Model save
Browse files- README.md +22 -21
- adapter_model.safetensors +1 -1
- all_results.json +17 -17
- eval_results.json +13 -13
- runs/Mar01_10-22-14_gpu4-119-4/events.out.tfevents.1709249008.gpu4-119-4.1841365.0 +2 -2
- runs/Mar01_10-22-14_gpu4-119-4/events.out.tfevents.1709257307.gpu4-119-4.1841365.1 +3 -0
- train_results.json +5 -5
- trainer_state.json +830 -78
README.md
CHANGED
|
@@ -2,13 +2,9 @@
|
|
| 2 |
license: mit
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
-
- alignment-handbook
|
| 6 |
-
- generated_from_trainer
|
| 7 |
- trl
|
| 8 |
- dpo
|
| 9 |
- generated_from_trainer
|
| 10 |
-
datasets:
|
| 11 |
-
- HuggingFaceH4/ultrafeedback_binarized
|
| 12 |
base_model: microsoft/phi-2
|
| 13 |
model-index:
|
| 14 |
- name: phi-2-dpo-ultrachat-lora
|
|
@@ -20,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 20 |
|
| 21 |
# phi-2-dpo-ultrachat-lora
|
| 22 |
|
| 23 |
-
This model is a fine-tuned version of [
|
| 24 |
It achieves the following results on the evaluation set:
|
| 25 |
-
- Loss: 0.
|
| 26 |
-
- Rewards/chosen: -0.
|
| 27 |
-
- Rewards/rejected: -0.
|
| 28 |
-
- Rewards/accuracies: 0.
|
| 29 |
-
- Rewards/margins: 0.
|
| 30 |
-
- Logps/rejected: -
|
| 31 |
-
- Logps/chosen: -
|
| 32 |
-
- Logits/rejected: 0.
|
| 33 |
-
- Logits/chosen: 0.
|
| 34 |
|
| 35 |
## Model description
|
| 36 |
|
|
@@ -61,16 +57,21 @@ The following hyperparameters were used during training:
|
|
| 61 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 62 |
- lr_scheduler_type: cosine
|
| 63 |
- lr_scheduler_warmup_ratio: 0.1
|
| 64 |
-
- num_epochs:
|
| 65 |
|
| 66 |
### Training results
|
| 67 |
|
| 68 |
-
| Training Loss | Epoch | Step |
|
| 69 |
-
|
| 70 |
-
| 0.693 | 0.21 | 100 | 0.
|
| 71 |
-
| 0.6922 | 0.42 | 200 | 0.
|
| 72 |
-
| 0.6917 | 0.63 | 300 | 0.
|
| 73 |
-
| 0.6905 | 0.84 | 400 | 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
|
| 75 |
|
| 76 |
### Framework versions
|
|
|
|
| 2 |
license: mit
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
|
|
|
|
|
|
| 5 |
- trl
|
| 6 |
- dpo
|
| 7 |
- generated_from_trainer
|
|
|
|
|
|
|
| 8 |
base_model: microsoft/phi-2
|
| 9 |
model-index:
|
| 10 |
- name: phi-2-dpo-ultrachat-lora
|
|
|
|
| 16 |
|
| 17 |
# phi-2-dpo-ultrachat-lora
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2) on the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6872
|
| 22 |
+
- Rewards/chosen: -0.0312
|
| 23 |
+
- Rewards/rejected: -0.0436
|
| 24 |
+
- Rewards/accuracies: 0.3340
|
| 25 |
+
- Rewards/margins: 0.0124
|
| 26 |
+
- Logps/rejected: -98.5542
|
| 27 |
+
- Logps/chosen: -94.8435
|
| 28 |
+
- Logits/rejected: 0.7532
|
| 29 |
+
- Logits/chosen: 0.7326
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 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: 2
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
+
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
| 65 |
+
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
| 66 |
+
| 0.693 | 0.21 | 100 | 0.7998 | 0.8176 | -91.7748 | -94.2804 | 0.6931 | 0.2680 | -0.0005 | 0.0004 | -0.0008 |
|
| 67 |
+
| 0.6922 | 0.42 | 200 | 0.7941 | 0.8121 | -91.9068 | -94.5141 | 0.6924 | 0.3020 | -0.0018 | 0.0014 | -0.0032 |
|
| 68 |
+
| 0.6917 | 0.63 | 300 | 0.7870 | 0.8057 | -92.2189 | -94.9659 | 0.6917 | 0.3100 | -0.0049 | 0.0028 | -0.0077 |
|
| 69 |
+
| 0.6905 | 0.84 | 400 | 0.7827 | 0.8012 | -92.4247 | -95.2509 | 0.6913 | 0.3280 | -0.0070 | 0.0036 | -0.0105 |
|
| 70 |
+
| 0.6898 | 1.05 | 500 | 0.6900 | -0.0142 | -0.0205 | 0.3360 | 0.0064 | -96.2490 | -93.1429 | 0.7903 | 0.7711 |
|
| 71 |
+
| 0.6882 | 1.26 | 600 | 0.6887 | -0.0217 | -0.0306 | 0.3340 | 0.0089 | -97.2594 | -93.8981 | 0.7722 | 0.7527 |
|
| 72 |
+
| 0.6858 | 1.47 | 700 | 0.6879 | -0.0274 | -0.0383 | 0.3280 | 0.0108 | -98.0249 | -94.4717 | 0.7600 | 0.7395 |
|
| 73 |
+
| 0.6857 | 1.67 | 800 | 0.6874 | -0.0303 | -0.0423 | 0.3340 | 0.0120 | -98.4270 | -94.7618 | 0.7548 | 0.7341 |
|
| 74 |
+
| 0.6866 | 1.88 | 900 | 0.6872 | -0.0313 | -0.0437 | 0.3420 | 0.0124 | -98.5655 | -94.8550 | 0.7528 | 0.7321 |
|
| 75 |
|
| 76 |
|
| 77 |
### Framework versions
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 41977616
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:03a564973f08ad1b996adb14089e1d24c9d0ffd3e6da6652e793b9c19210b312
|
| 3 |
size 41977616
|
all_results.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_logits/chosen": 0.
|
| 4 |
-
"eval_logits/rejected": 0.
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_rewards/accuracies": 0.
|
| 9 |
-
"eval_rewards/chosen": -0.
|
| 10 |
-
"eval_rewards/margins": 0.
|
| 11 |
-
"eval_rewards/rejected": -0.
|
| 12 |
-
"eval_runtime":
|
| 13 |
"eval_samples": 2000,
|
| 14 |
-
"eval_samples_per_second": 7.
|
| 15 |
-
"eval_steps_per_second": 0.
|
| 16 |
-
"train_loss": 0.
|
| 17 |
-
"train_runtime":
|
| 18 |
"train_samples": 30567,
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"eval_logits/chosen": 0.7325530052185059,
|
| 4 |
+
"eval_logits/rejected": 0.7531598806381226,
|
| 5 |
+
"eval_logps/chosen": -94.8434829711914,
|
| 6 |
+
"eval_logps/rejected": -98.55415344238281,
|
| 7 |
+
"eval_loss": 0.687246561050415,
|
| 8 |
+
"eval_rewards/accuracies": 0.33399999141693115,
|
| 9 |
+
"eval_rewards/chosen": -0.03116113506257534,
|
| 10 |
+
"eval_rewards/margins": 0.012410260736942291,
|
| 11 |
+
"eval_rewards/rejected": -0.04357139766216278,
|
| 12 |
+
"eval_runtime": 272.5724,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 7.337,
|
| 15 |
+
"eval_steps_per_second": 0.459,
|
| 16 |
+
"train_loss": 0.3994182998029441,
|
| 17 |
+
"train_runtime": 8026.3391,
|
| 18 |
"train_samples": 30567,
|
| 19 |
+
"train_samples_per_second": 7.617,
|
| 20 |
+
"train_steps_per_second": 0.119
|
| 21 |
}
|
eval_results.json
CHANGED
|
@@ -1,16 +1,16 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_logits/chosen": 0.
|
| 4 |
-
"eval_logits/rejected": 0.
