33.737464904785156,
|
| 1059 |
+
"loss": 0.2569,
|
| 1060 |
+
"rewards/accuracies": 0.875,
|
| 1061 |
+
"rewards/chosen": 0.32596832513809204,
|
| 1062 |
+
"rewards/margins": 0.37462925910949707,
|
| 1063 |
+
"rewards/rejected": -0.04866095632314682,
|
| 1064 |
+
"step": 650
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 1.71,
|
| 1068 |
+
"grad_norm": 32.5,
|
| 1069 |
+
"learning_rate": 3.5282177578265295e-06,
|
| 1070 |
+
"logits/chosen": -1.9535996913909912,
|
| 1071 |
+
"logits/rejected": -1.9540789127349854,
|
| 1072 |
+
"logps/chosen": -33.57703399658203,
|
| 1073 |
+
"logps/rejected": -31.611495971679688,
|
| 1074 |
+
"loss": 0.2509,
|
| 1075 |
+
"rewards/accuracies": 0.75,
|
| 1076 |
+
"rewards/chosen": 0.33281606435775757,
|
| 1077 |
+
"rewards/margins": 0.3366300165653229,
|
| 1078 |
+
"rewards/rejected": -0.0038140162359923124,
|
| 1079 |
+
"step": 660
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 1.74,
|
| 1083 |
+
"grad_norm": 46.75,
|
| 1084 |
+
"learning_rate": 3.476306177936961e-06,
|
| 1085 |
+
"logits/chosen": -1.9932886362075806,
|
| 1086 |
+
"logits/rejected": -1.983457326889038,
|
| 1087 |
+
"logps/chosen": -33.600196838378906,
|
| 1088 |
+
"logps/rejected": -32.383522033691406,
|
| 1089 |
+
"loss": 0.2964,
|
| 1090 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1091 |
+
"rewards/chosen": 0.36097538471221924,
|
| 1092 |
+
"rewards/margins": 0.4291335642337799,
|
| 1093 |
+
"rewards/rejected": -0.06815823167562485,
|
| 1094 |
+
"step": 670
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 1.77,
|
| 1098 |
+
"grad_norm": 53.25,
|
| 1099 |
+
"learning_rate": 3.423893017450324e-06,
|
| 1100 |
+
"logits/chosen": -1.8898265361785889,
|
| 1101 |
+
"logits/rejected": -1.8866018056869507,
|
| 1102 |
+
"logps/chosen": -31.05093002319336,
|
| 1103 |
+
"logps/rejected": -34.363487243652344,
|
| 1104 |
+
"loss": 0.2359,
|
| 1105 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1106 |
+
"rewards/chosen": 0.36902502179145813,
|
| 1107 |
+
"rewards/margins": 0.4094383120536804,
|
| 1108 |
+
"rewards/rejected": -0.0404132641851902,
|
| 1109 |
+
"step": 680
|
| 1110 |
+
},
|
| 1111 |
+
{
|
| 1112 |
+
"epoch": 1.79,
|
| 1113 |
+
"grad_norm": 99.5,
|
| 1114 |
+
"learning_rate": 3.3710052038048794e-06,
|
| 1115 |
+
"logits/chosen": -1.9469459056854248,
|
| 1116 |
+
"logits/rejected": -1.9461286067962646,
|
| 1117 |
+
"logps/chosen": -34.85956954956055,
|
| 1118 |
+
"logps/rejected": -35.397823333740234,
|
| 1119 |
+
"loss": 0.313,
|
| 1120 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1121 |
+
"rewards/chosen": 0.36964648962020874,
|
| 1122 |
+
"rewards/margins": 0.4092490077018738,
|
| 1123 |
+
"rewards/rejected": -0.03960253670811653,
|
| 1124 |
+
"step": 690
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 1.82,
|
| 1128 |
+
"grad_norm": 60.0,
|
| 1129 |
+
"learning_rate": 3.3176699082935546e-06,
|
| 1130 |
+
"logits/chosen": -1.9282480478286743,
|
| 1131 |
+
"logits/rejected": -1.9313417673110962,
|
| 1132 |
+
"logps/chosen": -32.19511032104492,
|
| 1133 |
+
"logps/rejected": -35.74637985229492,
|
| 1134 |
+
"loss": 0.2695,
|
| 1135 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1136 |
+
"rewards/chosen": 0.3669659197330475,
|
| 1137 |
+
"rewards/margins": 0.46120119094848633,
|
| 1138 |
+
"rewards/rejected": -0.09423529356718063,
|
| 1139 |
+
"step": 700
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 1.82,
|
| 1143 |
+
"eval_logits/chosen": -2.247775077819824,
|
| 1144 |
+
"eval_logits/rejected": -2.2429134845733643,
|
| 1145 |
+
"eval_logps/chosen": -33.943904876708984,
|
| 1146 |
+
"eval_logps/rejected": -37.41926193237305,
|
| 1147 |
+
"eval_loss": 0.402832955121994,
|
| 1148 |
+
"eval_rewards/accuracies": 0.5078904032707214,
|
| 1149 |
+
"eval_rewards/chosen": 0.08158490061759949,
|
| 1150 |
+
"eval_rewards/margins": -0.006037927698343992,
|
| 1151 |
+
"eval_rewards/rejected": 0.08762282878160477,
|
| 1152 |
+
"eval_runtime": 145.2777,
|
| 1153 |
+
"eval_samples_per_second": 2.361,
|
| 1154 |
+
"eval_steps_per_second": 0.296,
|
| 1155 |
+
"step": 700
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 1.84,
|
| 1159 |
+
"grad_norm": 51.75,
|
| 1160 |
+
"learning_rate": 3.2639145321045933e-06,
|
| 1161 |
+
"logits/chosen": -2.040821075439453,
|
| 1162 |
+
"logits/rejected": -2.0436055660247803,
|
| 1163 |
+
"logps/chosen": -34.49801254272461,
|
| 1164 |
+
"logps/rejected": -34.32398223876953,
|
| 1165 |
+
"loss": 0.3608,
|
| 1166 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1167 |
+
"rewards/chosen": 0.35781916975975037,
|
| 1168 |
+
"rewards/margins": 0.40626224875450134,
|
| 1169 |
+
"rewards/rejected": -0.04844306781888008,
|
| 1170 |
+
"step": 710
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 1.87,
|
| 1174 |
+
"grad_norm": 110.5,
|
| 1175 |
+
"learning_rate": 3.2097666922441107e-06,
|
| 1176 |
+
"logits/chosen": -1.8965297937393188,
|
| 1177 |
+
"logits/rejected": -1.8905613422393799,
|
| 1178 |
+
"logps/chosen": -34.4133415222168,
|
| 1179 |
+
"logps/rejected": -32.38819885253906,
|
| 1180 |
+
"loss": 0.2806,
|
| 1181 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1182 |
+
"rewards/chosen": 0.3779004216194153,
|
| 1183 |
+
"rewards/margins": 0.4106220602989197,
|
| 1184 |
+
"rewards/rejected": -0.03272164613008499,
|
| 1185 |
+
"step": 720
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.9,
|
| 1189 |
+
"grad_norm": 39.5,
|
| 1190 |
+
"learning_rate": 3.1552542073477554e-06,
|
| 1191 |
+
"logits/chosen": -2.0699164867401123,
|
| 1192 |
+
"logits/rejected": -2.0668702125549316,
|
| 1193 |
+
"logps/chosen": -30.270557403564453,
|
| 1194 |
+
"logps/rejected": -31.52914047241211,
|
| 1195 |
+
"loss": 0.2811,
|
| 1196 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1197 |
+
"rewards/chosen": 0.3441884219646454,
|
| 1198 |
+
"rewards/margins": 0.41703981161117554,
|
| 1199 |
+
"rewards/rejected": -0.07285144180059433,
|
| 1200 |
+
"step": 730
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 1.92,
|
| 1204 |
+
"grad_norm": 63.0,
|
| 1205 |
+
"learning_rate": 3.100405083388799e-06,
|
| 1206 |
+
"logits/chosen": -1.9103057384490967,
|
| 1207 |
+
"logits/rejected": -1.910112738609314,
|
| 1208 |
+
"logps/chosen": -33.07831573486328,
|
| 1209 |
+
"logps/rejected": -37.679222106933594,
|
| 1210 |
+
"loss": 0.2643,
|
| 1211 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1212 |
+
"rewards/chosen": 0.3463733494281769,
|
| 1213 |
+
"rewards/margins": 0.39264675974845886,
|
| 1214 |
+
"rewards/rejected": -0.04627341777086258,
|
| 1215 |
+
"step": 740
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"epoch": 1.95,
|
| 1219 |
+
"grad_norm": 60.75,
|
| 1220 |
+
"learning_rate": 3.0452474992899645e-06,
|
| 1221 |
+
"logits/chosen": -1.7918742895126343,
|
| 1222 |
+
"logits/rejected": -1.7970874309539795,
|
| 1223 |
+
"logps/chosen": -36.49778366088867,
|
| 1224 |
+
"logps/rejected": -34.00720977783203,
|
| 1225 |
+
"loss": 0.2697,
|
| 1226 |
+
"rewards/accuracies": 0.875,
|
| 1227 |
+
"rewards/chosen": 0.3912108838558197,
|
| 1228 |
+
"rewards/margins": 0.4225967526435852,
|
| 1229 |
+
"rewards/rejected": -0.031385865062475204,
|
| 1230 |
+
"step": 750
|
| 1231 |
+
},
|
| 1232 |
+
{
|
| 1233 |
+
"epoch": 1.97,
|
| 1234 |
+
"grad_norm": 51.25,
|
| 1235 |
+
"learning_rate": 2.989809792446417e-06,
|
| 1236 |
+
"logits/chosen": -2.0016860961914062,
|
| 1237 |
+
"logits/rejected": -2.003040313720703,
|
| 1238 |
+
"logps/chosen": -32.25410461425781,
|
| 1239 |
+
"logps/rejected": -32.861629486083984,
|
| 1240 |
+
"loss": 0.337,
|
| 1241 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1242 |
+
"rewards/chosen": 0.31641632318496704,
|
| 1243 |
+
"rewards/margins": 0.3905896544456482,
|
| 1244 |
+
"rewards/rejected": -0.07417333871126175,
|
| 1245 |
+
