Model save
Browse files- README.md +90 -0
- all_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +2562 -0
README.md
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
library_name: peft
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-mypo3_sim-qlora-lr5e6-beta0.30
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/1014579852qq-tsinghua-university/huggingface/runs/wnnhas0b)
|
| 18 |
+
# zephyr-7b-mypo3_sim-qlora-lr5e6-beta0.30
|
| 19 |
+
|
| 20 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on an unknown dataset.
|
| 21 |
+
It achieves the following results on the evaluation set:
|
| 22 |
+
- Loss: 1.3541
|
| 23 |
+
- Rewards/chosen: -0.0637
|
| 24 |
+
- Rewards/rejected: -0.3677
|
| 25 |
+
- Rewards/accuracies: 0.7025
|
| 26 |
+
- Rewards/margins: 0.3039
|
| 27 |
+
- Logps/rejected: -2.2834
|
| 28 |
+
- Logps/chosen: -1.1380
|
| 29 |
+
- Logits/rejected: -1.9704
|
| 30 |
+
- Logits/chosen: -2.0513
|
| 31 |
+
|
| 32 |
+
## Model description
|
| 33 |
+
|
| 34 |
+
More information needed
|
| 35 |
+
|
| 36 |
+
## Intended uses & limitations
|
| 37 |
+
|
| 38 |
+
More information needed
|
| 39 |
+
|
| 40 |
+
## Training and evaluation data
|
| 41 |
+
|
| 42 |
+
More information needed
|
| 43 |
+
|
| 44 |
+
## Training procedure
|
| 45 |
+
|
| 46 |
+
### Training hyperparameters
|
| 47 |
+
|
| 48 |
+
The following hyperparameters were used during training:
|
| 49 |
+
- learning_rate: 5e-06
|
| 50 |
+
- train_batch_size: 2
|
| 51 |
+
- eval_batch_size: 4
|
| 52 |
+
- seed: 42
|
| 53 |
+
- distributed_type: multi-GPU
|
| 54 |
+
- num_devices: 5
|
| 55 |
+
- gradient_accumulation_steps: 4
|
| 56 |
+
- total_train_batch_size: 40
|
| 57 |
+
- total_eval_batch_size: 20
|
| 58 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 59 |
+
- lr_scheduler_type: cosine
|
| 60 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 61 |
+
- num_epochs: 1
|
| 62 |
+
|
| 63 |
+
### Training results
|
| 64 |
+
|
| 65 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 66 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 67 |
+
| 1.3799 | 0.0654 | 100 | 1.3804 | -0.0062 | -0.0408 | 0.6600 | 0.0346 | -1.1940 | -0.9462 | -2.1974 | -2.2810 |
|
| 68 |
+
| 1.3728 | 0.1308 | 200 | 1.3734 | -0.0308 | -0.1119 | 0.6900 | 0.0811 | -1.4310 | -1.0283 | -2.2618 | -2.3330 |
|
| 69 |
+
| 1.3605 | 0.1963 | 300 | 1.3670 | -0.0656 | -0.2070 | 0.7200 | 0.1414 | -1.7478 | -1.1442 | -2.1971 | -2.2674 |
|
| 70 |
+
| 1.3607 | 0.2617 | 400 | 1.3637 | -0.0644 | -0.2551 | 0.6975 | 0.1908 | -1.9084 | -1.1401 | -2.2602 | -2.3277 |
|
| 71 |
+
| 1.3642 | 0.3271 | 500 | 1.3625 | -0.0744 | -0.3109 | 0.6875 | 0.2366 | -2.0943 | -1.1734 | -2.1841 | -2.2534 |
|
| 72 |
+
| 1.3489 | 0.3925 | 600 | 1.3649 | -0.1095 | -0.4197 | 0.6850 | 0.3101 | -2.4568 | -1.2906 | -2.0263 | -2.1039 |
|
| 73 |
+
| 1.3735 | 0.4580 | 700 | 1.3653 | -0.1046 | -0.4143 | 0.7000 | 0.3097 | -2.4389 | -1.2743 | -1.9237 | -2.0155 |
|
| 74 |
+
| 1.3606 | 0.5234 | 800 | 1.3592 | -0.0745 | -0.3701 | 0.6950 | 0.2956 | -2.2915 | -1.1739 | -1.9493 | -2.0356 |
|
| 75 |
+
| 1.3462 | 0.5888 | 900 | 1.3568 | -0.0854 | -0.3668 | 0.7050 | 0.2815 | -2.2807 | -1.2100 | -1.9785 | -2.0609 |
|
| 76 |
+
| 1.3527 | 0.6542 | 1000 | 1.3548 | -0.0626 | -0.3514 | 0.7050 | 0.2888 | -2.2291 | -1.1342 | -1.9978 | -2.0771 |
|
| 77 |
+
| 1.3483 | 0.7197 | 1100 | 1.3558 | -0.0665 | -0.3741 | 0.7025 | 0.3076 | -2.3048 | -1.1471 | -1.9802 | -2.0598 |
|
| 78 |
+
| 1.3558 | 0.7851 | 1200 | 1.3542 | -0.0628 | -0.3646 | 0.7050 | 0.3018 | -2.2733 | -1.1348 | -1.9719 | -2.0522 |
|
| 79 |
+
| 1.3515 | 0.8505 | 1300 | 1.3543 | -0.0644 | -0.3702 | 0.7050 | 0.3058 | -2.2918 | -1.1402 | -1.9694 | -2.0505 |
|
| 80 |
+
| 1.3572 | 0.9159 | 1400 | 1.3540 | -0.0639 | -0.3674 | 0.7075 | 0.3035 | -2.2825 | -1.1385 | -1.9716 | -2.0522 |
|
| 81 |
+
| 1.3527 | 0.9814 | 1500 | 1.3541 | -0.0637 | -0.3677 | 0.7025 | 0.3039 | -2.2834 | -1.1380 | -1.9704 | -2.0513 |
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
### Framework versions
|
| 85 |
+
|
| 86 |
+
- PEFT 0.10.0
|
| 87 |
+
- Transformers 4.43.1
|
| 88 |
+
- Pytorch 2.1.2+cu121
|
| 89 |
+
- Datasets 2.18.0
|