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_rewards/accuracies": 0.
|
| 9 |
-
"eval_rewards/chosen": -0.
|
| 10 |
-
"eval_rewards/margins": 0.
|
| 11 |
-
"eval_rewards/rejected": -0.
|
| 12 |
-
"eval_runtime":
|
| 13 |
"eval_samples": 2000,
|
| 14 |
-
"eval_samples_per_second": 7.
|
| 15 |
-
"eval_steps_per_second": 0.
|
| 16 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"eval_logits/chosen": 0.7325530052185059,
|
| 4 |
+
"eval_logits/rejected": 0.7531598806381226,
|
| 5 |
+
"eval_logps/chosen": -94.8434829711914,
|
| 6 |
+
"eval_logps/rejected": -98.55415344238281,
|
| 7 |
+
"eval_loss": 0.687246561050415,
|
| 8 |
+
"eval_rewards/accuracies": 0.33399999141693115,
|
| 9 |
+
"eval_rewards/chosen": -0.03116113506257534,
|
| 10 |
+
"eval_rewards/margins": 0.012410260736942291,
|
| 11 |
+
"eval_rewards/rejected": -0.04357139766216278,
|
| 12 |
+
"eval_runtime": 272.5724,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 7.337,
|
| 15 |
+
"eval_steps_per_second": 0.459
|
| 16 |
}
|
runs/Mar01_10-22-14_gpu4-119-4/events.out.tfevents.1709249008.gpu4-119-4.1841365.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53fa2c7156918af2e86794d36bc57146c6256b04f9f11982581c5d9d612686d6
|
| 3 |
+
size 43754
|
runs/Mar01_10-22-14_gpu4-119-4/events.out.tfevents.1709257307.gpu4-119-4.1841365.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5fe03d74aa04b2919e7ad4d520740505dc619f70188a5eca953e98fc156aaf23
|
| 3 |
+
size 828
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 30567,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"train_loss": 0.3994182998029441,
|
| 4 |
+
"train_runtime": 8026.3391,
|
| 5 |
"train_samples": 30567,
|
| 6 |
+
"train_samples_per_second": 7.617,
|
| 7 |
+
"train_steps_per_second": 0.119
|
| 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,
|
|
@@ -648,116 +648,868 @@
|
|
| 648 |
},
|
| 649 |
{
|
| 650 |
"epoch": 0.86,
|
| 651 |
-
"learning_rate":
|
| 652 |
-
"logits/chosen": 0.
|
| 653 |
-
"logits/rejected": 0.
|
| 654 |
-
"logps/chosen": -95.
|
| 655 |
-
"logps/rejected": -80.
|
| 656 |
"loss": 0.6916,
|
| 657 |
-
"rewards/accuracies": 0.
|
| 658 |
-
"rewards/chosen": -0.
|
| 659 |
-
"rewards/margins": 0.
|
| 660 |
-
"rewards/rejected": -0.
|
| 661 |
"step": 410
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.88,
|
| 665 |
-
"learning_rate":
|
| 666 |
-
"logits/chosen": 0.
|
| 667 |
-
"logits/rejected": 0.
|
| 668 |
-
"logps/chosen": -95.
|
| 669 |
-
"logps/rejected": -97.
|
| 670 |
-
"loss": 0.
|
| 671 |
-
"rewards/accuracies": 0.
|
| 672 |
-
"rewards/chosen": -0.
|
| 673 |
-
"rewards/margins": 0.
|
| 674 |
-
"rewards/rejected": -0.
|
| 675 |
"step": 420
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.9,
|
| 679 |
-
"learning_rate":
|
| 680 |
-
"logits/chosen": 0.
|
| 681 |
-
"logits/rejected": 0.
|
| 682 |
-
"logps/chosen": -83.
|
| 683 |
-
"logps/rejected": -90.
|
| 684 |
-
"loss": 0.
|
| 685 |
-
"rewards/accuracies": 0.
|
| 686 |
-
"rewards/chosen": -0.
|
| 687 |
-
"rewards/margins": 0.
|
| 688 |
-
"rewards/rejected": -0.
|
| 689 |
"step": 430
|
| 690 |
},
|
| 691 |
{
|
| 692 |
"epoch": 0.92,
|
| 693 |
-
"learning_rate":
|
| 694 |
-
"logits/chosen": 0.
|
| 695 |
-
"logits/rejected": 0.
|
| 696 |
-
"logps/chosen": -109.
|
| 697 |
-
"logps/rejected": -82.
|
| 698 |
-
"loss": 0.
|
| 699 |
-
"rewards/accuracies": 0.
|
| 700 |
-
"rewards/chosen": -0.
|
| 701 |
-
"rewards/margins": 0.
|
| 702 |
-
"rewards/rejected": -0.
|
| 703 |
"step": 440
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.94,
|
| 707 |
-
"learning_rate":
|
| 708 |
-
"logits/chosen": 0.
|
| 709 |
-
"logits/rejected": 0.
|
| 710 |
-
"logps/chosen": -81.
|
| 711 |
-
"logps/rejected": -
|
| 712 |
-
"loss": 0.
|
| 713 |
-
"rewards/accuracies": 0.
|
| 714 |
-
"rewards/chosen": -0.
|
| 715 |
-
"rewards/margins": 0.
|
| 716 |
-
"rewards/rejected": -0.
|
| 717 |
"step": 450
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.96,
|
| 721 |
-
"learning_rate":
|
| 722 |
-
"logits/chosen": 0.
|
| 723 |
-
"logits/rejected": 0.
|
| 724 |
-
"logps/chosen": -
|
| 725 |
-
"logps/rejected": -
|
| 726 |
-
"loss": 0.
|
| 727 |
-
"rewards/accuracies": 0.
|
| 728 |
-
"rewards/chosen": -0.
|
| 729 |
-
"rewards/margins": 0.
|
| 730 |
-
"rewards/rejected": -0.
|
| 731 |
"step": 460
|
| 732 |
},
|
| 733 |
{
|
| 734 |
"epoch": 0.98,
|
| 735 |
-
"learning_rate": 3.
|
| 736 |
-
"logits/chosen": 0.
|
| 737 |
-
"logits/rejected": 0.
|
| 738 |
-
"logps/chosen": -
|
| 739 |
-
"logps/rejected": -
|
| 740 |
-
"loss": 0.
|
| 741 |
-
"rewards/accuracies": 0.
|
| 742 |
-
"rewards/chosen": -0.
|
| 743 |
-
"rewards/margins": 0.
|
| 744 |
-
"rewards/rejected": -0.
|
| 745 |
"step": 470
|
| 746 |
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 750 |
"total_flos": 0.0,
|
| 751 |
-
"train_loss": 0.
|
| 752 |
-
"train_runtime":
|
| 753 |
-
"train_samples_per_second":
|
| 754 |
-
"train_steps_per_second": 0.