"step": 760
|
| 1246 |
+
},
|
| 1247 |
+
{
|
| 1248 |
+
"epoch": 2.0,
|
| 1249 |
+
"grad_norm": 67.0,
|
| 1250 |
+
"learning_rate": 2.9341204441673267e-06,
|
| 1251 |
+
"logits/chosen": -1.9714769124984741,
|
| 1252 |
+
"logits/rejected": -1.9706627130508423,
|
| 1253 |
+
"logps/chosen": -31.806903839111328,
|
| 1254 |
+
"logps/rejected": -34.992042541503906,
|
| 1255 |
+
"loss": 0.3126,
|
| 1256 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1257 |
+
"rewards/chosen": 0.3318381905555725,
|
| 1258 |
+
"rewards/margins": 0.29933661222457886,
|
| 1259 |
+
"rewards/rejected": 0.03250158578157425,
|
| 1260 |
+
"step": 770
|
| 1261 |
+
},
|
| 1262 |
+
{
|
| 1263 |
+
"epoch": 2.03,
|
| 1264 |
+
"grad_norm": 43.25,
|
| 1265 |
+
"learning_rate": 2.878208065043501e-06,
|
| 1266 |
+
"logits/chosen": -1.971329689025879,
|
| 1267 |
+
"logits/rejected": -1.9705543518066406,
|
| 1268 |
+
"logps/chosen": -34.23518371582031,
|
| 1269 |
+
"logps/rejected": -31.475284576416016,
|
| 1270 |
+
"loss": 0.2613,
|
| 1271 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1272 |
+
"rewards/chosen": 0.371949702501297,
|
| 1273 |
+
"rewards/margins": 0.42213401198387146,
|
| 1274 |
+
"rewards/rejected": -0.05018440634012222,
|
| 1275 |
+
"step": 780
|
| 1276 |
+
},
|
| 1277 |
+
{
|
| 1278 |
+
"epoch": 2.05,
|
| 1279 |
+
"grad_norm": 41.0,
|
| 1280 |
+
"learning_rate": 2.8221013802485974e-06,
|
| 1281 |
+
"logits/chosen": -2.003086566925049,
|
| 1282 |
+
"logits/rejected": -2.0015482902526855,
|
| 1283 |
+
"logps/chosen": -29.147836685180664,
|
| 1284 |
+
"logps/rejected": -32.72181701660156,
|
| 1285 |
+
"loss": 0.1641,
|
| 1286 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1287 |
+
"rewards/chosen": 0.37381279468536377,
|
| 1288 |
+
"rewards/margins": 0.450767457485199,
|
| 1289 |
+
"rewards/rejected": -0.07695472240447998,
|
| 1290 |
+
"step": 790
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"epoch": 2.08,
|
| 1294 |
+
"grad_norm": 28.375,
|
| 1295 |
+
"learning_rate": 2.76582921478147e-06,
|
| 1296 |
+
"logits/chosen": -2.0545146465301514,
|
| 1297 |
+
"logits/rejected": -2.0508596897125244,
|
| 1298 |
+
"logps/chosen": -31.972219467163086,
|
| 1299 |
+
"logps/rejected": -34.25817108154297,
|
| 1300 |
+
"loss": 0.1725,
|
| 1301 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1302 |
+
"rewards/chosen": 0.4094924330711365,
|
| 1303 |
+
"rewards/margins": 0.46271318197250366,
|
| 1304 |
+
"rewards/rejected": -0.05322074890136719,
|
| 1305 |
+
"step": 800
|
| 1306 |
+
},
|
| 1307 |
+
{
|
| 1308 |
+
"epoch": 2.08,
|
| 1309 |
+
"eval_logits/chosen": -2.246337890625,
|
| 1310 |
+
"eval_logits/rejected": -2.2414615154266357,
|
| 1311 |
+
"eval_logps/chosen": -33.92709732055664,
|
| 1312 |
+
"eval_logps/rejected": -37.39698791503906,
|
| 1313 |
+
"eval_loss": 0.4082644581794739,
|
| 1314 |
+
"eval_rewards/accuracies": 0.4821428656578064,
|
| 1315 |
+
"eval_rewards/chosen": 0.09670902043581009,
|
| 1316 |
+
"eval_rewards/margins": -0.010959767736494541,
|
| 1317 |
+
"eval_rewards/rejected": 0.10766878724098206,
|
| 1318 |
+
"eval_runtime": 145.3536,
|
| 1319 |
+
"eval_samples_per_second": 2.36,
|
| 1320 |
+
"eval_steps_per_second": 0.296,
|
| 1321 |
+
"step": 800
|
| 1322 |
+
},
|
| 1323 |
+
{
|
| 1324 |
+
"epoch": 2.1,
|
| 1325 |
+
"grad_norm": 25.625,
|
| 1326 |
+
"learning_rate": 2.7094204786572254e-06,
|
| 1327 |
+
"logits/chosen": -1.8916804790496826,
|
| 1328 |
+
"logits/rejected": -1.8837049007415771,
|
| 1329 |
+
"logps/chosen": -34.002410888671875,
|
| 1330 |
+
"logps/rejected": -34.613807678222656,
|
| 1331 |
+
"loss": 0.2292,
|
| 1332 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1333 |
+
"rewards/chosen": 0.4084460735321045,
|
| 1334 |
+
"rewards/margins": 0.4567646086215973,
|
| 1335 |
+
"rewards/rejected": -0.04831857234239578,
|
| 1336 |
+
"step": 810
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 2.13,
|
| 1340 |
+
"grad_norm": 42.0,
|
| 1341 |
+
"learning_rate": 2.6529041520546072e-06,
|
| 1342 |
+
"logits/chosen": -1.967140555381775,
|
| 1343 |
+
"logits/rejected": -1.9777908325195312,
|
| 1344 |
+
"logps/chosen": -35.31997299194336,
|
| 1345 |
+
"logps/rejected": -32.446533203125,
|
| 1346 |
+
"loss": 0.358,
|
| 1347 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1348 |
+
"rewards/chosen": 0.33814600110054016,
|
| 1349 |
+
"rewards/margins": 0.35173219442367554,
|
| 1350 |
+
"rewards/rejected": -0.013586151413619518,
|
| 1351 |
+
"step": 820
|
| 1352 |
+
},
|
| 1353 |
+
{
|
| 1354 |
+
"epoch": 2.16,
|
| 1355 |
+
"grad_norm": 52.0,
|
| 1356 |
+
"learning_rate": 2.5963092704273302e-06,
|
| 1357 |
+
"logits/chosen": -2.01682710647583,
|
| 1358 |
+
"logits/rejected": -2.0210189819335938,
|
| 1359 |
+
"logps/chosen": -34.396095275878906,
|
| 1360 |
+
"logps/rejected": -29.328195571899414,
|
| 1361 |
+
"loss": 0.1479,
|
| 1362 |
+
"rewards/accuracies": 0.9375,
|
| 1363 |
+
"rewards/chosen": 0.4560183584690094,
|
| 1364 |
+
"rewards/margins": 0.5099755525588989,
|
| 1365 |
+
"rewards/rejected": -0.05395715311169624,
|
| 1366 |
+
"step": 830
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 2.18,
|
| 1370 |
+
"grad_norm": 41.25,
|
| 1371 |
+
"learning_rate": 2.53966490958702e-06,
|
| 1372 |
+
"logits/chosen": -2.016162157058716,
|
| 1373 |
+
"logits/rejected": -2.0238842964172363,
|
| 1374 |
+
"logps/chosen": -34.38965606689453,
|
| 1375 |
+
"logps/rejected": -29.9200439453125,
|
| 1376 |
+
"loss": 0.2734,
|
| 1377 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1378 |
+
"rewards/chosen": 0.3585363030433655,
|
| 1379 |
+
"rewards/margins": 0.42364779114723206,
|
| 1380 |
+
"rewards/rejected": -0.06511145085096359,
|
| 1381 |
+
"step": 840
|
| 1382 |
+
},
|
| 1383 |
+
{
|
| 1384 |
+
"epoch": 2.21,
|
| 1385 |
+
"grad_norm": 51.0,
|
| 1386 |
+
"learning_rate": 2.4830001707654135e-06,
|
| 1387 |
+
"logits/chosen": -1.9484409093856812,
|
| 1388 |
+
"logits/rejected": -1.9388576745986938,
|
| 1389 |
+
"logps/chosen": -31.393484115600586,
|
| 1390 |
+
"logps/rejected": -32.06694030761719,
|
| 1391 |
+
"loss": 0.1541,
|
| 1392 |
+
"rewards/accuracies": 0.9375,
|
| 1393 |
+
"rewards/chosen": 0.35973984003067017,
|
| 1394 |
+
"rewards/margins": 0.40756136178970337,
|
| 1395 |
+
"rewards/rejected": -0.04782148450613022,
|
| 1396 |
+
"step": 850
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 2.23,
|
| 1400 |
+
"grad_norm": 28.75,
|
| 1401 |
+
"learning_rate": 2.4263441656635054e-06,
|
| 1402 |
+
"logits/chosen": -2.0893542766571045,
|
| 1403 |
+
"logits/rejected": -2.079577922821045,
|
| 1404 |
+
"logps/chosen": -25.801733016967773,
|
| 1405 |
+
"logps/rejected": -29.940242767333984,
|
| 1406 |
+
"loss": 0.1453,
|
| 1407 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1408 |
+
"rewards/chosen": 0.38010627031326294,
|
| 1409 |
+
"rewards/margins": 0.44122153520584106,
|
| 1410 |
+
"rewards/rejected": -0.06111522391438484,
|
| 1411 |
+
"step": 860
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.26,
|
| 1415 |
+
"grad_norm": 63.5,
|
| 1416 |
+
"learning_rate": 2.3697260014953107e-06,
|
| 1417 |
+
"logits/chosen": -1.9405052661895752,
|
| 1418 |
+
"logits/rejected": -1.9417035579681396,
|
| 1419 |
+
"logps/chosen": -33.93370819091797,
|
| 1420 |
+
"logps/rejected": -30.114543914794922,
|
| 1421 |
+
"loss": 0.2379,
|
| 1422 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1423 |
+
"rewards/chosen": 0.35936418175697327,
|
| 1424 |
+
"rewards/margins": 0.478200763463974,
|
| 1425 |
+
"rewards/rejected": -0.11883661895990372,
|
| 1426 |
+
"step": 870
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"epoch": 2.29,
|
| 1430 |
+
"grad_norm": 29.25,
|
| 1431 |
+
"learning_rate": 2.3131747660339396e-06,
|
| 1432 |
+
"logits/chosen": -1.9570789337158203,
|
| 1433 |
+
"logits/rejected": -1.957831621170044,
|