| 90 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9996728819103696,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 1.3600662709530735,
|
| 5 |
+
"train_runtime": 13966.5548,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 4.377,
|
| 8 |
+
"train_steps_per_second": 0.109
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9996728819103696,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 1.3600662709530735,
|
| 5 |
+
"train_runtime": 13966.5548,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 4.377,
|
| 8 |
+
"train_steps_per_second": 0.109
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2562 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9996728819103696,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 1528,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.006542361792607131,
|
| 13 |
+
"grad_norm": 0.2294921875,
|
| 14 |
+
"learning_rate": 3.267973856209151e-07,
|
| 15 |
+
"logits/chosen": -2.5530099868774414,
|
| 16 |
+
"logits/rejected": -2.454012632369995,
|
| 17 |
+
"logps/chosen": -0.937676727771759,
|
| 18 |
+
"logps/rejected": -1.109251856803894,
|
| 19 |
+
"loss": 1.3863,
|
| 20 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 21 |
+
"rewards/chosen": -9.286254498874769e-05,
|
| 22 |
+
"rewards/margins": -7.532918971264735e-05,
|
| 23 |
+
"rewards/rejected": -1.7533380741951987e-05,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.013084723585214262,
|
| 28 |
+
"grad_norm": 0.1435546875,
|
| 29 |
+
"learning_rate": 6.535947712418302e-07,
|
| 30 |
+
"logits/chosen": -2.480818510055542,
|
| 31 |
+
"logits/rejected": -2.3557686805725098,
|
| 32 |
+
"logps/chosen": -0.9598627090454102,
|
| 33 |
+
"logps/rejected": -0.9752475619316101,
|
| 34 |
+
"loss": 1.3863,
|
| 35 |
+
"rewards/accuracies": 0.5,
|
| 36 |
+
"rewards/chosen": -7.735938197583891e-06,
|
| 37 |
+
"rewards/margins": 0.00011255677964072675,
|
| 38 |
+
"rewards/rejected": -0.00012029272329527885,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.019627085377821395,
|
| 43 |
+
"grad_norm": 0.1630859375,
|
| 44 |
+
"learning_rate": 9.80392156862745e-07,
|
| 45 |
+
"logits/chosen": -2.4215729236602783,
|
| 46 |
+
"logits/rejected": -2.3662164211273193,
|
| 47 |
+
"logps/chosen": -0.9298206567764282,
|
| 48 |
+
"logps/rejected": -1.1405647993087769,
|
| 49 |
+
"loss": 1.3862,
|
| 50 |
+
"rewards/accuracies": 0.550000011920929,
|
| 51 |
+
"rewards/chosen": -0.00022919184993952513,
|
| 52 |
+
"rewards/margins": 0.0003895942063536495,
|
| 53 |
+
"rewards/rejected": -0.0006187860853970051,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.026169447170428524,
|
| 58 |
+
"grad_norm": 0.2197265625,
|
| 59 |
+
"learning_rate": 1.3071895424836604e-06,
|
| 60 |
+
"logits/chosen": -2.3722825050354004,
|
| 61 |
+
"logits/rejected": -2.3141565322875977,
|
| 62 |
+
"logps/chosen": -1.0184338092803955,
|
| 63 |
+
"logps/rejected": -1.0292493104934692,
|
| 64 |
+
"loss": 1.386,
|
| 65 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 66 |
+
"rewards/chosen": 0.00016174458141904324,
|
| 67 |
+
"rewards/margins": 0.00046144178486429155,
|
| 68 |
+
"rewards/rejected": -0.00029969721799716353,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.03271180896303565,
|
| 73 |
+
"grad_norm": 0.1962890625,
|
| 74 |
+
"learning_rate": 1.6339869281045753e-06,
|
| 75 |
+
"logits/chosen": -2.4260201454162598,
|
| 76 |
+
"logits/rejected": -2.3685925006866455,
|
| 77 |
+
"logps/chosen": -0.840053915977478,
|
| 78 |
+
"logps/rejected": -1.0239614248275757,
|
| 79 |
+
"loss": 1.3853,
|
| 80 |
+
"rewards/accuracies": 0.762499988079071,
|
| 81 |
+
"rewards/chosen": 0.0067680226638913155,
|
| 82 |
+
"rewards/margins": 0.005519701633602381,
|
| 83 |
+
"rewards/rejected": 0.0012483217287808657,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.03925417075564279,
|
| 88 |
+
"grad_norm": 0.251953125,
|
| 89 |
+
"learning_rate": 1.96078431372549e-06,
|
| 90 |
+
"logits/chosen": -2.5593574047088623,
|
| 91 |
+
"logits/rejected": -2.252335786819458,
|
| 92 |
+
"logps/chosen": -0.8980849981307983,
|
| 93 |
+
"logps/rejected": -1.0699880123138428,
|
| 94 |
+
"loss": 1.385,
|
| 95 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 96 |
+
"rewards/chosen": 0.0039029642939567566,