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
-
"max_steps":
|
| 759 |
"num_input_tokens_seen": 0,
|
| 760 |
-
"num_train_epochs":
|
| 761 |
"save_steps": 100,
|
| 762 |
"total_flos": 0.0,
|
| 763 |
"train_batch_size": 4,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.9968602825745683,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 954,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 648 |
},
|
| 649 |
{
|
| 650 |
"epoch": 0.86,
|
| 651 |
+
"learning_rate": 3.5218566107988872e-06,
|
| 652 |
+
"logits/chosen": 0.8185564279556274,
|
| 653 |
+
"logits/rejected": 0.8151994943618774,
|
| 654 |
+
"logps/chosen": -95.8418960571289,
|
| 655 |
+
"logps/rejected": -80.64591979980469,
|
| 656 |
"loss": 0.6916,
|
| 657 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 658 |
+
"rewards/chosen": -0.008152286522090435,
|
| 659 |
+
"rewards/margins": 0.003440212458372116,
|
| 660 |
+
"rewards/rejected": -0.011592499911785126,
|
| 661 |
"step": 410
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.88,
|
| 665 |
+
"learning_rate": 3.437648009023905e-06,
|
| 666 |
+
"logits/chosen": 0.7549771070480347,
|
| 667 |
+
"logits/rejected": 0.8158624768257141,
|
| 668 |
+
"logps/chosen": -95.72089385986328,
|
| 669 |
+
"logps/rejected": -97.96694946289062,
|
| 670 |
+
"loss": 0.6914,
|
| 671 |
+
"rewards/accuracies": 0.34375,
|
| 672 |
+
"rewards/chosen": -0.00923125259578228,
|
| 673 |
+
"rewards/margins": 0.0032367813400924206,
|
| 674 |
+
"rewards/rejected": -0.012468034401535988,
|
| 675 |
"step": 420
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.9,
|
| 679 |
+
"learning_rate": 3.352182461642929e-06,
|
| 680 |
+
"logits/chosen": 0.862383246421814,
|
| 681 |
+
"logits/rejected": 0.8848444819450378,
|
| 682 |
+
"logps/chosen": -83.51128387451172,
|
| 683 |
+
"logps/rejected": -90.3624038696289,
|
| 684 |
+
"loss": 0.6909,
|
| 685 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 686 |
+
"rewards/chosen": -0.0076131029054522514,
|
| 687 |
+
"rewards/margins": 0.005502562969923019,
|
| 688 |
+
"rewards/rejected": -0.013115664944052696,
|
| 689 |
"step": 430
|
| 690 |
},
|
| 691 |
{
|
| 692 |
"epoch": 0.92,
|
| 693 |
+
"learning_rate": 3.265574537815398e-06,
|
| 694 |
+
"logits/chosen": 0.7784366011619568,
|
| 695 |
+
"logits/rejected": 0.8296224474906921,
|
| 696 |
+
"logps/chosen": -109.9617919921875,
|
| 697 |
+
"logps/rejected": -82.34262084960938,
|
| 698 |
+
"loss": 0.6907,
|
| 699 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 700 |
+
"rewards/chosen": -0.006960420869290829,
|
| 701 |
+
"rewards/margins": 0.007026113569736481,
|
| 702 |
+
"rewards/rejected": -0.013986535370349884,
|
| 703 |
"step": 440
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.94,
|
| 707 |
+
"learning_rate": 3.177940338091043e-06,
|
| 708 |
+
"logits/chosen": 0.769826352596283,
|
| 709 |
+
"logits/rejected": 0.8450337648391724,
|
| 710 |
+
"logps/chosen": -81.3624496459961,
|
| 711 |
+
"logps/rejected": -72.23616027832031,
|
| 712 |
+
"loss": 0.6903,
|
| 713 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 714 |
+
"rewards/chosen": -0.00711588840931654,
|
| 715 |
+
"rewards/margins": 0.005811682902276516,
|
| 716 |
+
"rewards/rejected": -0.012927571311593056,
|
| 717 |
"step": 450
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.96,
|
| 721 |
+
"learning_rate": 3.089397338773569e-06,
|
| 722 |
+
"logits/chosen": 0.7976155877113342,
|
| 723 |
+
"logits/rejected": 0.8512013554573059,
|
| 724 |
+
"logps/chosen": -106.02879333496094,
|
| 725 |
+
"logps/rejected": -88.0615234375,
|
| 726 |
+
"loss": 0.6899,
|
| 727 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 728 |
+
"rewards/chosen": -0.007455066777765751,
|
| 729 |
+
"rewards/margins": 0.00965816993266344,
|
| 730 |
+
"rewards/rejected": -0.017113234847784042,
|
| 731 |
"step": 460
|
| 732 |
},
|
| 733 |
{
|
| 734 |
"epoch": 0.98,
|
| 735 |
+
"learning_rate": 3.0000642344401115e-06,
|
| 736 |
+
"logits/chosen": 0.8498729467391968,
|
| 737 |
+
"logits/rejected": 0.8459364771842957,
|
| 738 |
+
"logps/chosen": -116.07499694824219,
|
| 739 |
+
"logps/rejected": -111.56224060058594,
|
| 740 |
+
"loss": 0.6892,
|
| 741 |
+
"rewards/accuracies": 0.375,
|
| 742 |
+
"rewards/chosen": -0.010130222886800766,
|
| 743 |
+
"rewards/margins": 0.009535903111100197,
|
| 744 |
+
"rewards/rejected": -0.019666125997900963,
|
| 745 |
"step": 470
|
| 746 |
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
+
"learning_rate": 2.9100607788275547e-06,
|
| 750 |
+
"logits/chosen": 0.8643083572387695,
|
| 751 |
+
"logits/rejected": 0.8725606799125671,
|
| 752 |
+
"logps/chosen": -76.1094741821289,
|
| 753 |
+
"logps/rejected": -66.09565734863281,
|
| 754 |
+
"loss": 0.6903,
|
| 755 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 756 |
+
"rewards/chosen": -0.010615186765789986,
|
| 757 |
+
"rewards/margins": 0.004579311236739159,
|
| 758 |
+
"rewards/rejected": -0.015194499865174294,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.03,
|
| 763 |
+
"learning_rate": 2.8195076242990124e-06,
|
| 764 |
+
"logits/chosen": 0.8051859736442566,
|
| 765 |
+
"logits/rejected": 0.8318646550178528,
|
| 766 |
+
"logps/chosen": -93.38316345214844,
|
| 767 |
+
"logps/rejected": -100.01469421386719,
|
| 768 |
+
"loss": 0.6903,
|
| 769 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 770 |
+
"rewards/chosen": -0.014997744932770729,
|
| 771 |
+
"rewards/margins": 0.004119081888347864,
|
| 772 |
+
"rewards/rejected": -0.019116824492812157,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.05,
|
| 777 |
+
"learning_rate": 2.72852616010567e-06,
|
| 778 |
+
"logits/chosen": 0.8241451978683472,
|
| 779 |
+
"logits/rejected": 0.8089090585708618,
|
| 780 |
+
"logps/chosen": -102.48905181884766,
|
| 781 |
+
"logps/rejected": -102.73881530761719,
|
| 782 |
+
"loss": 0.6898,
|