| 1434 |
+
"logps/chosen": -31.962173461914062,
|
| 1435 |
+
"logps/rejected": -33.07716751098633,
|
| 1436 |
+
"loss": 0.2589,
|
| 1437 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1438 |
+
"rewards/chosen": 0.3843681216239929,
|
| 1439 |
+
"rewards/margins": 0.40799227356910706,
|
| 1440 |
+
"rewards/rejected": -0.023624107241630554,
|
| 1441 |
+
"step": 880
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.31,
|
| 1445 |
+
"grad_norm": 30.625,
|
| 1446 |
+
"learning_rate": 2.256719512667651e-06,
|
| 1447 |
+
"logits/chosen": -1.860047698020935,
|
| 1448 |
+
"logits/rejected": -1.858214020729065,
|
| 1449 |
+
"logps/chosen": -35.269798278808594,
|
| 1450 |
+
"logps/rejected": -35.90485382080078,
|
| 1451 |
+
"loss": 0.2443,
|
| 1452 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1453 |
+
"rewards/chosen": 0.3556040823459625,
|
| 1454 |
+
"rewards/margins": 0.3804590404033661,
|
| 1455 |
+
"rewards/rejected": -0.024854950606822968,
|
| 1456 |
+
"step": 890
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"epoch": 2.34,
|
| 1460 |
+
"grad_norm": 104.5,
|
| 1461 |
+
"learning_rate": 2.2003892454735786e-06,
|
| 1462 |
+
"logits/chosen": -2.0041489601135254,
|
| 1463 |
+
"logits/rejected": -1.9973175525665283,
|
| 1464 |
+
"logps/chosen": -31.497005462646484,
|
| 1465 |
+
"logps/rejected": -32.764808654785156,
|
| 1466 |
+
"loss": 0.2502,
|
| 1467 |
+
"rewards/accuracies": 0.875,
|
| 1468 |
+
"rewards/chosen": 0.45506516098976135,
|
| 1469 |
+
"rewards/margins": 0.46113377809524536,
|
| 1470 |
+
"rewards/rejected": -0.006068611983209848,
|
| 1471 |
+
"step": 900
|
| 1472 |
+
},
|
| 1473 |
+
{
|
| 1474 |
+
"epoch": 2.34,
|
| 1475 |
+
"eval_logits/chosen": -2.248673439025879,
|
| 1476 |
+
"eval_logits/rejected": -2.2437942028045654,
|
| 1477 |
+
"eval_logps/chosen": -33.90636444091797,
|
| 1478 |
+
"eval_logps/rejected": -37.37092971801758,
|
| 1479 |
+
"eval_loss": 0.40991199016571045,
|
| 1480 |
+
"eval_rewards/accuracies": 0.49003320932388306,
|
| 1481 |
+
"eval_rewards/chosen": 0.11537089198827744,
|
| 1482 |
+
"eval_rewards/margins": -0.015749912708997726,
|
| 1483 |
+
"eval_rewards/rejected": 0.13112080097198486,
|
| 1484 |
+
"eval_runtime": 145.395,
|
| 1485 |
+
"eval_samples_per_second": 2.359,
|
| 1486 |
+
"eval_steps_per_second": 0.296,
|
| 1487 |
+
"step": 900
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 2.36,
|
| 1491 |
+
"grad_norm": 15.8125,
|
| 1492 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 1493 |
+
"logits/chosen": -2.0507731437683105,
|
| 1494 |
+
"logits/rejected": -2.0459985733032227,
|
| 1495 |
+
"logps/chosen": -33.123680114746094,
|
| 1496 |
+
"logps/rejected": -34.560203552246094,
|
| 1497 |
+
"loss": 0.1461,
|
| 1498 |
+
"rewards/accuracies": 0.875,
|
| 1499 |
+
"rewards/chosen": 0.39060264825820923,
|
| 1500 |
+
"rewards/margins": 0.4190269410610199,
|
| 1501 |
+
"rewards/rejected": -0.028424326330423355,
|
| 1502 |
+
"step": 910
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 2.39,
|
| 1506 |
+
"grad_norm": 57.25,
|
| 1507 |
+
"learning_rate": 2.088219349982323e-06,
|
| 1508 |
+
"logits/chosen": -1.9670826196670532,
|
| 1509 |
+
"logits/rejected": -1.9720354080200195,
|
| 1510 |
+
"logps/chosen": -34.90446090698242,
|
| 1511 |
+
"logps/rejected": -32.519283294677734,
|
| 1512 |
+
"loss": 0.2881,
|
| 1513 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1514 |
+
"rewards/chosen": 0.3997105062007904,
|
| 1515 |
+
"rewards/margins": 0.4171512722969055,
|
| 1516 |
+
"rewards/rejected": -0.01744077168405056,
|
| 1517 |
+
"step": 920
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 2.42,
|
| 1521 |
+
"grad_norm": 26.375,
|
| 1522 |
+
"learning_rate": 2.0324373493478803e-06,
|
| 1523 |
+
"logits/chosen": -2.0626909732818604,
|
| 1524 |
+
"logits/rejected": -2.053279161453247,
|
| 1525 |
+
"logps/chosen": -31.813243865966797,
|
| 1526 |
+
"logps/rejected": -34.46527099609375,
|
| 1527 |
+
"loss": 0.1157,
|
| 1528 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1529 |
+
"rewards/chosen": 0.4245172142982483,
|
| 1530 |
+
"rewards/margins": 0.44474273920059204,
|
| 1531 |
+
"rewards/rejected": -0.020225506275892258,
|
| 1532 |
+
"step": 930
|
| 1533 |
+
},
|
| 1534 |
+
{
|
| 1535 |
+
"epoch": 2.44,
|
| 1536 |
+
"grad_norm": 85.0,
|
| 1537 |
+
"learning_rate": 1.976895560604729e-06,
|
| 1538 |
+
"logits/chosen": -1.98918879032135,
|
| 1539 |
+
"logits/rejected": -1.986096739768982,
|
| 1540 |
+
"logps/chosen": -30.814416885375977,
|
| 1541 |
+
"logps/rejected": -32.54015350341797,
|
| 1542 |
+
"loss": 0.2202,
|
| 1543 |
+
"rewards/accuracies": 0.875,
|
| 1544 |
+
"rewards/chosen": 0.4266343116760254,
|
| 1545 |
+
"rewards/margins": 0.49393558502197266,
|
| 1546 |
+
"rewards/rejected": -0.06730121374130249,
|
| 1547 |
+
"step": 940
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 2.47,
|
| 1551 |
+
"grad_norm": 15.1875,
|
| 1552 |
+
"learning_rate": 1.921622518534466e-06,
|
| 1553 |
+
"logits/chosen": -1.932186484336853,
|
| 1554 |
+
"logits/rejected": -1.9396946430206299,
|
| 1555 |
+
"logps/chosen": -32.45960235595703,
|
| 1556 |
+
"logps/rejected": -35.62566375732422,
|
| 1557 |
+
"loss": 0.1934,
|
| 1558 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1559 |
+
"rewards/chosen": 0.3703078627586365,
|
| 1560 |
+
"rewards/margins": 0.43846622109413147,
|
| 1561 |
+
"rewards/rejected": -0.06815837323665619,
|
| 1562 |
+
"step": 950
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 2.49,
|
| 1566 |
+
"grad_norm": 73.5,
|
| 1567 |
+
"learning_rate": 1.8666466198491794e-06,
|
| 1568 |
+
"logits/chosen": -1.940987229347229,
|
| 1569 |
+
"logits/rejected": -1.934152603149414,
|
| 1570 |
+
"logps/chosen": -33.384063720703125,
|
| 1571 |
+
"logps/rejected": -36.64484405517578,
|
| 1572 |
+
"loss": 0.1978,
|
| 1573 |
+
"rewards/accuracies": 0.875,
|
| 1574 |
+
"rewards/chosen": 0.4000038206577301,
|
| 1575 |
+
"rewards/margins": 0.44143620133399963,
|
| 1576 |
+
"rewards/rejected": -0.041432395577430725,
|
| 1577 |
+
"step": 960
|
| 1578 |
+
},
|
| 1579 |
+
{
|
| 1580 |
+
"epoch": 2.52,
|
| 1581 |
+
"grad_norm": 18.0,
|
| 1582 |
+
"learning_rate": 1.8119961086025376e-06,
|
| 1583 |
+
"logits/chosen": -1.942278265953064,
|
| 1584 |
+
"logits/rejected": -1.9429972171783447,
|
| 1585 |
+
"logps/chosen": -30.803075790405273,
|
| 1586 |
+
"logps/rejected": -33.071388244628906,
|
| 1587 |
+
"loss": 0.2326,
|
| 1588 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1589 |
+
"rewards/chosen": 0.38999220728874207,
|
| 1590 |
+
"rewards/margins": 0.47110748291015625,
|
| 1591 |
+
"rewards/rejected": -0.0811152458190918,
|
| 1592 |
+
"step": 970
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 2.55,
|
| 1596 |
+
"grad_norm": 31.75,
|
| 1597 |
+
"learning_rate": 1.7576990616793139e-06,
|
| 1598 |
+
"logits/chosen": -1.946674108505249,
|
| 1599 |
+
"logits/rejected": -1.9579198360443115,
|
| 1600 |
+
"logps/chosen": -32.60912322998047,
|
| 1601 |
+
"logps/rejected": -34.126827239990234,
|
| 1602 |
+
"loss": 0.264,
|
| 1603 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1604 |
+
"rewards/chosen": 0.4208148419857025,
|
| 1605 |
+
"rewards/margins": 0.4168703556060791,
|
| 1606 |
+
"rewards/rejected": 0.003944486379623413,
|
| 1607 |
+
"step": 980
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 2.57,
|
| 1611 |
+
"grad_norm": 42.0,
|
| 1612 |
+
"learning_rate": 1.7037833743707892e-06,
|
| 1613 |
+
"logits/chosen": -2.0197396278381348,
|
| 1614 |
+
"logits/rejected": -2.017103672027588,
|
| 1615 |
+
"logps/chosen": -35.6496696472168,
|
| 1616 |
+
"logps/rejected": -32.41356658935547,
|
| 1617 |
+
"loss": 0.2859,
|
| 1618 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1619 |
+
"rewards/chosen": 0.4242428243160248,
|
| 1620 |
+
"rewards/margins": 0.4371464252471924,