|
| 97 |
+
"rewards/margins": 0.0045288316905498505,
|
| 98 |
+
"rewards/rejected": -0.0006258675130084157,
|
| 99 |
+
"step": 60
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.04579653254824992,
|
| 103 |
+
"grad_norm": 0.271484375,
|
| 104 |
+
"learning_rate": 2.2875816993464053e-06,
|
| 105 |
+
"logits/chosen": -2.476616382598877,
|
| 106 |
+
"logits/rejected": -2.4485809803009033,
|
| 107 |
+
"logps/chosen": -0.987118124961853,
|
| 108 |
+
"logps/rejected": -1.2191505432128906,
|
| 109 |
+
"loss": 1.3832,
|
| 110 |
+
"rewards/accuracies": 0.637499988079071,
|
| 111 |
+
"rewards/chosen": -0.01106469426304102,
|
| 112 |
+
"rewards/margins": 0.01529990416020155,
|
| 113 |
+
"rewards/rejected": -0.02636459842324257,
|
| 114 |
+
"step": 70
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.05233889434085705,
|
| 118 |
+
"grad_norm": 0.30078125,
|
| 119 |
+
"learning_rate": 2.6143790849673208e-06,
|
| 120 |
+
"logits/chosen": -2.353388786315918,
|
| 121 |
+
"logits/rejected": -2.254671335220337,
|
| 122 |
+
"logps/chosen": -1.0187981128692627,
|
| 123 |
+
"logps/rejected": -1.1903177499771118,
|
| 124 |
+
"loss": 1.3827,
|
| 125 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 126 |
+
"rewards/chosen": -0.03571772575378418,
|
| 127 |
+
"rewards/margins": 0.013263550586998463,
|
| 128 |
+
"rewards/rejected": -0.04898127168416977,
|
| 129 |
+
"step": 80
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.058881256133464184,
|
| 133 |
+
"grad_norm": 0.33984375,
|
| 134 |
+
"learning_rate": 2.9411764705882355e-06,
|
| 135 |
+
"logits/chosen": -2.3827168941497803,
|
| 136 |
+
"logits/rejected": -2.307100772857666,
|
| 137 |
+
"logps/chosen": -0.9496015310287476,
|
| 138 |
+
"logps/rejected": -1.1855210065841675,
|
| 139 |
+
"loss": 1.3815,
|
| 140 |
+
"rewards/accuracies": 0.699999988079071,
|
| 141 |
+
"rewards/chosen": -0.018529271706938744,
|
| 142 |
+
"rewards/margins": 0.020624618977308273,
|
| 143 |
+
"rewards/rejected": -0.03915388882160187,
|
| 144 |
+
"step": 90
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.0654236179260713,
|
| 148 |
+
"grad_norm": 0.2333984375,
|
| 149 |
+
"learning_rate": 3.2679738562091506e-06,
|
| 150 |
+
"logits/chosen": -2.4862723350524902,
|
| 151 |
+
"logits/rejected": -2.4109530448913574,
|
| 152 |
+
"logps/chosen": -0.9562190175056458,
|
| 153 |
+
"logps/rejected": -1.266717553138733,
|
| 154 |
+
"loss": 1.3799,
|
| 155 |
+
"rewards/accuracies": 0.5625,
|
| 156 |
+
"rewards/chosen": -0.022339094430208206,
|
| 157 |
+
"rewards/margins": 0.035685621201992035,
|
| 158 |
+
"rewards/rejected": -0.05802471563220024,
|
| 159 |
+
"step": 100
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.0654236179260713,
|
| 163 |
+
"eval_logits/chosen": -2.2809672355651855,
|
| 164 |
+
"eval_logits/rejected": -2.1974074840545654,
|
| 165 |
+
"eval_logps/chosen": -0.9461590647697449,
|
| 166 |
+
"eval_logps/rejected": -1.194018840789795,
|
| 167 |
+
"eval_loss": 1.3803561925888062,
|
| 168 |
+
"eval_rewards/accuracies": 0.6600000262260437,
|
| 169 |
+
"eval_rewards/chosen": -0.006188389845192432,
|
| 170 |
+
"eval_rewards/margins": 0.034646403044462204,
|
| 171 |
+
"eval_rewards/rejected": -0.04083478823304176,
|
| 172 |
+
"eval_runtime": 193.2216,
|
| 173 |
+
"eval_samples_per_second": 10.351,
|
| 174 |
+
"eval_steps_per_second": 0.518,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.07196597971867845,
|
| 179 |
+
"grad_norm": 0.80859375,
|
| 180 |
+
"learning_rate": 3.5947712418300657e-06,
|
| 181 |
+
"logits/chosen": -2.346219301223755,
|
| 182 |
+
"logits/rejected": -2.3201358318328857,
|
| 183 |
+
"logps/chosen": -1.0299032926559448,
|
| 184 |
+
"logps/rejected": -1.1942877769470215,
|
| 185 |
+
"loss": 1.3801,
|
| 186 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 187 |
+
"rewards/chosen": -0.01987696811556816,
|
| 188 |
+
"rewards/margins": 0.039442867040634155,
|
| 189 |
+
"rewards/rejected": -0.05931984260678291,
|
| 190 |
+
"step": 110
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.07850834151128558,
|
| 194 |
+
"grad_norm": 0.427734375,
|
| 195 |
+
"learning_rate": 3.92156862745098e-06,
|
| 196 |
+
"logits/chosen": -2.382317543029785,
|