| 783 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 784 |
+
"rewards/chosen": -0.013014930300414562,
|
| 785 |
+
"rewards/margins": 0.011042198166251183,
|
| 786 |
+
"rewards/rejected": -0.02405713126063347,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.05,
|
| 791 |
+
"eval_logits/chosen": 0.7711244225502014,
|
| 792 |
+
"eval_logits/rejected": 0.7902986407279968,
|
| 793 |
+
"eval_logps/chosen": -93.14293670654297,
|
| 794 |
+
"eval_logps/rejected": -96.24901580810547,
|
| 795 |
+
"eval_loss": 0.6899796724319458,
|
| 796 |
+
"eval_rewards/accuracies": 0.335999995470047,
|
| 797 |
+
"eval_rewards/chosen": -0.014155692420899868,
|
| 798 |
+
"eval_rewards/margins": 0.006364365108311176,
|
| 799 |
+
"eval_rewards/rejected": -0.020520057529211044,
|
| 800 |
+
"eval_runtime": 272.796,
|
| 801 |
+
"eval_samples_per_second": 7.331,
|
| 802 |
+
"eval_steps_per_second": 0.458,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 1.07,
|
| 807 |
+
"learning_rate": 2.637238349660819e-06,
|
| 808 |
+
"logits/chosen": 0.8259177207946777,
|
| 809 |
+
"logits/rejected": 0.8255017995834351,
|
| 810 |
+
"logps/chosen": -92.70307159423828,
|
| 811 |
+
"logps/rejected": -90.75215148925781,
|
| 812 |
+
"loss": 0.6893,
|
| 813 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 814 |
+
"rewards/chosen": -0.016096794977784157,
|
| 815 |
+
"rewards/margins": 0.006517867557704449,
|
| 816 |
+
"rewards/rejected": -0.02261466160416603,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.09,
|
| 821 |
+
"learning_rate": 2.5457665670441937e-06,
|
| 822 |
+
"logits/chosen": 0.8241807222366333,
|
| 823 |
+
"logits/rejected": 0.8674284219741821,
|
| 824 |
+
"logps/chosen": -103.48587799072266,
|
| 825 |
+
"logps/rejected": -95.00794219970703,
|
| 826 |
+
"loss": 0.6887,
|
| 827 |
+
"rewards/accuracies": 0.375,
|
| 828 |
+
"rewards/chosen": -0.012511089444160461,
|
| 829 |
+
"rewards/margins": 0.011780844070017338,
|
| 830 |
+
"rewards/rejected": -0.024291934445500374,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.11,
|
| 835 |
+
"learning_rate": 2.4542334329558075e-06,
|
| 836 |
+
"logits/chosen": 0.8090038299560547,
|
| 837 |
+
"logits/rejected": 0.8909789323806763,
|
| 838 |
+
"logps/chosen": -108.34500885009766,
|
| 839 |
+
"logps/rejected": -90.74520111083984,
|
| 840 |
+
"loss": 0.6897,
|
| 841 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 842 |
+
"rewards/chosen": -0.013912905938923359,
|
| 843 |
+
"rewards/margins": 0.01008325070142746,
|
| 844 |
+
"rewards/rejected": -0.023996157571673393,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.13,
|
| 849 |
+
"learning_rate": 2.3627616503391813e-06,
|
| 850 |
+
"logits/chosen": 0.7575622797012329,
|
| 851 |
+
"logits/rejected": 0.7894734144210815,
|
| 852 |
+
"logps/chosen": -103.4853515625,
|
| 853 |
+
"logps/rejected": -89.14204406738281,
|
| 854 |
+
"loss": 0.6886,
|
| 855 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 856 |
+
"rewards/chosen": -0.01601695641875267,
|
| 857 |
+
"rewards/margins": 0.009688997641205788,
|
| 858 |
+
"rewards/rejected": -0.025705954059958458,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.15,
|
| 863 |
+
"learning_rate": 2.271473839894331e-06,
|
| 864 |
+
"logits/chosen": 0.7494063377380371,
|
| 865 |
+
"logits/rejected": 0.7767230272293091,
|
| 866 |
+
"logps/chosen": -93.18793487548828,
|
| 867 |
+
"logps/rejected": -85.53865814208984,
|
| 868 |
+
"loss": 0.689,
|
| 869 |
+
"rewards/accuracies": 0.34375,
|
| 870 |
+
"rewards/chosen": -0.015046611428260803,
|
| 871 |
+
"rewards/margins": 0.009437174536287785,
|
| 872 |
+
"rewards/rejected": -0.024483786895871162,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.17,
|
| 877 |
+
"learning_rate": 2.1804923757009885e-06,
|
| 878 |
+
"logits/chosen": 0.7773897051811218,
|
| 879 |
+
"logits/rejected": 0.8226049542427063,
|
| 880 |
+
"logps/chosen": -118.14359283447266,
|
| 881 |
+
"logps/rejected": -118.2892837524414,
|
| 882 |
+
"loss": 0.6883,
|
| 883 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 884 |
+
"rewards/chosen": -0.01944814994931221,
|
| 885 |
+
"rewards/margins": 0.012502019293606281,
|
| 886 |
+
"rewards/rejected": -0.03195016831159592,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.19,
|
| 891 |
+
"learning_rate": 2.089939221172446e-06,
|
| 892 |
+
"logits/chosen": 0.7639147639274597,
|
| 893 |
+
"logits/rejected": 0.8038470149040222,
|
| 894 |
+
"logps/chosen": -96.55006408691406,
|
| 895 |
+
"logps/rejected": -94.61703491210938,
|
| 896 |
+
"loss": 0.6893,
|
| 897 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 898 |
+
"rewards/chosen": -0.01976608671247959,
|
| 899 |
+
"rewards/margins": 0.01228870265185833,
|
| 900 |
+
"rewards/rejected": -0.03205478936433792,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.21,
|
| 905 |
+
"learning_rate": 1.9999357655598894e-06,
|
| 906 |
+
"logits/chosen": 0.799372673034668,
|
| 907 |
+
"logits/rejected": 0.7758508920669556,
|
| 908 |
+
"logps/chosen": -107.0300521850586,
|
| 909 |
+
"logps/rejected": -101.98541259765625,
|
| 910 |
+
"loss": 0.6898,
|
| 911 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 912 |
+
"rewards/chosen": -0.023156706243753433,
|
| 913 |
+
"rewards/margins": 0.0033106685150414705,
|
| 914 |
+
"rewards/rejected": -0.026467373594641685,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.23,
|
| 919 |
+
"learning_rate": 1.9106026612264316e-06,
|
| 920 |
+
"logits/chosen": 0.8342978358268738,
|
| 921 |
+
"logits/rejected": 0.818748950958252,
|
| 922 |
+
"logps/chosen": -82.849609375,
|
| 923 |
+
"logps/rejected": -82.92151641845703,
|
| 924 |
+
"loss": 0.6875,
|
| 925 |
+
"rewards/accuracies": 0.34375,
|
| 926 |
+
"rewards/chosen": -0.021987412124872208,
|
| 927 |
+
"rewards/margins": 0.00925590842962265,
|
| 928 |
+
"rewards/rejected": -0.03124331869184971,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.26,