|
| 1621 |
+
"rewards/rejected": -0.012903605587780476,
|
| 1622 |
+
"step": 990
|
| 1623 |
+
},
|
| 1624 |
+
{
|
| 1625 |
+
"epoch": 2.6,
|
| 1626 |
+
"grad_norm": 35.25,
|
| 1627 |
+
"learning_rate": 1.6502767460434588e-06,
|
| 1628 |
+
"logits/chosen": -1.9874387979507446,
|
| 1629 |
+
"logits/rejected": -1.9911830425262451,
|
| 1630 |
+
"logps/chosen": -34.0231819152832,
|
| 1631 |
+
"logps/rejected": -34.56896209716797,
|
| 1632 |
+
"loss": 0.1529,
|
| 1633 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1634 |
+
"rewards/chosen": 0.39610037207603455,
|
| 1635 |
+
"rewards/margins": 0.47072356939315796,
|
| 1636 |
+
"rewards/rejected": -0.07462319731712341,
|
| 1637 |
+
"step": 1000
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 2.6,
|
| 1641 |
+
"eval_logits/chosen": -2.2477176189422607,
|
| 1642 |
+
"eval_logits/rejected": -2.242847442626953,
|
| 1643 |
+
"eval_logps/chosen": -33.89875411987305,
|
| 1644 |
+
"eval_logps/rejected": -37.377010345458984,
|
| 1645 |
+
"eval_loss": 0.3879217505455017,
|
| 1646 |
+
"eval_rewards/accuracies": 0.5215947031974792,
|
| 1647 |
+
"eval_rewards/chosen": 0.12221997231245041,
|
| 1648 |
+
"eval_rewards/margins": -0.003430851735174656,
|
| 1649 |
+
"eval_rewards/rejected": 0.1256508231163025,
|
| 1650 |
+
"eval_runtime": 145.3271,
|
| 1651 |
+
"eval_samples_per_second": 2.36,
|
| 1652 |
+
"eval_steps_per_second": 0.296,
|
| 1653 |
+
"step": 1000
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 2.62,
|
| 1657 |
+
"grad_norm": 40.25,
|
| 1658 |
+
"learning_rate": 1.5972066659083796e-06,
|
| 1659 |
+
"logits/chosen": -1.9966695308685303,
|
| 1660 |
+
"logits/rejected": -2.0017590522766113,
|
| 1661 |
+
"logps/chosen": -32.33439636230469,
|
| 1662 |
+
"logps/rejected": -32.539424896240234,
|
| 1663 |
+
"loss": 0.1576,
|
| 1664 |
+
"rewards/accuracies": 0.875,
|
| 1665 |
+
"rewards/chosen": 0.41942542791366577,
|
| 1666 |
+
"rewards/margins": 0.43577662110328674,
|
| 1667 |
+
"rewards/rejected": -0.01635116897523403,
|
| 1668 |
+
"step": 1010
|
| 1669 |
+
},
|
| 1670 |
+
{
|
| 1671 |
+
"epoch": 2.65,
|
| 1672 |
+
"grad_norm": 53.25,
|
| 1673 |
+
"learning_rate": 1.5446003988985041e-06,
|
| 1674 |
+
"logits/chosen": -1.959848165512085,
|
| 1675 |
+
"logits/rejected": -1.962561011314392,
|
| 1676 |
+
"logps/chosen": -30.024948120117188,
|
| 1677 |
+
"logps/rejected": -31.70804786682129,
|
| 1678 |
+
"loss": 0.2273,
|
| 1679 |
+
"rewards/accuracies": 0.875,
|
| 1680 |
+
"rewards/chosen": 0.3786233961582184,
|
| 1681 |
+
"rewards/margins": 0.3866207003593445,
|
| 1682 |
+
"rewards/rejected": -0.007997283712029457,
|
| 1683 |
+
"step": 1020
|
| 1684 |
+
},
|
| 1685 |
+
{
|
| 1686 |
+
"epoch": 2.68,
|
| 1687 |
+
"grad_norm": 23.125,
|
| 1688 |
+
"learning_rate": 1.4924849716612211e-06,
|
| 1689 |
+
"logits/chosen": -2.017338514328003,
|
| 1690 |
+
"logits/rejected": -2.0114893913269043,
|
| 1691 |
+
"logps/chosen": -33.496673583984375,
|
| 1692 |
+
"logps/rejected": -33.55792999267578,
|
| 1693 |
+
"loss": 0.1431,
|
| 1694 |
+
"rewards/accuracies": 0.9375,
|
| 1695 |
+
"rewards/chosen": 0.4431969225406647,
|
| 1696 |
+
"rewards/margins": 0.4791356027126312,
|
| 1697 |
+
"rewards/rejected": -0.035938672721385956,
|
| 1698 |
+
"step": 1030
|
| 1699 |
+
},
|
| 1700 |
+
{
|
| 1701 |
+
"epoch": 2.7,
|
| 1702 |
+
"grad_norm": 56.0,
|
| 1703 |
+
"learning_rate": 1.440887158673332e-06,
|
| 1704 |
+
"logits/chosen": -1.9453651905059814,
|
| 1705 |
+
"logits/rejected": -1.9490783214569092,
|
| 1706 |
+
"logps/chosen": -35.259788513183594,
|
| 1707 |
+
"logps/rejected": -34.503028869628906,
|
| 1708 |
+
"loss": 0.2102,
|
| 1709 |
+
"rewards/accuracies": 0.9375,
|
| 1710 |
+
"rewards/chosen": 0.4445478916168213,
|
| 1711 |
+
"rewards/margins": 0.519649863243103,
|
| 1712 |
+
"rewards/rejected": -0.07510194182395935,
|
| 1713 |
+
"step": 1040
|
| 1714 |
+
},
|
| 1715 |
+
{
|
| 1716 |
+
"epoch": 2.73,
|
| 1717 |
+
"grad_norm": 35.75,
|
| 1718 |
+
"learning_rate": 1.3898334684855647e-06,
|
| 1719 |
+
"logits/chosen": -1.9455549716949463,
|
| 1720 |
+
"logits/rejected": -1.95868718624115,
|
| 1721 |
+
"logps/chosen": -31.420841217041016,
|
| 1722 |
+
"logps/rejected": -33.076026916503906,
|
| 1723 |
+
"loss": 0.2334,
|
| 1724 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1725 |
+
"rewards/chosen": 0.39784207940101624,
|
| 1726 |
+
"rewards/margins": 0.4482339024543762,
|
| 1727 |
+
"rewards/rejected": -0.05039183050394058,
|
| 1728 |
+
"step": 1050
|
| 1729 |
+
},
|
| 1730 |
+
{
|
| 1731 |
+
"epoch": 2.75,
|
| 1732 |
+
"grad_norm": 58.75,
|
| 1733 |
+
"learning_rate": 1.3393501301037245e-06,
|
| 1734 |
+
"logits/chosen": -1.9375944137573242,
|
| 1735 |
+
"logits/rejected": -1.9310089349746704,
|
| 1736 |
+
"logps/chosen": -31.681589126586914,
|
| 1737 |
+
"logps/rejected": -33.36168670654297,
|
| 1738 |
+
"loss": 0.1253,
|
| 1739 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1740 |
+
"rewards/chosen": 0.42997032403945923,
|
| 1741 |
+
"rewards/margins": 0.46931910514831543,
|
| 1742 |
+
"rewards/rejected": -0.03934880718588829,
|
| 1743 |
+
"step": 1060
|
| 1744 |
+
},
|
| 1745 |
+
{
|
| 1746 |
+
"epoch": 2.78,
|
| 1747 |
+
"grad_norm": 40.5,
|
| 1748 |
+
"learning_rate": 1.2894630795134454e-06,
|
| 1749 |
+
"logits/chosen": -2.0641446113586426,
|
| 1750 |
+
"logits/rejected": -2.0649874210357666,
|
| 1751 |
+
"logps/chosen": -33.09912109375,
|
| 1752 |
+
"logps/rejected": -32.90647888183594,
|
| 1753 |
+
"loss": 0.1818,
|
| 1754 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1755 |
+
"rewards/chosen": 0.4031626284122467,
|
| 1756 |
+
"rewards/margins": 0.4587678909301758,
|
| 1757 |
+
"rewards/rejected": -0.05560523271560669,
|
| 1758 |
+
"step": 1070
|
| 1759 |
+
},
|
| 1760 |
+
{
|
| 1761 |
+
"epoch": 2.81,
|
| 1762 |
+
"grad_norm": 25.25,
|
| 1763 |
+
"learning_rate": 1.2401979463554984e-06,
|
| 1764 |
+
"logits/chosen": -1.9966720342636108,
|
| 1765 |
+
"logits/rejected": -1.9955018758773804,
|
| 1766 |
+
"logps/chosen": -33.611534118652344,
|
| 1767 |
+
"logps/rejected": -33.192691802978516,
|
| 1768 |
+
"loss": 0.1975,
|
| 1769 |
+
"rewards/accuracies": 0.8125,
|
| 1770 |
+
"rewards/chosen": 0.43046802282333374,
|
| 1771 |
+
"rewards/margins": 0.431241899728775,
|
| 1772 |
+
"rewards/rejected": -0.0007738359272480011,
|
| 1773 |
+
"step": 1080
|
| 1774 |
+
},
|
| 1775 |
+
{
|
| 1776 |
+
"epoch": 2.83,
|
| 1777 |
+
"grad_norm": 24.625,
|
| 1778 |
+
"learning_rate": 1.1915800407584705e-06,
|
| 1779 |
+
"logits/chosen": -2.0245463848114014,
|
| 1780 |
+
"logits/rejected": -2.0170249938964844,
|
| 1781 |
+
"logps/chosen": -33.81514358520508,
|
| 1782 |
+
"logps/rejected": -31.061843872070312,
|
| 1783 |
+
"loss": 0.1443,
|
| 1784 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1785 |
+
"rewards/chosen": 0.3696584105491638,
|
| 1786 |
+
"rewards/margins": 0.387938916683197,
|
| 1787 |
+
"rewards/rejected": -0.018280521035194397,
|
| 1788 |
+
"step": 1090
|
| 1789 |
+
},
|
| 1790 |
+
{
|
| 1791 |
+
"epoch": 2.86,
|
| 1792 |
+
"grad_norm": 88.0,
|
| 1793 |
+
"learning_rate": 1.1436343403356019e-06,
|
| 1794 |
+
"logits/chosen": -1.9909054040908813,
|
| 1795 |
+
"logits/rejected": -1.9901247024536133,
|
| 1796 |
+
"logps/chosen": -34.939388275146484,
|
| 1797 |
+
"logps/rejected": -36.37110137939453,
|
| 1798 |
+
"loss": 0.1583,
|
| 1799 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1800 |
+
"rewards/chosen": 0.43112340569496155,
|
| 1801 |
+
"rewards/margins": 0.44767728447914124,
|
| 1802 |
+
"rewards/rejected": -0.01655389368534088,
|
| 1803 |
+
"step": 1100
|
| 1804 |
+
},
|
| 1805 |
+
{
|
| 1806 |
+
"epoch": 2.86,
|
| 1807 |
+
"eval_logits/chosen": -2.2481679916381836,
|
| 1808 |
+
"eval_logits/rejected": -2.243297576904297,
|
| 1809 |
+
"eval_logps/chosen": -33.90199279785156,