| 197 |
+
"logits/rejected": -2.2658298015594482,
|
| 198 |
+
"logps/chosen": -0.9744423627853394,
|
| 199 |
+
"logps/rejected": -1.2910969257354736,
|
| 200 |
+
"loss": 1.3775,
|
| 201 |
+
"rewards/accuracies": 0.737500011920929,
|
| 202 |
+
"rewards/chosen": -0.018291741609573364,
|
| 203 |
+
"rewards/margins": 0.041886262595653534,
|
| 204 |
+
"rewards/rejected": -0.0601780004799366,
|
| 205 |
+
"step": 120
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.08505070330389271,
|
| 209 |
+
"grad_norm": 1.59375,
|
| 210 |
+
"learning_rate": 4.2483660130718954e-06,
|
| 211 |
+
"logits/chosen": -2.4718565940856934,
|
| 212 |
+
"logits/rejected": -2.3606743812561035,
|
| 213 |
+
"logps/chosen": -1.1133302450180054,
|
| 214 |
+
"logps/rejected": -1.388488531112671,
|
| 215 |
+
"loss": 1.3854,
|
| 216 |
+
"rewards/accuracies": 0.612500011920929,
|
| 217 |
+
"rewards/chosen": -0.06918147951364517,
|
| 218 |
+
"rewards/margins": 0.031170833855867386,
|
| 219 |
+
"rewards/rejected": -0.10035230964422226,
|
| 220 |
+
"step": 130
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.09159306509649984,
|
| 224 |
+
"grad_norm": 0.5234375,
|
| 225 |
+
"learning_rate": 4.5751633986928105e-06,
|
| 226 |
+
"logits/chosen": -2.500530242919922,
|
| 227 |
+
"logits/rejected": -2.386270523071289,
|
| 228 |
+
"logps/chosen": -0.9971791505813599,
|
| 229 |
+
"logps/rejected": -1.2755845785140991,
|
| 230 |
+
"loss": 1.375,
|
| 231 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 232 |
+
"rewards/chosen": -0.0014015674823895097,
|
| 233 |
+
"rewards/margins": 0.06071232631802559,
|
| 234 |
+
"rewards/rejected": -0.06211389973759651,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.09813542688910697,
|
| 239 |
+
"grad_norm": 0.41015625,
|
| 240 |
+
"learning_rate": 4.901960784313726e-06,
|
| 241 |
+
"logits/chosen": -2.470759391784668,
|
| 242 |
+
"logits/rejected": -2.4136905670166016,
|
| 243 |
+
"logps/chosen": -1.0106420516967773,
|
| 244 |
+
"logps/rejected": -1.3554531335830688,
|
| 245 |
+
"loss": 1.3771,
|
| 246 |
+
"rewards/accuracies": 0.762499988079071,
|
| 247 |
+
"rewards/chosen": -0.013763192109763622,
|
| 248 |
+
"rewards/margins": 0.06293424218893051,
|
| 249 |
+
"rewards/rejected": -0.076697438955307,
|
| 250 |
+
"step": 150
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.1046777886817141,
|
| 254 |
+
"grad_norm": 0.6953125,
|
| 255 |
+
"learning_rate": 4.999680264259825e-06,
|
| 256 |
+
"logits/chosen": -2.5543665885925293,
|
| 257 |
+
"logits/rejected": -2.3195226192474365,
|
| 258 |
+
"logps/chosen": -0.8638038635253906,
|
| 259 |
+
"logps/rejected": -1.1180086135864258,
|
| 260 |
+
"loss": 1.3734,
|
| 261 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 262 |
+
"rewards/chosen": 9.241541556548327e-05,
|
| 263 |
+
"rewards/margins": 0.05282307416200638,
|
| 264 |
+
"rewards/rejected": -0.052730657160282135,
|
| 265 |
+
"step": 160
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.11122015047432122,
|
| 269 |
+
"grad_norm": 0.6171875,
|
| 270 |
+
"learning_rate": 4.998114408534616e-06,
|
| 271 |
+
"logits/chosen": -2.4441475868225098,
|
| 272 |
+
"logits/rejected": -2.248377561569214,
|
| 273 |
+
"logps/chosen": -0.9274452328681946,
|
| 274 |
+
"logps/rejected": -1.4115386009216309,
|
| 275 |
+
"loss": 1.3736,
|
| 276 |
+
"rewards/accuracies": 0.6875,
|
| 277 |
+
"rewards/chosen": -0.019874438643455505,
|
| 278 |
+
"rewards/margins": 0.09404312819242477,
|
| 279 |
+
"rewards/rejected": -0.11391757428646088,
|
| 280 |
+
"step": 170
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.11776251226692837,
|
| 284 |
+
"grad_norm": 0.53515625,
|
| 285 |
+
"learning_rate": 4.995244522215781e-06,
|
| 286 |
+
"logits/chosen": -2.4685444831848145,
|
| 287 |
+
"logits/rejected": -2.441394329071045,
|
| 288 |
+
"logps/chosen": -1.172699213027954,
|
| 289 |
+
"logps/rejected": -1.4395049810409546,
|
| 290 |
+
"loss": 1.373,
|
| 291 |
+
"rewards/accuracies": 0.612500011920929,
|
| 292 |
+
"rewards/chosen": -0.07093976438045502,
|
| 293 |
+
"rewards/margins": 0.05027998611330986,
|
| 294 |
+
"rewards/rejected": -0.12121973931789398,
|
| 295 |
+
"step": 180
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.1243048740595355,