|
| 933 |
+
"learning_rate": 1.8220596619089576e-06,
|
| 934 |
+
"logits/chosen": 0.7792515158653259,
|
| 935 |
+
"logits/rejected": 0.7631284594535828,
|
| 936 |
+
"logps/chosen": -108.453369140625,
|
| 937 |
+
"logps/rejected": -109.84110260009766,
|
| 938 |
+
"loss": 0.6882,
|
| 939 |
+
"rewards/accuracies": 0.3812499940395355,
|
| 940 |
+
"rewards/chosen": -0.024591121822595596,
|
| 941 |
+
"rewards/margins": 0.010857349261641502,
|
| 942 |
+
"rewards/rejected": -0.03544846922159195,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.26,
|
| 947 |
+
"eval_logits/chosen": 0.7526758909225464,
|
| 948 |
+
"eval_logits/rejected": 0.7722000479698181,
|
| 949 |
+
"eval_logps/chosen": -93.8980941772461,
|
| 950 |
+
"eval_logps/rejected": -97.25941467285156,
|
| 951 |
+
"eval_loss": 0.6886565089225769,
|
| 952 |
+
"eval_rewards/accuracies": 0.33399999141693115,
|
| 953 |
+
"eval_rewards/chosen": -0.02170729637145996,
|
| 954 |
+
"eval_rewards/margins": 0.008916829712688923,
|
| 955 |
+
"eval_rewards/rejected": -0.03062412701547146,
|
| 956 |
+
"eval_runtime": 272.575,
|
| 957 |
+
"eval_samples_per_second": 7.337,
|
| 958 |
+
"eval_steps_per_second": 0.459,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 1.28,
|
| 963 |
+
"learning_rate": 1.7344254621846018e-06,
|
| 964 |
+
"logits/chosen": 0.7705439329147339,
|
| 965 |
+
"logits/rejected": 0.7898679375648499,
|
| 966 |
+
"logps/chosen": -114.97818756103516,
|
| 967 |
+
"logps/rejected": -105.07425689697266,
|
| 968 |
+
"loss": 0.6871,
|
| 969 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 970 |
+
"rewards/chosen": -0.020355116575956345,
|
| 971 |
+
"rewards/margins": 0.01100640743970871,
|
| 972 |
+
"rewards/rejected": -0.031361524015665054,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.3,
|
| 977 |
+
"learning_rate": 1.647817538357072e-06,
|
| 978 |
+
"logits/chosen": 0.8015000224113464,
|
| 979 |
+
"logits/rejected": 0.8592731356620789,
|
| 980 |
+
"logps/chosen": -105.63825988769531,
|
| 981 |
+
"logps/rejected": -101.30821228027344,
|
| 982 |
+
"loss": 0.6877,
|
| 983 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 984 |
+
"rewards/chosen": -0.025122905150055885,
|
| 985 |
+
"rewards/margins": 0.012796345166862011,
|
| 986 |
+
"rewards/rejected": -0.03791924566030502,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.32,
|
| 991 |
+
"learning_rate": 1.5623519909760953e-06,
|
| 992 |
+
"logits/chosen": 0.72571861743927,
|
| 993 |
+
"logits/rejected": 0.7672609090805054,
|
| 994 |
+
"logps/chosen": -95.39134216308594,
|
| 995 |
+
"logps/rejected": -91.43064880371094,
|
| 996 |
+
"loss": 0.6875,
|
| 997 |
+
"rewards/accuracies": 0.3125,
|
| 998 |
+
"rewards/chosen": -0.023522889241576195,
|
| 999 |
+
"rewards/margins": 0.008042030036449432,
|
| 1000 |
+
"rewards/rejected": -0.031564921140670776,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.34,
|
| 1005 |
+
"learning_rate": 1.4781433892011132e-06,
|
| 1006 |
+
"logits/chosen": 0.7983066439628601,
|
| 1007 |
+
"logits/rejected": 0.828285813331604,
|
| 1008 |
+
"logps/chosen": -91.58604431152344,
|
| 1009 |
+
"logps/rejected": -98.34529876708984,
|
| 1010 |
+
"loss": 0.6877,
|
| 1011 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 1012 |
+
"rewards/chosen": -0.02216607704758644,
|
| 1013 |
+
"rewards/margins": 0.009402362629771233,
|
| 1014 |
+
"rewards/rejected": -0.03156844526529312,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.36,
|
| 1019 |
+
"learning_rate": 1.3953046172178413e-06,
|
| 1020 |
+
"logits/chosen": 0.8052563667297363,
|
| 1021 |
+
"logits/rejected": 0.861487090587616,
|
| 1022 |
+
"logps/chosen": -102.22875213623047,
|
| 1023 |
+
"logps/rejected": -115.90311431884766,
|
| 1024 |
+
"loss": 0.688,
|
| 1025 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 1026 |
+
"rewards/chosen": -0.02779082953929901,
|
| 1027 |
+
"rewards/margins": 0.011276346631348133,
|
| 1028 |
+
"rewards/rejected": -0.03906718268990517,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 1.38,
|
| 1033 |
+
"learning_rate": 1.3139467229135999e-06,
|
| 1034 |
+
"logits/chosen": 0.7473502159118652,
|
| 1035 |
+
"logits/rejected": 0.7619670629501343,
|
| 1036 |
+
"logps/chosen": -81.72496032714844,
|
| 1037 |
+
"logps/rejected": -88.32991027832031,
|
| 1038 |
+
"loss": 0.6867,
|
| 1039 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 1040 |
+
"rewards/chosen": -0.023514125496149063,
|
| 1041 |
+
"rewards/margins": 0.01081976480782032,
|
| 1042 |
+
"rewards/rejected": -0.034333888441324234,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 1.4,
|
| 1047 |
+
"learning_rate": 1.2341787690142436e-06,
|
| 1048 |
+
"logits/chosen": 0.677151083946228,
|
| 1049 |
+
"logits/rejected": 0.7038652300834656,
|
| 1050 |
+
"logps/chosen": -116.34004211425781,
|
| 1051 |
+
"logps/rejected": -98.69480895996094,
|
| 1052 |
+
"loss": 0.6883,
|
| 1053 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 1054 |
+
"rewards/chosen": -0.024621013551950455,
|
| 1055 |
+
"rewards/margins": 0.016095632687211037,
|
| 1056 |
+
"rewards/rejected": -0.040716640651226044,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 1.42,
|
| 1061 |
+
"learning_rate": 1.1561076868822756e-06,
|
| 1062 |
+
"logits/chosen": 0.7057438492774963,
|
| 1063 |
+
"logits/rejected": 0.734573483467102,
|
| 1064 |
+
"logps/chosen": -77.70423889160156,
|
| 1065 |
+
"logps/rejected": -83.44449615478516,
|
| 1066 |
+
"loss": 0.6875,
|
| 1067 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 1068 |
+
"rewards/chosen": -0.029654154554009438,
|
| 1069 |
+
"rewards/margins": 0.00593325262889266,
|
| 1070 |
+
"rewards/rejected": -0.035587407648563385,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 1.44,
|
| 1075 |
+
"learning_rate": 1.079838133172111e-06,
|
| 1076 |
+
"logits/chosen": 0.7637497186660767,
|
| 1077 |
+
"logits/rejected": 0.8190714716911316,