|
| 1810 |
+
"eval_logps/rejected": -37.37769317626953,
|
| 1811 |
+
"eval_loss": 0.3968319892883301,
|
| 1812 |
+
"eval_rewards/accuracies": 0.4875415563583374,
|
| 1813 |
+
"eval_rewards/chosen": 0.11930261552333832,
|
| 1814 |
+
"eval_rewards/margins": -0.0057318080216646194,
|
| 1815 |
+
"eval_rewards/rejected": 0.1250344216823578,
|
| 1816 |
+
"eval_runtime": 145.2343,
|
| 1817 |
+
"eval_samples_per_second": 2.362,
|
| 1818 |
+
"eval_steps_per_second": 0.296,
|
| 1819 |
+
"step": 1100
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"epoch": 2.88,
|
| 1823 |
+
"grad_norm": 34.0,
|
| 1824 |
+
"learning_rate": 1.0963854773524548e-06,
|
| 1825 |
+
"logits/chosen": -2.0373494625091553,
|
| 1826 |
+
"logits/rejected": -2.0427050590515137,
|
| 1827 |
+
"logps/chosen": -35.03519821166992,
|
| 1828 |
+
"logps/rejected": -35.56550216674805,
|
| 1829 |
+
"loss": 0.1969,
|
| 1830 |
+
"rewards/accuracies": 0.875,
|
| 1831 |
+
"rewards/chosen": 0.38872379064559937,
|
| 1832 |
+
"rewards/margins": 0.40845227241516113,
|
| 1833 |
+
"rewards/rejected": -0.01972847804427147,
|
| 1834 |
+
"step": 1110
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"epoch": 2.91,
|
| 1838 |
+
"grad_norm": 91.0,
|
| 1839 |
+
"learning_rate": 1.049857726072005e-06,
|
| 1840 |
+
"logits/chosen": -2.009969472885132,
|
| 1841 |
+
"logits/rejected": -2.0080225467681885,
|
| 1842 |
+
"logps/chosen": -32.2147216796875,
|
| 1843 |
+
"logps/rejected": -32.992332458496094,
|
| 1844 |
+
"loss": 0.2049,
|
| 1845 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1846 |
+
"rewards/chosen": 0.36133700609207153,
|
| 1847 |
+
"rewards/margins": 0.36995649337768555,
|
| 1848 |
+
"rewards/rejected": -0.008619492873549461,
|
| 1849 |
+
"step": 1120
|
| 1850 |
+
},
|
| 1851 |
+
{
|
| 1852 |
+
"epoch": 2.94,
|
| 1853 |
+
"grad_norm": 42.0,
|
| 1854 |
+
"learning_rate": 1.0040749902836508e-06,
|
| 1855 |
+
"logits/chosen": -1.9126440286636353,
|
| 1856 |
+
"logits/rejected": -1.9155027866363525,
|
| 1857 |
+
"logps/chosen": -29.281444549560547,
|
| 1858 |
+
"logps/rejected": -30.237567901611328,
|
| 1859 |
+
"loss": 0.18,
|
| 1860 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1861 |
+
"rewards/chosen": 0.42047175765037537,
|
| 1862 |
+
"rewards/margins": 0.4137454926967621,
|
| 1863 |
+
"rewards/rejected": 0.006726269610226154,
|
| 1864 |
+
"step": 1130
|
| 1865 |
+
},
|
| 1866 |
+
{
|
| 1867 |
+
"epoch": 2.96,
|
| 1868 |
+
"grad_norm": 33.0,
|
| 1869 |
+
"learning_rate": 9.59060791022566e-07,
|
| 1870 |
+
"logits/chosen": -2.009428024291992,
|
| 1871 |
+
"logits/rejected": -2.0063159465789795,
|
| 1872 |
+
"logps/chosen": -32.74279022216797,
|
| 1873 |
+
"logps/rejected": -32.62069320678711,
|
| 1874 |
+
"loss": 0.1567,
|
| 1875 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1876 |
+
"rewards/chosen": 0.4148639142513275,
|
| 1877 |
+
"rewards/margins": 0.49929580092430115,
|
| 1878 |
+
"rewards/rejected": -0.08443190902471542,
|
| 1879 |
+
"step": 1140
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 2.99,
|
| 1883 |
+
"grad_norm": 50.25,
|
| 1884 |
+
"learning_rate": 9.148382544856885e-07,
|
| 1885 |
+
"logits/chosen": -1.931823968887329,
|
| 1886 |
+
"logits/rejected": -1.925490140914917,
|
| 1887 |
+
"logps/chosen": -27.730154037475586,
|
| 1888 |
+
"logps/rejected": -32.26335906982422,
|
| 1889 |
+
"loss": 0.1296,
|
| 1890 |
+
"rewards/accuracies": 0.9375,
|
| 1891 |
+
"rewards/chosen": 0.3670499324798584,
|
| 1892 |
+
"rewards/margins": 0.47060221433639526,
|
| 1893 |
+
"rewards/rejected": -0.10355228185653687,
|
| 1894 |
+
"step": 1150
|
| 1895 |
+
},
|
| 1896 |
+
{
|
| 1897 |
+
"epoch": 3.01,
|
| 1898 |
+
"grad_norm": 35.25,
|
| 1899 |
+
"learning_rate": 8.714301001505568e-07,
|
| 1900 |
+
"logits/chosen": -2.0501132011413574,
|
| 1901 |
+
"logits/rejected": -2.047624111175537,
|
| 1902 |
+
"logps/chosen": -32.637855529785156,
|
| 1903 |
+
"logps/rejected": -35.0833625793457,
|
| 1904 |
+
"loss": 0.0915,
|
| 1905 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1906 |
+
"rewards/chosen": 0.41239070892333984,
|
| 1907 |
+
"rewards/margins": 0.4050343632698059,
|
| 1908 |
+
"rewards/rejected": 0.007356312125921249,
|
| 1909 |
+
"step": 1160
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.04,
|
| 1913 |
+
"grad_norm": 17.625,
|
| 1914 |
+
"learning_rate": 8.288586291031025e-07,
|
| 1915 |
+
"logits/chosen": -2.0300636291503906,
|
| 1916 |
+
"logits/rejected": -2.0281882286071777,
|
| 1917 |
+
"logps/chosen": -31.161453247070312,
|
| 1918 |
+
"logps/rejected": -32.45496368408203,
|
| 1919 |
+
"loss": 0.0941,
|
| 1920 |
+
"rewards/accuracies": 0.9375,
|
| 1921 |
+
"rewards/chosen": 0.4076564908027649,
|
| 1922 |
+
"rewards/margins": 0.4894997179508209,
|
| 1923 |
+
"rewards/rejected": -0.08184324204921722,
|
| 1924 |
+
"step": 1170
|
| 1925 |
+
},
|
| 1926 |
+
{
|
| 1927 |
+
"epoch": 3.06,
|
| 1928 |
+
"grad_norm": 32.25,
|
| 1929 |
+
"learning_rate": 7.871457125803897e-07,
|
| 1930 |
+
"logits/chosen": -2.02311110496521,
|
| 1931 |
+
"logits/rejected": -2.0107522010803223,
|
| 1932 |
+
"logps/chosen": -35.538490295410156,
|
| 1933 |
+
"logps/rejected": -34.12282180786133,
|
| 1934 |
+
"loss": 0.1417,
|
| 1935 |
+
"rewards/accuracies": 0.9375,
|
| 1936 |
+
"rewards/chosen": 0.4561285972595215,
|
| 1937 |
+
"rewards/margins": 0.5023427605628967,
|
| 1938 |
+
"rewards/rejected": -0.046214111149311066,
|
| 1939 |
+
"step": 1180
|
| 1940 |
+
},
|
| 1941 |
+
{
|
| 1942 |
+
"epoch": 3.09,
|
| 1943 |
+
"grad_norm": 55.0,
|
| 1944 |
+
"learning_rate": 7.463127807341966e-07,
|
| 1945 |
+
"logits/chosen": -1.9117752313613892,
|
| 1946 |
+
"logits/rejected": -1.9064388275146484,
|
| 1947 |
+
"logps/chosen": -34.895999908447266,
|
| 1948 |
+
"logps/rejected": -34.239891052246094,
|
| 1949 |
+
"loss": 0.1368,
|
| 1950 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1951 |
+
"rewards/chosen": 0.4083231985569,
|
| 1952 |
+
"rewards/margins": 0.41019004583358765,
|
| 1953 |
+
"rewards/rejected": -0.001866865553893149,
|
| 1954 |
+
"step": 1190
|
| 1955 |
+
},
|
| 1956 |
+
{
|
| 1957 |
+
"epoch": 3.12,
|
| 1958 |
+
"grad_norm": 54.5,
|
| 1959 |
+
"learning_rate": 7.063808116212021e-07,
|
| 1960 |
+
"logits/chosen": -1.9550445079803467,
|
| 1961 |
+
"logits/rejected": -1.958034873008728,
|
| 1962 |
+
"logps/chosen": -32.1730842590332,
|
| 1963 |
+
"logps/rejected": -31.49075698852539,
|
| 1964 |
+
"loss": 0.113,
|
| 1965 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1966 |
+
"rewards/chosen": 0.39780890941619873,
|
| 1967 |
+
"rewards/margins": 0.4019277095794678,
|
| 1968 |
+
"rewards/rejected": -0.004118794109672308,
|
| 1969 |
+
"step": 1200
|
| 1970 |
+
},
|
| 1971 |
+
{
|
| 1972 |
+
"epoch": 3.12,
|
| 1973 |
+
"eval_logits/chosen": -2.2470004558563232,
|
| 1974 |
+
"eval_logits/rejected": -2.2421324253082275,
|
| 1975 |
+
"eval_logps/chosen": -33.908164978027344,
|
| 1976 |
+
"eval_logps/rejected": -37.38742446899414,
|
| 1977 |
+
"eval_loss": 0.3848971724510193,
|
| 1978 |
+
"eval_rewards/accuracies": 0.47840529680252075,
|
| 1979 |
+
"eval_rewards/chosen": 0.11374907940626144,
|
| 1980 |
+
"eval_rewards/margins": -0.002530111465603113,
|
| 1981 |
+
"eval_rewards/rejected": 0.11627919226884842,
|
| 1982 |
+
"eval_runtime": 145.1536,
|
| 1983 |
+
"eval_samples_per_second": 2.363,
|
| 1984 |
+
"eval_steps_per_second": 0.296,
|
| 1985 |
+
"step": 1200
|
| 1986 |
+
},
|
| 1987 |
+
{
|
| 1988 |
+
"epoch": 3.14,
|
| 1989 |
+
"grad_norm": 23.875,
|
| 1990 |
+
"learning_rate": 6.673703204254348e-07,
|
| 1991 |
+
"logits/chosen": -1.9873783588409424,
|
| 1992 |
+
"logits/rejected": -1.9827861785888672,
|
| 1993 |
+
"logps/chosen": -30.8137264251709,
|
| 1994 |
+
"logps/rejected": -30.97092628479004,
|
| 1995 |
+
"loss": 0.1001,
|