|
| 299 |
+
"grad_norm": 1.25,
|
| 300 |
+
"learning_rate": 4.9910721034010655e-06,
|
| 301 |
+
"logits/chosen": -2.487239360809326,
|
| 302 |
+
"logits/rejected": -2.382844924926758,
|
| 303 |
+
"logps/chosen": -0.9922649264335632,
|
| 304 |
+
"logps/rejected": -1.4326521158218384,
|
| 305 |
+
"loss": 1.3774,
|
| 306 |
+
"rewards/accuracies": 0.675000011920929,
|
| 307 |
+
"rewards/chosen": -0.02712084725499153,
|
| 308 |
+
"rewards/margins": 0.08723781257867813,
|
| 309 |
+
"rewards/rejected": -0.11435864865779877,
|
| 310 |
+
"step": 190
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.1308472358521426,
|
| 314 |
+
"grad_norm": 0.333984375,
|
| 315 |
+
"learning_rate": 4.985599330117931e-06,
|
| 316 |
+
"logits/chosen": -2.43881893157959,
|
| 317 |
+
"logits/rejected": -2.389829635620117,
|
| 318 |
+
"logps/chosen": -0.9726101160049438,
|
| 319 |
+
"logps/rejected": -1.330091953277588,
|
| 320 |
+
"loss": 1.3728,
|
| 321 |
+
"rewards/accuracies": 0.699999988079071,
|
| 322 |
+
"rewards/chosen": -0.026006106287240982,
|
| 323 |
+
"rewards/margins": 0.07863597571849823,
|
| 324 |
+
"rewards/rejected": -0.10464207828044891,
|
| 325 |
+
"step": 200
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.1308472358521426,
|
| 329 |
+
"eval_logits/chosen": -2.332975387573242,
|
| 330 |
+
"eval_logits/rejected": -2.261765241622925,
|
| 331 |
+
"eval_logps/chosen": -1.0283117294311523,
|
| 332 |
+
"eval_logps/rejected": -1.4310046434402466,
|
| 333 |
+
"eval_loss": 1.373443603515625,
|
| 334 |
+
"eval_rewards/accuracies": 0.6899999976158142,
|
| 335 |
+
"eval_rewards/chosen": -0.030834216624498367,
|
| 336 |
+
"eval_rewards/margins": 0.08109636604785919,
|
| 337 |
+
"eval_rewards/rejected": -0.11193057149648666,
|
| 338 |
+
"eval_runtime": 192.5903,
|
| 339 |
+
"eval_samples_per_second": 10.385,
|
| 340 |
+
"eval_steps_per_second": 0.519,
|
| 341 |
+
"step": 200
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"epoch": 0.13738959764474976,
|
| 345 |
+
"grad_norm": 0.33984375,
|
| 346 |
+
"learning_rate": 4.978829059186611e-06,
|
| 347 |
+
"logits/chosen": -2.504112482070923,
|
| 348 |
+
"logits/rejected": -2.4700379371643066,
|
| 349 |
+
"logps/chosen": -1.1215603351593018,
|
| 350 |
+
"logps/rejected": -1.4604188203811646,
|
| 351 |
+
"loss": 1.3733,
|
| 352 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 353 |
+
"rewards/chosen": -0.042933590710163116,
|
| 354 |
+
"rewards/margins": 0.06642533093690872,
|
| 355 |
+
"rewards/rejected": -0.10935892164707184,
|
| 356 |
+
"step": 210
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.1439319594373569,
|
| 360 |
+
"grad_norm": 0.6484375,
|
| 361 |
+
"learning_rate": 4.97076482472884e-06,
|
| 362 |
+
"logits/chosen": -2.3071188926696777,
|
| 363 |
+
"logits/rejected": -2.3290657997131348,
|
| 364 |
+
"logps/chosen": -1.1499958038330078,
|
| 365 |
+
"logps/rejected": -1.5844924449920654,
|
| 366 |
+
"loss": 1.3679,
|
| 367 |
+
"rewards/accuracies": 0.6875,
|
| 368 |
+
"rewards/chosen": -0.054123081266880035,
|
| 369 |
+
"rewards/margins": 0.1018981784582138,
|
| 370 |
+
"rewards/rejected": -0.15602126717567444,
|
| 371 |
+
"step": 220
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.15047432122996401,
|
| 375 |
+
"grad_norm": 0.44140625,
|
| 376 |
+
"learning_rate": 4.961410836323014e-06,
|
| 377 |
+
"logits/chosen": -2.4832520484924316,
|
| 378 |
+
"logits/rejected": -2.376649856567383,
|
| 379 |
+
"logps/chosen": -1.0944504737854004,
|
| 380 |
+
"logps/rejected": -1.3434072732925415,
|
| 381 |
+
"loss": 1.3733,
|
| 382 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 383 |
+
"rewards/chosen": -0.04292039945721626,
|
| 384 |
+
"rewards/margins": 0.057192735373973846,
|
| 385 |
+
"rewards/rejected": -0.10011313855648041,
|
| 386 |
+
"step": 230
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.15701668302257116,
|
| 390 |
+
"grad_norm": 0.453125,
|
| 391 |
+
"learning_rate": 4.950771976806769e-06,
|
| 392 |
+
"logits/chosen": -2.599358081817627,
|
| 393 |
+
"logits/rejected": -2.4423699378967285,
|
| 394 |
+
"logps/chosen": -0.8888974189758301,
|
| 395 |
+
"logps/rejected": -1.405906081199646,
|
| 396 |
+
"loss": 1.3718,