|
| 1078 |
+
"logps/chosen": -95.7881088256836,
|
| 1079 |
+
"logps/rejected": -82.04200744628906,
|
| 1080 |
+
"loss": 0.6882,
|
| 1081 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 1082 |
+
"rewards/chosen": -0.026545196771621704,
|
| 1083 |
+
"rewards/margins": 0.014608954079449177,
|
| 1084 |
+
"rewards/rejected": -0.041154149919748306,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 1.47,
|
| 1089 |
+
"learning_rate": 1.0054723495346484e-06,
|
| 1090 |
+
"logits/chosen": 0.8055804371833801,
|
| 1091 |
+
"logits/rejected": 0.8074037432670593,
|
| 1092 |
+
"logps/chosen": -106.52473449707031,
|
| 1093 |
+
"logps/rejected": -99.84661865234375,
|
| 1094 |
+
"loss": 0.6858,
|
| 1095 |
+
"rewards/accuracies": 0.375,
|
| 1096 |
+
"rewards/chosen": -0.03077777661383152,
|
| 1097 |
+
"rewards/margins": 0.016430744901299477,
|
| 1098 |
+
"rewards/rejected": -0.0472085177898407,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 1.47,
|
| 1103 |
+
"eval_logits/chosen": 0.7395281791687012,
|
| 1104 |
+
"eval_logits/rejected": 0.7599726319313049,
|
| 1105 |
+
"eval_logps/chosen": -94.4716567993164,
|
| 1106 |
+
"eval_logps/rejected": -98.02493286132812,
|
| 1107 |
+
"eval_loss": 0.6878523826599121,
|
| 1108 |
+
"eval_rewards/accuracies": 0.328000009059906,
|
| 1109 |
+
"eval_rewards/chosen": -0.02744293212890625,
|
| 1110 |
+
"eval_rewards/margins": 0.010836233384907246,
|
| 1111 |
+
"eval_rewards/rejected": -0.03827916085720062,
|
| 1112 |
+
"eval_runtime": 272.6941,
|
| 1113 |
+
"eval_samples_per_second": 7.334,
|
| 1114 |
+
"eval_steps_per_second": 0.458,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 1.49,
|
| 1119 |
+
"learning_rate": 9.331100255592437e-07,
|
| 1120 |
+
"logits/chosen": 0.7606201767921448,
|
| 1121 |
+
"logits/rejected": 0.8386918306350708,
|
| 1122 |
+
"logps/chosen": -89.20939636230469,
|
| 1123 |
+
"logps/rejected": -82.81990814208984,
|
| 1124 |
+
"loss": 0.6861,
|
| 1125 |
+
"rewards/accuracies": 0.3062500059604645,
|
| 1126 |
+
"rewards/chosen": -0.022489020600914955,
|
| 1127 |
+
"rewards/margins": 0.012643699534237385,
|
| 1128 |
+
"rewards/rejected": -0.03513271361589432,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 1.51,
|
| 1133 |
+
"learning_rate": 8.628481651367876e-07,
|
| 1134 |
+
"logits/chosen": 0.771535336971283,
|
| 1135 |
+
"logits/rejected": 0.8548757433891296,
|
| 1136 |
+
"logps/chosen": -113.86335754394531,
|
| 1137 |
+
"logps/rejected": -101.86293029785156,
|
| 1138 |
+
"loss": 0.6856,
|
| 1139 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 1140 |
+
"rewards/chosen": -0.02375701069831848,
|
| 1141 |
+
"rewards/margins": 0.01785343512892723,
|
| 1142 |
+
"rewards/rejected": -0.041610442101955414,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 1.53,
|
| 1147 |
+
"learning_rate": 7.947809564230446e-07,
|
| 1148 |
+
"logits/chosen": 0.7982994914054871,
|
| 1149 |
+
"logits/rejected": 0.8568431735038757,
|
| 1150 |
+
"logps/chosen": -98.13008880615234,
|
| 1151 |
+
"logps/rejected": -97.87364196777344,
|
| 1152 |
+
"loss": 0.6874,
|
| 1153 |
+
"rewards/accuracies": 0.375,
|
| 1154 |
+
"rewards/chosen": -0.03274186700582504,
|
| 1155 |
+
"rewards/margins": 0.010785898193717003,
|
| 1156 |
+
"rewards/rejected": -0.0435277596116066,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 1.55,
|
| 1161 |
+
"learning_rate": 7.289996455765749e-07,
|
| 1162 |
+
"logits/chosen": 0.7194357514381409,
|
| 1163 |
+
"logits/rejected": 0.7669156789779663,
|
| 1164 |
+
"logps/chosen": -88.61851501464844,
|
| 1165 |
+
"logps/rejected": -82.62269592285156,
|
| 1166 |
+
"loss": 0.6889,
|
| 1167 |
+
"rewards/accuracies": 0.2562499940395355,
|
| 1168 |
+
"rewards/chosen": -0.02295432984828949,
|
| 1169 |
+
"rewards/margins": 0.008415495045483112,
|
| 1170 |
+
"rewards/rejected": -0.03136982396245003,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 1.57,
|
| 1175 |
+
"learning_rate": 6.655924144404907e-07,
|
| 1176 |
+
"logits/chosen": 0.7268679141998291,
|
| 1177 |
+
"logits/rejected": 0.7419520616531372,
|
| 1178 |
+
"logps/chosen": -90.82257843017578,
|
| 1179 |
+
"logps/rejected": -87.84976196289062,
|
| 1180 |
+
"loss": 0.687,
|
| 1181 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 1182 |
+
"rewards/chosen": -0.024226779118180275,
|
| 1183 |
+
"rewards/margins": 0.014533626846969128,
|
| 1184 |
+
"rewards/rejected": -0.03876040503382683,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 1.59,
|
| 1189 |
+
"learning_rate": 6.046442623320145e-07,
|
| 1190 |
+
"logits/chosen": 0.775734543800354,
|
| 1191 |
+
"logits/rejected": 0.8094171285629272,
|
| 1192 |
+
"logps/chosen": -111.5390853881836,
|
| 1193 |
+
"logps/rejected": -107.2262954711914,
|
| 1194 |
+
"loss": 0.6868,
|
| 1195 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 1196 |
+
"rewards/chosen": -0.035627130419015884,
|
| 1197 |
+
"rewards/margins": 0.011030396446585655,
|
| 1198 |
+
"rewards/rejected": -0.04665752500295639,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 1.61,
|
| 1203 |
+
"learning_rate": 5.462368920983249e-07,
|
| 1204 |
+
"logits/chosen": 0.7696752548217773,
|
| 1205 |
+
"logits/rejected": 0.7860345840454102,
|
| 1206 |
+
"logps/chosen": -86.94595336914062,
|
| 1207 |
+
"logps/rejected": -74.9552230834961,
|
| 1208 |
+
"loss": 0.6867,
|
| 1209 |
+
"rewards/accuracies": 0.3125,
|
| 1210 |
+
"rewards/chosen": -0.02532133087515831,
|
| 1211 |
+
"rewards/margins": 0.010699031874537468,
|
| 1212 |
+
"rewards/rejected": -0.03602036461234093,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 1.63,
|
| 1217 |
+
"learning_rate": 4.904486005914027e-07,
|
| 1218 |
+
"logits/chosen": 0.7500615119934082,
|
| 1219 |
+
"logits/rejected": 0.7661712765693665,
|
| 1220 |
+
"logps/chosen": -119.89558410644531,
|
| 1221 |
+
"logps/rejected": -111.26094055175781,
|
| 1222 |
+
"loss": 0.6851,
|