| 1996 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1997 |
+
"rewards/chosen": 0.41394877433776855,
|
| 1998 |
+
"rewards/margins": 0.4704950451850891,
|
| 1999 |
+
"rewards/rejected": -0.05654625967144966,
|
| 2000 |
+
"step": 1210
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 3.17,
|
| 2004 |
+
"grad_norm": 18.875,
|
| 2005 |
+
"learning_rate": 6.293013489185315e-07,
|
| 2006 |
+
"logits/chosen": -1.9986066818237305,
|
| 2007 |
+
"logits/rejected": -1.9930963516235352,
|
| 2008 |
+
"logps/chosen": -34.82990264892578,
|
| 2009 |
+
"logps/rejected": -34.601905822753906,
|
| 2010 |
+
"loss": 0.1136,
|
| 2011 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2012 |
+
"rewards/chosen": 0.399636447429657,
|
| 2013 |
+
"rewards/margins": 0.4284568727016449,
|
| 2014 |
+
"rewards/rejected": -0.02882050909101963,
|
| 2015 |
+
"step": 1220
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 3.19,
|
| 2019 |
+
"grad_norm": 35.0,
|
| 2020 |
+
"learning_rate": 5.921934551632086e-07,
|
| 2021 |
+
"logits/chosen": -2.000286102294922,
|
| 2022 |
+
"logits/rejected": -1.9870882034301758,
|
| 2023 |
+
"logps/chosen": -32.6419677734375,
|
| 2024 |
+
"logps/rejected": -34.18825149536133,
|
| 2025 |
+
"loss": 0.1139,
|
| 2026 |
+
"rewards/accuracies": 0.875,
|
| 2027 |
+
"rewards/chosen": 0.4068435728549957,
|
| 2028 |
+
"rewards/margins": 0.4684053957462311,
|
| 2029 |
+
"rewards/rejected": -0.06156182289123535,
|
| 2030 |
+
"step": 1230
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 3.22,
|
| 2034 |
+
"grad_norm": 29.0,
|
| 2035 |
+
"learning_rate": 5.560657034652405e-07,
|
| 2036 |
+
"logits/chosen": -2.044322967529297,
|
| 2037 |
+
"logits/rejected": -2.0414676666259766,
|
| 2038 |
+
"logps/chosen": -34.62710189819336,
|
| 2039 |
+
"logps/rejected": -31.878192901611328,
|
| 2040 |
+
"loss": 0.1302,
|
| 2041 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2042 |
+
"rewards/chosen": 0.36521658301353455,
|
| 2043 |
+
"rewards/margins": 0.399711012840271,
|
| 2044 |
+
"rewards/rejected": -0.03449442610144615,
|
| 2045 |
+
"step": 1240
|
| 2046 |
+
},
|
| 2047 |
+
{
|
| 2048 |
+
"epoch": 3.25,
|
| 2049 |
+
"grad_norm": 51.5,
|
| 2050 |
+
"learning_rate": 5.2093665457911e-07,
|
| 2051 |
+
"logits/chosen": -1.9766407012939453,
|
| 2052 |
+
"logits/rejected": -1.9737846851348877,
|
| 2053 |
+
"logps/chosen": -33.98097610473633,
|
| 2054 |
+
"logps/rejected": -34.84381103515625,
|
| 2055 |
+
"loss": 0.1699,
|
| 2056 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2057 |
+
"rewards/chosen": 0.37366223335266113,
|
| 2058 |
+
"rewards/margins": 0.4156786799430847,
|
| 2059 |
+
"rewards/rejected": -0.04201648384332657,
|
| 2060 |
+
"step": 1250
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 3.27,
|
| 2064 |
+
"grad_norm": 30.75,
|
| 2065 |
+
"learning_rate": 4.868243561723535e-07,
|
| 2066 |
+
"logits/chosen": -2.038552761077881,
|
| 2067 |
+
"logits/rejected": -2.0333991050720215,
|
| 2068 |
+
"logps/chosen": -30.33209800720215,
|
| 2069 |
+
"logps/rejected": -32.44548797607422,
|
| 2070 |
+
"loss": 0.0978,
|
| 2071 |
+
"rewards/accuracies": 0.9375,
|
| 2072 |
+
"rewards/chosen": 0.3858305513858795,
|
| 2073 |
+
"rewards/margins": 0.44247692823410034,
|
| 2074 |
+
"rewards/rejected": -0.05664638429880142,
|
| 2075 |
+
"step": 1260
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 3.3,
|
| 2079 |
+
"grad_norm": 28.375,
|
| 2080 |
+
"learning_rate": 4.537463335535161e-07,
|
| 2081 |
+
"logits/chosen": -2.065164089202881,
|
| 2082 |
+
"logits/rejected": -2.069824457168579,
|
| 2083 |
+
"logps/chosen": -32.27547073364258,
|
| 2084 |
+
"logps/rejected": -31.700992584228516,
|
| 2085 |
+
"loss": 0.1001,
|
| 2086 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2087 |
+
"rewards/chosen": 0.3919737637042999,
|
| 2088 |
+
"rewards/margins": 0.4096389412879944,
|
| 2089 |
+
"rewards/rejected": -0.017665188759565353,
|
| 2090 |
+
"step": 1270
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"epoch": 3.32,
|
| 2094 |
+
"grad_norm": 18.625,
|
| 2095 |
+
"learning_rate": 4.217195806684629e-07,
|
| 2096 |
+
"logits/chosen": -1.9599748849868774,
|
| 2097 |
+
"logits/rejected": -1.9670900106430054,
|
| 2098 |
+
"logps/chosen": -34.481956481933594,
|
| 2099 |
+
"logps/rejected": -32.67145538330078,
|
| 2100 |
+
"loss": 0.1083,
|
| 2101 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2102 |
+
"rewards/chosen": 0.3827614486217499,
|
| 2103 |
+
"rewards/margins": 0.4486331343650818,
|
| 2104 |
+
"rewards/rejected": -0.06587164849042892,
|
| 2105 |
+
"step": 1280
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 3.35,
|
| 2109 |
+
"grad_norm": 33.25,
|
| 2110 |
+
"learning_rate": 3.907605513696808e-07,
|
| 2111 |
+
"logits/chosen": -1.862562894821167,
|
| 2112 |
+
"logits/rejected": -1.864749550819397,
|
| 2113 |
+
"logps/chosen": -32.70376968383789,
|
| 2114 |
+
"logps/rejected": -36.58349609375,
|
| 2115 |
+
"loss": 0.1038,
|
| 2116 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2117 |
+
"rewards/chosen": 0.3853146433830261,
|
| 2118 |
+
"rewards/margins": 0.41162100434303284,
|
| 2119 |
+
"rewards/rejected": -0.026306327432394028,
|
| 2120 |
+
"step": 1290
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 3.38,
|
| 2124 |
+
"grad_norm": 64.0,
|
| 2125 |
+
"learning_rate": 3.6088515096305675e-07,
|
| 2126 |
+
"logits/chosen": -1.9185287952423096,
|
| 2127 |
+
"logits/rejected": -1.921884298324585,
|
| 2128 |
+
"logps/chosen": -32.8210334777832,
|
| 2129 |
+
"logps/rejected": -32.53430938720703,
|
| 2130 |
+
"loss": 0.0937,
|
| 2131 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2132 |
+
"rewards/chosen": 0.35607367753982544,
|
| 2133 |
+
"rewards/margins": 0.41920334100723267,
|
| 2134 |
+
"rewards/rejected": -0.06312969326972961,
|
| 2135 |
+
"step": 1300
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 3.38,
|
| 2139 |
+
"eval_logits/chosen": -2.2471628189086914,
|
| 2140 |
+
"eval_logits/rejected": -2.2422919273376465,
|
| 2141 |
+
"eval_logps/chosen": -33.897342681884766,
|
| 2142 |
+
"eval_logps/rejected": -37.385860443115234,
|
| 2143 |
+
"eval_loss": 0.37376365065574646,
|
| 2144 |
+
"eval_rewards/accuracies": 0.5045680999755859,
|
| 2145 |
+
"eval_rewards/chosen": 0.1234876811504364,
|
| 2146 |
+
"eval_rewards/margins": 0.005806357134133577,
|
| 2147 |
+
"eval_rewards/rejected": 0.1176813393831253,
|
| 2148 |
+
"eval_runtime": 145.1552,
|
| 2149 |
+
"eval_samples_per_second": 2.363,
|
| 2150 |
+
"eval_steps_per_second": 0.296,
|
| 2151 |
+
"step": 1300
|
| 2152 |
+
},
|
| 2153 |
+
{
|
| 2154 |
+
"epoch": 3.4,
|
| 2155 |
+
"grad_norm": 37.5,
|
| 2156 |
+
"learning_rate": 3.321087280364757e-07,
|
| 2157 |
+
"logits/chosen": -2.027667999267578,
|
| 2158 |
+
"logits/rejected": -2.0075087547302246,
|
| 2159 |
+
"logps/chosen": -30.6841983795166,
|
| 2160 |
+
"logps/rejected": -35.201026916503906,
|
| 2161 |
+
"loss": 0.135,
|
| 2162 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2163 |
+
"rewards/chosen": 0.38628271222114563,
|
| 2164 |
+
"rewards/margins": 0.4174365997314453,
|
| 2165 |
+
"rewards/rejected": -0.031153928488492966,
|
| 2166 |
+
"step": 1310
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"epoch": 3.43,
|
| 2170 |
+
"grad_norm": 141.0,
|
| 2171 |
+
"learning_rate": 3.044460665744284e-07,
|
| 2172 |
+
"logits/chosen": -2.0063624382019043,
|
| 2173 |
+
"logits/rejected": -2.012112855911255,
|
| 2174 |
+
"logps/chosen": -32.609859466552734,
|
| 2175 |
+
"logps/rejected": -32.543277740478516,
|
| 2176 |
+
"loss": 0.1282,
|
| 2177 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2178 |
+
"rewards/chosen": 0.38153642416000366,
|
| 2179 |
+
"rewards/margins": 0.4288099706172943,
|
| 2180 |
+
"rewards/rejected": -0.04727357253432274,
|
| 2181 |
+
"step": 1320
|
| 2182 |
+
},
|
| 2183 |
+
{
|
| 2184 |