|
| 397 |
+
"rewards/accuracies": 0.762499988079071,
|
| 398 |
+
"rewards/chosen": -0.01539047621190548,
|
| 399 |
+
"rewards/margins": 0.09842763096094131,
|
| 400 |
+
"rewards/rejected": -0.11381809413433075,
|
| 401 |
+
"step": 240
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.16355904481517827,
|
| 405 |
+
"grad_norm": 0.58984375,
|
| 406 |
+
"learning_rate": 4.938853799728112e-06,
|
| 407 |
+
"logits/chosen": -2.4943268299102783,
|
| 408 |
+
"logits/rejected": -2.328460216522217,
|
| 409 |
+
"logps/chosen": -1.0880366563796997,
|
| 410 |
+
"logps/rejected": -1.6129534244537354,
|
| 411 |
+
"loss": 1.3694,
|
| 412 |
+
"rewards/accuracies": 0.699999988079071,
|
| 413 |
+
"rewards/chosen": -0.044886864721775055,
|
| 414 |
+
"rewards/margins": 0.12210796028375626,
|
| 415 |
+
"rewards/rejected": -0.1669948399066925,
|
| 416 |
+
"step": 250
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.17010140660778542,
|
| 420 |
+
"grad_norm": 0.578125,
|
| 421 |
+
"learning_rate": 4.925662526446431e-06,
|
| 422 |
+
"logits/chosen": -2.386685848236084,
|
| 423 |
+
"logits/rejected": -2.312121868133545,
|
| 424 |
+
"logps/chosen": -1.2352960109710693,
|
| 425 |
+
"logps/rejected": -1.6638290882110596,
|
| 426 |
+
"loss": 1.3658,
|
| 427 |
+
"rewards/accuracies": 0.699999988079071,
|
| 428 |
+
"rewards/chosen": -0.07152961194515228,
|
| 429 |
+
"rewards/margins": 0.11174267530441284,
|
| 430 |
+
"rewards/rejected": -0.18327227234840393,
|
| 431 |
+
"step": 260
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.17664376840039253,
|
| 435 |
+
"grad_norm": 0.78515625,
|
| 436 |
+
"learning_rate": 4.911205042884912e-06,
|
| 437 |
+
"logits/chosen": -2.4438509941101074,
|
| 438 |
+
"logits/rejected": -2.317202091217041,
|
| 439 |
+
"logps/chosen": -1.4406167268753052,
|
| 440 |
+
"logps/rejected": -1.5704759359359741,
|
| 441 |
+
"loss": 1.3685,
|
| 442 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 443 |
+
"rewards/chosen": -0.13397307693958282,
|
| 444 |
+
"rewards/margins": 0.051210492849349976,
|
| 445 |
+
"rewards/rejected": -0.1851835697889328,
|
| 446 |
+
"step": 270
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.18318613019299967,
|
| 450 |
+
"grad_norm": 0.3828125,
|
| 451 |
+
"learning_rate": 4.895488895936047e-06,
|
| 452 |
+
"logits/chosen": -2.3865578174591064,
|
| 453 |
+
"logits/rejected": -2.3630199432373047,
|
| 454 |
+
"logps/chosen": -1.1934704780578613,
|
| 455 |
+
"logps/rejected": -1.8741092681884766,
|
| 456 |
+
"loss": 1.362,
|
| 457 |
+
"rewards/accuracies": 0.737500011920929,
|
| 458 |
+
"rewards/chosen": -0.09065061807632446,
|
| 459 |
+
"rewards/margins": 0.14752301573753357,
|
| 460 |
+
"rewards/rejected": -0.23817363381385803,
|
| 461 |
+
"step": 280
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.18972849198560682,
|
| 465 |
+
"grad_norm": 0.4609375,
|
| 466 |
+
"learning_rate": 4.8785222895221075e-06,
|
| 467 |
+
"logits/chosen": -2.611616611480713,
|
| 468 |
+
"logits/rejected": -2.496882438659668,
|
| 469 |
+
"logps/chosen": -1.1736793518066406,
|
| 470 |
+
"logps/rejected": -1.4841904640197754,
|
| 471 |
+
"loss": 1.3704,
|
| 472 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 473 |
+
"rewards/chosen": -0.09048116952180862,
|
| 474 |
+
"rewards/margins": 0.07914741337299347,
|
| 475 |
+
"rewards/rejected": -0.1696285754442215,
|
| 476 |
+
"step": 290
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.19627085377821393,
|
| 480 |
+
"grad_norm": 0.4609375,
|
| 481 |
+
"learning_rate": 4.860314080312651e-06,
|
| 482 |
+
"logits/chosen": -2.4393398761749268,
|
| 483 |
+
"logits/rejected": -2.359809398651123,
|
| 484 |
+
"logps/chosen": -1.144270896911621,
|
| 485 |
+
"logps/rejected": -1.7192535400390625,
|
| 486 |
+
"loss": 1.3605,
|
| 487 |
+
"rewards/accuracies": 0.737500011920929,
|
| 488 |
+
"rewards/chosen": -0.07278771698474884,
|
| 489 |
+
"rewards/margins": 0.1271217167377472,
|
| 490 |
+
"rewards/rejected": -0.19990943372249603,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.19627085377821393,
|
| 495 |
+
"eval_logits/chosen": -2.26737117767334,
|
| 496 |
+
"eval_logits/rejected": -2.197129011154175,
|
| 497 |
+