| 1223 |
+
"rewards/accuracies": 0.3812499940395355,
|
| 1224 |
+
"rewards/chosen": -0.02393215522170067,
|
| 1225 |
+
"rewards/margins": 0.01878645084798336,
|
| 1226 |
+
"rewards/rejected": -0.04271860793232918,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 1.65,
|
| 1231 |
+
"learning_rate": 4.373541737087264e-07,
|
| 1232 |
+
"logits/chosen": 0.7226775288581848,
|
| 1233 |
+
"logits/rejected": 0.7507297992706299,
|
| 1234 |
+
"logps/chosen": -103.21110534667969,
|
| 1235 |
+
"logps/rejected": -99.74224853515625,
|
| 1236 |
+
"loss": 0.6876,
|
| 1237 |
+
"rewards/accuracies": 0.29374998807907104,
|
| 1238 |
+
"rewards/chosen": -0.028694171458482742,
|
| 1239 |
+
"rewards/margins": 0.013696588575839996,
|
| 1240 |
+
"rewards/rejected": -0.04239075630903244,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 1.67,
|
| 1245 |
+
"learning_rate": 3.8702478614051353e-07,
|
| 1246 |
+
"logits/chosen": 0.7573332786560059,
|
| 1247 |
+
"logits/rejected": 0.7167325019836426,
|
| 1248 |
+
"logps/chosen": -101.85111236572266,
|
| 1249 |
+
"logps/rejected": -95.10759735107422,
|
| 1250 |
+
"loss": 0.6857,
|
| 1251 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 1252 |
+
"rewards/chosen": -0.02707928977906704,
|
| 1253 |
+
"rewards/margins": 0.022238872945308685,
|
| 1254 |
+
"rewards/rejected": -0.049318164587020874,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 1.67,
|
| 1259 |
+
"eval_logits/chosen": 0.7340908646583557,
|
| 1260 |
+
"eval_logits/rejected": 0.7547872066497803,
|
| 1261 |
+
"eval_logps/chosen": -94.76182556152344,
|
| 1262 |
+
"eval_logps/rejected": -98.42695617675781,
|
| 1263 |
+
"eval_loss": 0.687368631362915,
|
| 1264 |
+
"eval_rewards/accuracies": 0.33399999141693115,
|
| 1265 |
+
"eval_rewards/chosen": -0.030344627797603607,
|
| 1266 |
+
"eval_rewards/margins": 0.011954776011407375,
|
| 1267 |
+
"eval_rewards/rejected": -0.04229940101504326,
|
| 1268 |
+
"eval_runtime": 272.7101,
|
| 1269 |
+
"eval_samples_per_second": 7.334,
|
| 1270 |
+
"eval_steps_per_second": 0.458,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 1.7,
|
| 1275 |
+
"learning_rate": 3.3952790595787986e-07,
|
| 1276 |
+
"logits/chosen": 0.8084260821342468,
|
| 1277 |
+
"logits/rejected": 0.8206745982170105,
|
| 1278 |
+
"logps/chosen": -118.43504333496094,
|
| 1279 |
+
"logps/rejected": -105.46468353271484,
|
| 1280 |
+
"loss": 0.6876,
|
| 1281 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 1282 |
+
"rewards/chosen": -0.03346724063158035,
|
| 1283 |
+
"rewards/margins": 0.010504155419766903,
|
| 1284 |
+
"rewards/rejected": -0.04397139698266983,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 1.72,
|
| 1289 |
+
"learning_rate": 2.9492720416985004e-07,
|
| 1290 |
+
"logits/chosen": 0.6800391674041748,
|
| 1291 |
+
"logits/rejected": 0.6965945363044739,
|
| 1292 |
+
"logps/chosen": -96.83241271972656,
|
| 1293 |
+
"logps/rejected": -101.45925903320312,
|
| 1294 |
+
"loss": 0.6869,
|
| 1295 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 1296 |
+
"rewards/chosen": -0.027126986533403397,
|
| 1297 |
+
"rewards/margins": 0.01599222794175148,
|
| 1298 |
+
"rewards/rejected": -0.04311921447515488,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 1.74,
|
| 1303 |
+
"learning_rate": 2.5328246937043526e-07,
|
| 1304 |
+
"logits/chosen": 0.7581857442855835,
|
| 1305 |
+
"logits/rejected": 0.7956913113594055,
|
| 1306 |
+
"logps/chosen": -98.04803466796875,
|
| 1307 |
+
"logps/rejected": -102.8603515625,
|
| 1308 |
+
"loss": 0.6858,
|
| 1309 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 1310 |
+
"rewards/chosen": -0.034231387078762054,
|
| 1311 |
+
"rewards/margins": 0.015285378322005272,
|
| 1312 |
+
"rewards/rejected": -0.049516770988702774,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 1.76,
|
| 1317 |
+
"learning_rate": 2.1464952759020857e-07,
|
| 1318 |
+
"logits/chosen": 0.6740394830703735,
|
| 1319 |
+
"logits/rejected": 0.7218228578567505,
|
| 1320 |
+
"logps/chosen": -96.54594421386719,
|
| 1321 |
+
"logps/rejected": -99.02603912353516,
|
| 1322 |
+
"loss": 0.6878,
|
| 1323 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 1324 |
+
"rewards/chosen": -0.029307430610060692,
|
| 1325 |
+
"rewards/margins": 0.010790064930915833,
|
| 1326 |
+
"rewards/rejected": -0.040097493678331375,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 1.78,
|
| 1331 |
+
"learning_rate": 1.790801674598186e-07,
|
| 1332 |
+
"logits/chosen": 0.7388015985488892,
|
| 1333 |
+
"logits/rejected": 0.7736684083938599,
|
| 1334 |
+
"logps/chosen": -126.44425201416016,
|
| 1335 |
+
"logps/rejected": -102.01686096191406,
|
| 1336 |
+
"loss": 0.686,
|
| 1337 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 1338 |
+
"rewards/chosen": -0.030984923243522644,
|
| 1339 |
+
"rewards/margins": 0.01829494535923004,
|
| 1340 |
+
"rewards/rejected": -0.049279868602752686,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 1.8,
|
| 1345 |
+
"learning_rate": 1.4662207078575685e-07,
|
| 1346 |
+
"logits/chosen": 0.7661498785018921,
|
| 1347 |
+
"logits/rejected": 0.7773095965385437,
|
| 1348 |
+
"logps/chosen": -92.13005065917969,
|
| 1349 |
+
"logps/rejected": -91.37489318847656,
|
| 1350 |
+
"loss": 0.6844,
|
| 1351 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 1352 |
+
"rewards/chosen": -0.028016680851578712,
|
| 1353 |
+
"rewards/margins": 0.021260341629385948,
|
| 1354 |
+
"rewards/rejected": -0.04927702248096466,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 1.82,
|
| 1359 |
+
"learning_rate": 1.1731874863145143e-07,
|
| 1360 |
+
"logits/chosen": 0.7595205903053284,
|
| 1361 |
+
"logits/rejected": 0.8001340627670288,
|
| 1362 |
+
"logps/chosen": -94.4384765625,
|
| 1363 |
+
"logps/rejected": -92.37102508544922,
|
| 1364 |
+
"loss": 0.687,
|
| 1365 |
+
"rewards/accuracies": 0.3812499940395355,
|
| 1366 |
+