+
"epoch": 3.45,
|
| 2185 |
+
"grad_norm": 13.625,
|
| 2186 |
+
"learning_rate": 2.779113783626916e-07,
|
| 2187 |
+
"logits/chosen": -1.9746071100234985,
|
| 2188 |
+
"logits/rejected": -1.969681739807129,
|
| 2189 |
+
"logps/chosen": -32.591461181640625,
|
| 2190 |
+
"logps/rejected": -34.600608825683594,
|
| 2191 |
+
"loss": 0.1059,
|
| 2192 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2193 |
+
"rewards/chosen": 0.38793960213661194,
|
| 2194 |
+
"rewards/margins": 0.4134295880794525,
|
| 2195 |
+
"rewards/rejected": -0.025490064173936844,
|
| 2196 |
+
"step": 1330
|
| 2197 |
+
},
|
| 2198 |
+
{
|
| 2199 |
+
"epoch": 3.48,
|
| 2200 |
+
"grad_norm": 35.75,
|
| 2201 |
+
"learning_rate": 2.5251829568697204e-07,
|
| 2202 |
+
"logits/chosen": -1.8414928913116455,
|
| 2203 |
+
"logits/rejected": -1.8512153625488281,
|
| 2204 |
+
"logps/chosen": -33.299381256103516,
|
| 2205 |
+
"logps/rejected": -31.048131942749023,
|
| 2206 |
+
"loss": 0.0735,
|
| 2207 |
+
"rewards/accuracies": 0.987500011920929,
|
| 2208 |
+
"rewards/chosen": 0.4501881003379822,
|
| 2209 |
+
"rewards/margins": 0.4628051221370697,
|
| 2210 |
+
"rewards/rejected": -0.012616944499313831,
|
| 2211 |
+
"step": 1340
|
| 2212 |
+
},
|
| 2213 |
+
{
|
| 2214 |
+
"epoch": 3.51,
|
| 2215 |
+
"grad_norm": 22.5,
|
| 2216 |
+
"learning_rate": 2.2827986432927774e-07,
|
| 2217 |
+
"logits/chosen": -1.9015308618545532,
|
| 2218 |
+
"logits/rejected": -1.9062535762786865,
|
| 2219 |
+
"logps/chosen": -33.4455680847168,
|
| 2220 |
+
"logps/rejected": -33.258026123046875,
|
| 2221 |
+
"loss": 0.0823,
|
| 2222 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2223 |
+
"rewards/chosen": 0.4548335671424866,
|
| 2224 |
+
"rewards/margins": 0.5027266144752502,
|
| 2225 |
+
"rewards/rejected": -0.04789309203624725,
|
| 2226 |
+
"step": 1350
|
| 2227 |
+
},
|
| 2228 |
+
{
|
| 2229 |
+
"epoch": 3.53,
|
| 2230 |
+
"grad_norm": 56.25,
|
| 2231 |
+
"learning_rate": 2.0520853686560177e-07,
|
| 2232 |
+
"logits/chosen": -1.9281432628631592,
|
| 2233 |
+
"logits/rejected": -1.9337003231048584,
|
| 2234 |
+
"logps/chosen": -33.40667724609375,
|
| 2235 |
+
"logps/rejected": -34.605411529541016,
|
| 2236 |
+
"loss": 0.0912,
|
| 2237 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2238 |
+
"rewards/chosen": 0.38248568773269653,
|
| 2239 |
+
"rewards/margins": 0.42059844732284546,
|
| 2240 |
+
"rewards/rejected": -0.038112785667181015,
|
| 2241 |
+
"step": 1360
|
| 2242 |
+
},
|
| 2243 |
+
{
|
| 2244 |
+
"epoch": 3.56,
|
| 2245 |
+
"grad_norm": 11.375,
|
| 2246 |
+
"learning_rate": 1.833161662683672e-07,
|
| 2247 |
+
"logits/chosen": -2.017787456512451,
|
| 2248 |
+
"logits/rejected": -2.0132956504821777,
|
| 2249 |
+
"logps/chosen": -32.1794319152832,
|
| 2250 |
+
"logps/rejected": -30.92777442932129,
|
| 2251 |
+
"loss": 0.1078,
|
| 2252 |
+
"rewards/accuracies": 0.9375,
|
| 2253 |
+
"rewards/chosen": 0.4106277525424957,
|
| 2254 |
+
"rewards/margins": 0.4249129891395569,
|
| 2255 |
+
"rewards/rejected": -0.014285224489867687,
|
| 2256 |
+
"step": 1370
|
| 2257 |
+
},
|
| 2258 |
+
{
|
| 2259 |
+
"epoch": 3.58,
|
| 2260 |
+
"grad_norm": 19.875,
|
| 2261 |
+
"learning_rate": 1.626139998169246e-07,
|
| 2262 |
+
"logits/chosen": -1.9410808086395264,
|
| 2263 |
+
"logits/rejected": -1.9427919387817383,
|
| 2264 |
+
"logps/chosen": -29.26833152770996,
|
| 2265 |
+
"logps/rejected": -30.779708862304688,
|
| 2266 |
+
"loss": 0.0842,
|
| 2267 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2268 |
+
"rewards/chosen": 0.35187023878097534,
|
| 2269 |
+
"rewards/margins": 0.3838461935520172,
|
| 2270 |
+
"rewards/rejected": -0.031975992023944855,
|
| 2271 |
+
"step": 1380
|
| 2272 |
+
},
|
| 2273 |
+
{
|
| 2274 |
+
"epoch": 3.61,
|
| 2275 |
+
"grad_norm": 16.125,
|
| 2276 |
+
"learning_rate": 1.4311267331922535e-07,
|
| 2277 |
+
"logits/chosen": -1.926817536354065,
|
| 2278 |
+
"logits/rejected": -1.931318998336792,
|
| 2279 |
+
"logps/chosen": -31.545446395874023,
|
| 2280 |
+
"logps/rejected": -33.034786224365234,
|
| 2281 |
+
"loss": 0.0861,
|
| 2282 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2283 |
+
"rewards/chosen": 0.4266428053379059,
|
| 2284 |
+
"rewards/margins": 0.46869271993637085,
|
| 2285 |
+
"rewards/rejected": -0.042049963027238846,
|
| 2286 |
+
"step": 1390
|
| 2287 |
+
},
|
| 2288 |
+
{
|
| 2289 |
+
"epoch": 3.64,
|
| 2290 |
+
"grad_norm": 29.0,
|
| 2291 |
+
"learning_rate": 1.2482220564763669e-07,
|
| 2292 |
+
"logits/chosen": -1.9985454082489014,
|
| 2293 |
+
"logits/rejected": -2.00184965133667,
|
| 2294 |
+
"logps/chosen": -34.45569610595703,
|
| 2295 |
+
"logps/rejected": -33.49158477783203,
|
| 2296 |
+
"loss": 0.0815,
|
| 2297 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2298 |
+
"rewards/chosen": 0.4273880422115326,
|
| 2299 |
+
"rewards/margins": 0.4886067509651184,
|
| 2300 |
+
"rewards/rejected": -0.061218757182359695,
|
| 2301 |
+
"step": 1400
|
| 2302 |
+
},
|
| 2303 |
+
{
|
| 2304 |
+
"epoch": 3.64,
|
| 2305 |
+
"eval_logits/chosen": -2.2475900650024414,
|
| 2306 |
+
"eval_logits/rejected": -2.2427213191986084,
|
| 2307 |
+
"eval_logps/chosen": -33.8858757019043,
|
| 2308 |
+
"eval_logps/rejected": -37.38360595703125,
|
| 2309 |
+
"eval_loss": 0.35954806208610535,
|
| 2310 |
+
"eval_rewards/accuracies": 0.5224252939224243,
|
| 2311 |
+
"eval_rewards/chosen": 0.13380877673625946,
|
| 2312 |
+
"eval_rewards/margins": 0.014098312705755234,
|
| 2313 |
+
"eval_rewards/rejected": 0.11971045285463333,
|
| 2314 |
+
"eval_runtime": 145.2937,
|
| 2315 |
+
"eval_samples_per_second": 2.361,
|
| 2316 |
+
"eval_steps_per_second": 0.296,
|
| 2317 |
+
"step": 1400
|
| 2318 |
+
},
|
| 2319 |
+
{
|
| 2320 |
+
"epoch": 3.66,
|
| 2321 |
+
"grad_norm": 38.25,
|
| 2322 |
+
"learning_rate": 1.0775199359171346e-07,
|
| 2323 |
+
"logits/chosen": -2.0676581859588623,
|
| 2324 |
+
"logits/rejected": -2.060824155807495,
|
| 2325 |
+
"logps/chosen": -33.353126525878906,
|
| 2326 |
+
"logps/rejected": -34.27102279663086,
|
| 2327 |
+
"loss": 0.1028,
|
| 2328 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2329 |
+
"rewards/chosen": 0.4044764041900635,
|
| 2330 |
+
"rewards/margins": 0.42828792333602905,
|
| 2331 |
+
"rewards/rejected": -0.023811563849449158,
|
| 2332 |
+
"step": 1410
|
| 2333 |
+
},
|
| 2334 |
+
{
|
| 2335 |
+
"epoch": 3.69,
|
| 2336 |
+
"grad_norm": 46.0,
|
| 2337 |
+
"learning_rate": 9.191080703056604e-08,
|
| 2338 |
+
"logits/chosen": -1.9524263143539429,
|
| 2339 |
+
"logits/rejected": -1.9631872177124023,
|
| 2340 |
+
"logps/chosen": -33.7514533996582,
|
| 2341 |
+
"logps/rejected": -33.68230438232422,
|
| 2342 |
+
"loss": 0.1069,
|
| 2343 |
+
"rewards/accuracies": 0.9375,
|
| 2344 |
+
"rewards/chosen": 0.3580663800239563,
|
| 2345 |
+
"rewards/margins": 0.44016751646995544,
|
| 2346 |
+
"rewards/rejected": -0.08210111409425735,
|
| 2347 |
+
"step": 1420
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 3.71,
|
| 2351 |
+
"grad_norm": 33.0,
|
| 2352 |
+
"learning_rate": 7.730678442730539e-08,
|
| 2353 |
+
"logits/chosen": -2.038909912109375,
|
| 2354 |
+
"logits/rejected": -2.0507607460021973,
|
| 2355 |
+
"logps/chosen": -34.61919021606445,
|
| 2356 |
+
"logps/rejected": -33.713623046875,
|
| 2357 |
+
"loss": 0.1328,
|
| 2358 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2359 |
+
"rewards/chosen": 0.39745327830314636,
|
| 2360 |
+
"rewards/margins": 0.465212881565094,
|
| 2361 |
+
"rewards/rejected": -0.06775961816310883,
|
| 2362 |
+
"step": 1430
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"epoch": 3.74,
|
| 2366 |
+
"grad_norm": 20.125,
|
| 2367 |
+
"learning_rate": 6.394742864787806e-08,
|
| 2368 |
+
"logits/chosen": -2.0287575721740723,
|
| 2369 |
+
"logits/rejected": -2.030729055404663,
|
| 2370 |
+
"logps/chosen": -32.74304962158203,