"eval_logps/chosen": -1.1442248821258545,
|
| 498 |
+
"eval_logps/rejected": -1.7477951049804688,
|
| 499 |
+
"eval_loss": 1.3670034408569336,
|
| 500 |
+
"eval_rewards/accuracies": 0.7200000286102295,
|
| 501 |
+
"eval_rewards/chosen": -0.06560814380645752,
|
| 502 |
+
"eval_rewards/margins": 0.1413595825433731,
|
| 503 |
+
"eval_rewards/rejected": -0.20696774125099182,
|
| 504 |
+
"eval_runtime": 192.5244,
|
| 505 |
+
"eval_samples_per_second": 10.388,
|
| 506 |
+
"eval_steps_per_second": 0.519,
|
| 507 |
+
"step": 300
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.20281321557082108,
|
| 511 |
+
"grad_norm": 0.92578125,
|
| 512 |
+
"learning_rate": 4.840873773101287e-06,
|
| 513 |
+
"logits/chosen": -2.416212558746338,
|
| 514 |
+
"logits/rejected": -2.349287748336792,
|
| 515 |
+
"logps/chosen": -1.2875757217407227,
|
| 516 |
+
"logps/rejected": -1.9493656158447266,
|
| 517 |
+
"loss": 1.3632,
|
| 518 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 519 |
+
"rewards/chosen": -0.09226052463054657,
|
| 520 |
+
"rewards/margins": 0.16571097075939178,
|
| 521 |
+
"rewards/rejected": -0.25797149538993835,
|
| 522 |
+
"step": 310
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.2093555773634282,
|
| 526 |
+
"grad_norm": 0.5546875,
|
| 527 |
+
"learning_rate": 4.820211515844116e-06,
|
| 528 |
+
"logits/chosen": -2.4312744140625,
|
| 529 |
+
"logits/rejected": -2.428727865219116,
|
| 530 |
+
"logps/chosen": -1.1892143487930298,
|
| 531 |
+
"logps/rejected": -1.7102506160736084,
|
| 532 |
+
"loss": 1.3675,
|
| 533 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 534 |
+
"rewards/chosen": -0.08930884301662445,
|
| 535 |
+
"rewards/margins": 0.11253657191991806,
|
| 536 |
+
"rewards/rejected": -0.2018454372882843,
|
| 537 |
+
"step": 320
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.21589793915603533,
|
| 541 |
+
"grad_norm": 1.09375,
|
| 542 |
+
"learning_rate": 4.798338094362439e-06,
|
| 543 |
+
"logits/chosen": -2.4368550777435303,
|
| 544 |
+
"logits/rejected": -2.468419075012207,
|
| 545 |
+
"logps/chosen": -1.2478764057159424,
|
| 546 |
+
"logps/rejected": -1.4515265226364136,
|
| 547 |
+
"loss": 1.3713,
|
| 548 |
+
"rewards/accuracies": 0.75,
|
| 549 |
+
"rewards/chosen": -0.09131525456905365,
|
| 550 |
+
"rewards/margins": 0.06188271567225456,
|
| 551 |
+
"rewards/rejected": -0.1531979739665985,
|
| 552 |
+
"step": 330
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.22244030094864245,
|
| 556 |
+
"grad_norm": 2.1875,
|
| 557 |
+
"learning_rate": 4.775264926712489e-06,
|
| 558 |
+
"logits/chosen": -2.4024834632873535,
|
| 559 |
+
"logits/rejected": -2.3669588565826416,
|
| 560 |
+
"logps/chosen": -1.008465051651001,
|
| 561 |
+
"logps/rejected": -1.6137611865997314,
|
| 562 |
+
"loss": 1.367,
|
| 563 |
+
"rewards/accuracies": 0.699999988079071,
|
| 564 |
+
"rewards/chosen": -0.05111456662416458,
|
| 565 |
+
"rewards/margins": 0.1272425502538681,
|
| 566 |
+
"rewards/rejected": -0.17835713922977448,
|
| 567 |
+
"step": 340
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.2289826627412496,
|
| 571 |
+
"grad_norm": 0.59375,
|
| 572 |
+
"learning_rate": 4.751004057225147e-06,
|
| 573 |
+
"logits/chosen": -2.4048571586608887,
|
| 574 |
+
"logits/rejected": -2.3667616844177246,
|
| 575 |
+
"logps/chosen": -1.0595513582229614,
|
| 576 |
+
"logps/rejected": -1.6971559524536133,
|
| 577 |
+
"loss": 1.3641,
|
| 578 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 579 |
+
"rewards/chosen": -0.04516686871647835,
|
| 580 |
+
"rewards/margins": 0.1346011608839035,
|
| 581 |
+
"rewards/rejected": -0.17976802587509155,
|
| 582 |
+
"step": 350
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.23552502453385674,
|
| 586 |
+
"grad_norm": 1.3203125,
|
| 587 |
+
"learning_rate": 4.725568150218719e-06,
|
| 588 |
+
"logits/chosen": -2.3158469200134277,
|
| 589 |
+
"logits/rejected": -2.3420870304107666,
|
| 590 |
+
"logps/chosen": -1.2580420970916748,
|
| 591 |
+
"logps/rejected": -1.8185100555419922,
|
| 592 |
+
"loss": 1.3619,
|
| 593 |
+
"rewards/accuracies": 0.675000011920929,
|
| 594 |
+
"rewards/chosen": -0.09195388108491898,
|
| 595 |
+
"rewards/margins": 0.12128813564777374,
|
| 596 |
+