"rewards/chosen": -0.026580199599266052,
|
| 1367 |
+
"rewards/margins": 0.01631156913936138,
|
| 1368 |
+
"rewards/rejected": -0.042891766875982285,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 1.84,
|
| 1373 |
+
"learning_rate": 9.120948298936422e-08,
|
| 1374 |
+
"logits/chosen": 0.7684468030929565,
|
| 1375 |
+
"logits/rejected": 0.8293699026107788,
|
| 1376 |
+
"logps/chosen": -89.8713150024414,
|
| 1377 |
+
"logps/rejected": -87.5567855834961,
|
| 1378 |
+
"loss": 0.6844,
|
| 1379 |
+
"rewards/accuracies": 0.34375,
|
| 1380 |
+
"rewards/chosen": -0.026115071028470993,
|
| 1381 |
+
"rewards/margins": 0.01522884052246809,
|
| 1382 |
+
"rewards/rejected": -0.04134391248226166,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 1.86,
|
| 1387 |
+
"learning_rate": 6.832927412229017e-08,
|
| 1388 |
+
"logits/chosen": 0.7296298742294312,
|
| 1389 |
+
"logits/rejected": 0.769476056098938,
|
| 1390 |
+
"logps/chosen": -93.165283203125,
|
| 1391 |
+
"logps/rejected": -96.9769287109375,
|
| 1392 |
+
"loss": 0.6851,
|
| 1393 |
+
"rewards/accuracies": 0.34375,
|
| 1394 |
+
"rewards/chosen": -0.029177119955420494,
|
| 1395 |
+
"rewards/margins": 0.016743745654821396,
|
| 1396 |
+
"rewards/rejected": -0.04592086747288704,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 1.88,
|
| 1401 |
+
"learning_rate": 4.870879364444109e-08,
|
| 1402 |
+
"logits/chosen": 0.7372065782546997,
|
| 1403 |
+
"logits/rejected": 0.7822612524032593,
|
| 1404 |
+
"logps/chosen": -89.78585052490234,
|
| 1405 |
+
"logps/rejected": -89.9983901977539,
|
| 1406 |
+
"loss": 0.6866,
|
| 1407 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 1408 |
+
"rewards/chosen": -0.028756320476531982,
|
| 1409 |
+
"rewards/margins": 0.017704127356410027,
|
| 1410 |
+
"rewards/rejected": -0.04646044969558716,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 1.88,
|
| 1415 |
+
"eval_logits/chosen": 0.732122004032135,
|
| 1416 |
+
"eval_logits/rejected": 0.7528373599052429,
|
| 1417 |
+
"eval_logps/chosen": -94.85501861572266,
|
| 1418 |
+
"eval_logps/rejected": -98.56551361083984,
|
| 1419 |
+
"eval_loss": 0.6872289776802063,
|
| 1420 |
+
"eval_rewards/accuracies": 0.34200000762939453,
|
| 1421 |
+
"eval_rewards/chosen": -0.03127633407711983,
|
| 1422 |
+
"eval_rewards/margins": 0.012408819980919361,
|
| 1423 |
+
"eval_rewards/rejected": -0.043685153126716614,
|
| 1424 |
+
"eval_runtime": 272.6053,
|
| 1425 |
+
"eval_samples_per_second": 7.337,
|
| 1426 |
+
"eval_steps_per_second": 0.459,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 1.9,
|
| 1431 |
+
"learning_rate": 3.237434340521789e-08,
|
| 1432 |
+
"logits/chosen": 0.710226833820343,
|
| 1433 |
+
"logits/rejected": 0.7852008938789368,
|
| 1434 |
+
"logps/chosen": -103.32108306884766,
|
| 1435 |
+
"logps/rejected": -105.41300964355469,
|
| 1436 |
+
"loss": 0.6851,
|
| 1437 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 1438 |
+
"rewards/chosen": -0.030542368069291115,
|
| 1439 |
+
"rewards/margins": 0.015573601238429546,
|
| 1440 |
+
"rewards/rejected": -0.04611596092581749,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 1.93,
|
| 1445 |
+
"learning_rate": 1.93478202307823e-08,
|
| 1446 |
+
"logits/chosen": 0.7026702165603638,
|
| 1447 |
+
"logits/rejected": 0.7310872673988342,
|
| 1448 |
+
"logps/chosen": -98.82904815673828,
|
| 1449 |
+
"logps/rejected": -96.06184387207031,
|
| 1450 |
+
"loss": 0.6866,
|
| 1451 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 1452 |
+
"rewards/chosen": -0.02804415300488472,
|
| 1453 |
+
"rewards/margins": 0.011173558421432972,
|
| 1454 |
+
"rewards/rejected": -0.03921770304441452,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 1.95,
|
| 1459 |
+
"learning_rate": 9.646686570697062e-09,
|
| 1460 |
+
"logits/chosen": 0.7164516448974609,
|
| 1461 |
+
"logits/rejected": 0.8162258863449097,
|
| 1462 |
+
"logps/chosen": -104.56522369384766,
|
| 1463 |
+
"logps/rejected": -101.3372802734375,
|
| 1464 |
+
"loss": 0.6871,
|
| 1465 |
+
"rewards/accuracies": 0.3812499940395355,
|
| 1466 |
+
"rewards/chosen": -0.024641428142786026,
|
| 1467 |
+
"rewards/margins": 0.01806877739727497,
|
| 1468 |
+
"rewards/rejected": -0.042710207402706146,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 1.97,
|
| 1473 |
+
"learning_rate": 3.283947088983663e-09,
|
| 1474 |
+
"logits/chosen": 0.7736842632293701,
|
| 1475 |
+
"logits/rejected": 0.7780998945236206,
|
| 1476 |
+
"logps/chosen": -100.5300064086914,
|
| 1477 |
+
"logps/rejected": -111.0368881225586,
|
| 1478 |
+
"loss": 0.6875,
|
| 1479 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 1480 |
+
"rewards/chosen": -0.0371762290596962,
|
| 1481 |
+
"rewards/margins": 0.015147706493735313,
|
| 1482 |
+
"rewards/rejected": -0.05232393741607666,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 1.99,
|
| 1487 |
+
"learning_rate": 2.681312309735229e-10,
|
| 1488 |
+
"logits/chosen": 0.8036483526229858,
|
| 1489 |
+
"logits/rejected": 0.8175506591796875,
|
| 1490 |
+
"logps/chosen": -94.41046905517578,
|
| 1491 |
+
"logps/rejected": -104.76774597167969,
|
| 1492 |
+
"loss": 0.6875,
|
| 1493 |
+
"rewards/accuracies": 0.3062500059604645,
|
| 1494 |
+
"rewards/chosen": -0.03001193329691887,
|
| 1495 |
+
"rewards/margins": 0.011342789977788925,
|
| 1496 |
+
"rewards/rejected": -0.04135472699999809,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.0,
|
| 1501 |
+
"step": 954,
|
| 1502 |
"total_flos": 0.0,
|
| 1503 |
+
"train_loss": 0.3994182998029441,
|
| 1504 |
+
"train_runtime": 8026.3391,
|
| 1505 |
+
"train_samples_per_second": 7.617,
|
| 1506 |
+
"train_steps_per_second": 0.119
|
| 1507 |
}
|
| 1508 |
],
|
| 1509 |
"logging_steps": 10,
|
| 1510 |
+
"max_steps": 954,
|
| 1511 |
"num_input_tokens_seen": 0,
|
| 1512 |
+
"num_train_epochs": 2,
|
| 1513 |
"save_steps": 100,
|
| 1514 |
"total_flos": 0.0,
|
| 1515 |
"train_batch_size": 4,
|