|
| 2371 |
+
"logps/rejected": -33.90821075439453,
|
| 2372 |
+
"loss": 0.1046,
|
| 2373 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2374 |
+
"rewards/chosen": 0.4188024401664734,
|
| 2375 |
+
"rewards/margins": 0.46222352981567383,
|
| 2376 |
+
"rewards/rejected": -0.04342114180326462,
|
| 2377 |
+
"step": 1440
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 3.77,
|
| 2381 |
+
"grad_norm": 20.875,
|
| 2382 |
+
"learning_rate": 5.183960310644748e-08,
|
| 2383 |
+
"logits/chosen": -2.010601758956909,
|
| 2384 |
+
"logits/rejected": -2.0033953189849854,
|
| 2385 |
+
"logps/chosen": -35.09027862548828,
|
| 2386 |
+
"logps/rejected": -33.78215789794922,
|
| 2387 |
+
"loss": 0.0945,
|
| 2388 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2389 |
+
"rewards/chosen": 0.43024691939353943,
|
| 2390 |
+
"rewards/margins": 0.4408265948295593,
|
| 2391 |
+
"rewards/rejected": -0.010579647496342659,
|
| 2392 |
+
"step": 1450
|
| 2393 |
+
},
|
| 2394 |
+
{
|
| 2395 |
+
"epoch": 3.79,
|
| 2396 |
+
"grad_norm": 58.5,
|
| 2397 |
+
"learning_rate": 4.098952823928693e-08,
|
| 2398 |
+
"logits/chosen": -1.9762957096099854,
|
| 2399 |
+
"logits/rejected": -1.9823601245880127,
|
| 2400 |
+
"logps/chosen": -30.78021240234375,
|
| 2401 |
+
"logps/rejected": -33.74589538574219,
|
| 2402 |
+
"loss": 0.0957,
|
| 2403 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2404 |
+
"rewards/chosen": 0.38966792821884155,
|
| 2405 |
+
"rewards/margins": 0.45797547698020935,
|
| 2406 |
+
"rewards/rejected": -0.06830750405788422,
|
| 2407 |
+
"step": 1460
|
| 2408 |
+
},
|
| 2409 |
+
{
|
| 2410 |
+
"epoch": 3.82,
|
| 2411 |
+
"grad_norm": 28.375,
|
| 2412 |
+
"learning_rate": 3.1402778309014284e-08,
|
| 2413 |
+
"logits/chosen": -1.9236361980438232,
|
| 2414 |
+
"logits/rejected": -1.9213138818740845,
|
| 2415 |
+
"logps/chosen": -30.394033432006836,
|
| 2416 |
+
"logps/rejected": -30.590625762939453,
|
| 2417 |
+
"loss": 0.1267,
|
| 2418 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2419 |
+
"rewards/chosen": 0.40636783838272095,
|
| 2420 |
+
"rewards/margins": 0.4031190872192383,
|
| 2421 |
+
"rewards/rejected": 0.003248719498515129,
|
| 2422 |
+
"step": 1470
|
| 2423 |
+
},
|
| 2424 |
+
{
|
| 2425 |
+
"epoch": 3.84,
|
| 2426 |
+
"grad_norm": 18.25,
|
| 2427 |
+
"learning_rate": 2.3084278540791427e-08,
|
| 2428 |
+
"logits/chosen": -2.09725284576416,
|
| 2429 |
+
"logits/rejected": -2.091542959213257,
|
| 2430 |
+
"logps/chosen": -34.822288513183594,
|
| 2431 |
+
"logps/rejected": -32.077293395996094,
|
| 2432 |
+
"loss": 0.0885,
|
| 2433 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2434 |
+
"rewards/chosen": 0.4166763424873352,
|
| 2435 |
+
"rewards/margins": 0.4352034032344818,
|
| 2436 |
+
"rewards/rejected": -0.018527105450630188,
|
| 2437 |
+
"step": 1480
|
| 2438 |
+
},
|
| 2439 |
+
{
|
| 2440 |
+
"epoch": 3.87,
|
| 2441 |
+
"grad_norm": 16.0,
|
| 2442 |
+
"learning_rate": 1.6038302591975807e-08,
|
| 2443 |
+
"logits/chosen": -1.9885956048965454,
|
| 2444 |
+
"logits/rejected": -1.9906467199325562,
|
| 2445 |
+
"logps/chosen": -28.027481079101562,
|
| 2446 |
+
"logps/rejected": -28.051410675048828,
|
| 2447 |
+
"loss": 0.0847,
|
| 2448 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2449 |
+
"rewards/chosen": 0.40400925278663635,
|
| 2450 |
+
"rewards/margins": 0.4302259087562561,
|
| 2451 |
+
"rewards/rejected": -0.02621668018400669,
|
| 2452 |
+
"step": 1490
|
| 2453 |
+
},
|
| 2454 |
+
{
|
| 2455 |
+
"epoch": 3.9,
|
| 2456 |
+
"grad_norm": 28.375,
|
| 2457 |
+
"learning_rate": 1.0268470356514237e-08,
|
| 2458 |
+
"logits/chosen": -1.9982271194458008,
|
| 2459 |
+
"logits/rejected": -1.9925235509872437,
|
| 2460 |
+
"logps/chosen": -32.401649475097656,
|
| 2461 |
+
"logps/rejected": -32.05929946899414,
|
| 2462 |
+
"loss": 0.0757,
|
| 2463 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2464 |
+
"rewards/chosen": 0.359231173992157,
|
| 2465 |
+
"rewards/margins": 0.44054365158081055,
|
| 2466 |
+
"rewards/rejected": -0.08131246268749237,
|
| 2467 |
+
"step": 1500
|
| 2468 |
+
},
|
| 2469 |
+
{
|
| 2470 |
+
"epoch": 3.9,
|
| 2471 |
+
"eval_logits/chosen": -2.2469353675842285,
|
| 2472 |
+
"eval_logits/rejected": -2.2420578002929688,
|
| 2473 |
+
"eval_logps/chosen": -33.8865852355957,
|
| 2474 |
+
"eval_logps/rejected": -37.384151458740234,
|
| 2475 |
+
"eval_loss": 0.3542863130569458,
|
| 2476 |
+
"eval_rewards/accuracies": 0.5485880374908447,
|
| 2477 |
+
"eval_rewards/chosen": 0.1331672966480255,
|
| 2478 |
+
"eval_rewards/margins": 0.013947191648185253,
|
| 2479 |
+
"eval_rewards/rejected": 0.11922012269496918,
|
| 2480 |
+
"eval_runtime": 145.2538,
|
| 2481 |
+
"eval_samples_per_second": 2.361,
|
| 2482 |
+
"eval_steps_per_second": 0.296,
|
| 2483 |
+
"step": 1500
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 3.92,
|
| 2487 |
+
"grad_norm": 30.375,
|
| 2488 |
+
"learning_rate": 5.777746105209147e-09,
|
| 2489 |
+
"logits/chosen": -1.9242042303085327,
|
| 2490 |
+
"logits/rejected": -1.9285516738891602,
|
| 2491 |
+
"logps/chosen": -33.746910095214844,
|
| 2492 |
+
"logps/rejected": -34.46141815185547,
|
| 2493 |
+
"loss": 0.1179,
|
| 2494 |
+
"rewards/accuracies": 0.875,
|
| 2495 |
+
"rewards/chosen": 0.38167905807495117,
|
| 2496 |
+
"rewards/margins": 0.4276120662689209,
|
| 2497 |
+
"rewards/rejected": -0.04593303054571152,
|
| 2498 |
+
"step": 1510
|
| 2499 |
+
},
|
| 2500 |
+
{
|
| 2501 |
+
"epoch": 3.95,
|
| 2502 |
+
"grad_norm": 78.5,
|
| 2503 |
+
"learning_rate": 2.5684369628148352e-09,
|
| 2504 |
+
"logits/chosen": -1.9852806329727173,
|
| 2505 |
+
"logits/rejected": -1.9835399389266968,
|
| 2506 |
+
"logps/chosen": -30.321674346923828,
|
| 2507 |
+
"logps/rejected": -32.68654251098633,
|
| 2508 |
+
"loss": 0.1491,
|
| 2509 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2510 |
+
"rewards/chosen": 0.3712538182735443,
|
| 2511 |
+
"rewards/margins": 0.4583936631679535,
|
| 2512 |
+
"rewards/rejected": -0.08713988214731216,
|
| 2513 |
+
"step": 1520
|
| 2514 |
+
},
|
| 2515 |
+
{
|
| 2516 |
+
"epoch": 3.97,
|
| 2517 |
+
"grad_norm": 41.25,
|
| 2518 |
+
"learning_rate": 6.421917227455999e-10,
|
| 2519 |
+
"logits/chosen": -2.082426071166992,
|
| 2520 |
+
"logits/rejected": -2.074831008911133,
|
| 2521 |
+
"logps/chosen": -27.650592803955078,
|
| 2522 |
+
"logps/rejected": -28.735992431640625,
|
| 2523 |
+
"loss": 0.1085,
|
| 2524 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2525 |
+
"rewards/chosen": 0.3711165487766266,
|
| 2526 |
+
"rewards/margins": 0.4290649890899658,
|
| 2527 |
+
"rewards/rejected": -0.05794848874211311,
|
| 2528 |
+
"step": 1530
|
| 2529 |
+
},
|
| 2530 |
+
{
|
| 2531 |
+
"epoch": 4.0,
|
| 2532 |
+
"grad_norm": 50.25,
|
| 2533 |
+
"learning_rate": 0.0,
|
| 2534 |
+
"logits/chosen": -1.9804470539093018,
|
| 2535 |
+
"logits/rejected": -1.9702799320220947,
|
| 2536 |
+
"logps/chosen": -32.56533432006836,
|
| 2537 |
+
"logps/rejected": -35.504661560058594,
|
| 2538 |
+
"loss": 0.1078,
|
| 2539 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 2540 |
+
"rewards/chosen": 0.4507782459259033,
|
| 2541 |
+
"rewards/margins": 0.5349205732345581,
|
| 2542 |
+
"rewards/rejected": -0.0841422826051712,
|
| 2543 |
+
"step": 1540
|
| 2544 |
+
},
|
| 2545 |
+
{
|
| 2546 |
+
"epoch": 4.0,
|
| 2547 |
+
"step": 1540,
|
| 2548 |
"total_flos": 0.0,
|
| 2549 |
+
"train_loss": 0.17404492969636792,
|
| 2550 |
+
"train_runtime": 10767.1792,
|
| 2551 |
+
"train_samples_per_second": 1.144,
|
| 2552 |
+
"train_steps_per_second": 0.143
|
| 2553 |
}
|
| 2554 |
],
|
| 2555 |
"logging_steps": 10,
|
| 2556 |
+
"max_steps": 1540,
|
| 2557 |
"num_input_tokens_seen": 0,
|
| 2558 |
+
"num_train_epochs": 4,
|
| 2559 |
"save_steps": 100,
|
| 2560 |
"total_flos": 0.0,
|
| 2561 |
"train_batch_size": 4,
|