"rewards/rejected": -0.21324202418327332,
|
| 597 |
+
"step": 360
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.24206738632646385,
|
| 601 |
+
"grad_norm": 0.625,
|
| 602 |
+
"learning_rate": 4.6989704833880936e-06,
|
| 603 |
+
"logits/chosen": -2.413306474685669,
|
| 604 |
+
"logits/rejected": -2.269974708557129,
|
| 605 |
+
"logps/chosen": -1.261281967163086,
|
| 606 |
+
"logps/rejected": -1.7907254695892334,
|
| 607 |
+
"loss": 1.3625,
|
| 608 |
+
"rewards/accuracies": 0.625,
|
| 609 |
+
"rewards/chosen": -0.09402702748775482,
|
| 610 |
+
"rewards/margins": 0.1201835423707962,
|
| 611 |
+
"rewards/rejected": -0.21421055495738983,
|
| 612 |
+
"step": 370
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.248609748119071,
|
| 616 |
+
"grad_norm": 1.5078125,
|
| 617 |
+
"learning_rate": 4.671224940873704e-06,
|
| 618 |
+
"logits/chosen": -2.5112624168395996,
|
| 619 |
+
"logits/rejected": -2.4713196754455566,
|
| 620 |
+
"logps/chosen": -1.2645624876022339,
|
| 621 |
+
"logps/rejected": -2.146049976348877,
|
| 622 |
+
"loss": 1.3588,
|
| 623 |
+
"rewards/accuracies": 0.75,
|
| 624 |
+
"rewards/chosen": -0.09442068636417389,
|
| 625 |
+
"rewards/margins": 0.2256404459476471,
|
| 626 |
+
"rewards/rejected": -0.3200611174106598,
|
| 627 |
+
"step": 380
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.25515210991167814,
|
| 631 |
+
"grad_norm": 0.765625,
|
| 632 |
+
"learning_rate": 4.642346006013925e-06,
|
| 633 |
+
"logits/chosen": -2.540147066116333,
|
| 634 |
+
"logits/rejected": -2.479360580444336,
|
| 635 |
+
"logps/chosen": -1.2912054061889648,
|
| 636 |
+
"logps/rejected": -1.8986918926239014,
|
| 637 |
+
"loss": 1.3605,
|
| 638 |
+
"rewards/accuracies": 0.699999988079071,
|
| 639 |
+
"rewards/chosen": -0.10545520484447479,
|
| 640 |
+
"rewards/margins": 0.16333678364753723,
|
| 641 |
+
"rewards/rejected": -0.26879197359085083,
|
| 642 |
+
"step": 390
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.2616944717042852,
|
| 646 |
+
"grad_norm": 1.296875,
|
| 647 |
+
"learning_rate": 4.612348753784682e-06,
|
| 648 |
+
"logits/chosen": -2.5153968334198,
|
| 649 |
+
"logits/rejected": -2.3895232677459717,
|
| 650 |
+
"logps/chosen": -1.239985466003418,
|
| 651 |
+
"logps/rejected": -1.7698838710784912,
|
| 652 |
+
"loss": 1.3607,
|
| 653 |
+
"rewards/accuracies": 0.6875,
|
| 654 |
+
"rewards/chosen": -0.0712096244096756,
|
| 655 |
+
"rewards/margins": 0.15390561521053314,
|
| 656 |
+
"rewards/rejected": -0.22511525452136993,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.2616944717042852,
|
| 661 |
+
"eval_logits/chosen": -2.3277456760406494,
|
| 662 |
+
"eval_logits/rejected": -2.260209560394287,
|
| 663 |
+
"eval_logps/chosen": -1.140083909034729,
|
| 664 |
+
"eval_logps/rejected": -1.9083786010742188,
|
| 665 |
+
"eval_loss": 1.363683819770813,
|
| 666 |
+
"eval_rewards/accuracies": 0.6974999904632568,
|
| 667 |
+
"eval_rewards/chosen": -0.06436584144830704,
|
| 668 |
+
"eval_rewards/margins": 0.19077691435813904,
|
| 669 |
+
"eval_rewards/rejected": -0.2551427483558655,
|
| 670 |
+
"eval_runtime": 192.5656,
|
| 671 |
+
"eval_samples_per_second": 10.386,
|
| 672 |
+
"eval_steps_per_second": 0.519,
|
| 673 |
+
"step": 400
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"epoch": 0.26823683349689237,
|
| 677 |
+
"grad_norm": 0.57421875,
|
| 678 |
+
"learning_rate": 4.5812488429302245e-06,
|
| 679 |
+
"logits/chosen": -2.4453935623168945,
|
| 680 |
+
"logits/rejected": -2.369070053100586,
|
| 681 |
+
"logps/chosen": -1.1491725444793701,
|
| 682 |
+
"logps/rejected": -1.6909382343292236,
|
| 683 |
+
"loss": 1.3609,
|
| 684 |
+
"rewards/accuracies": 0.6875,
|
| 685 |
+
"rewards/chosen": -0.06435368955135345,
|
| 686 |
+
"rewards/margins": 0.14920566976070404,
|
| 687 |
+
"rewards/rejected": -0.2135593444108963,
|
| 688 |
+
"step": 410
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.2747791952894995,
|
| 692 |
+
"grad_norm": 1.0,
|
| 693 |
+
"learning_rate": 4.54906250778917e-06,
|
| 694 |
+
"logits/chosen": -2.410181760787964,
|
| 695 |
+
"logits/rejected": -2.414539337158203,
|
| 696 |
+
"logps/chosen": -1.1408861875534058,
|
| 697 |
+
"logps/rejected": -2.0139617919921875,
|
| 698 |
+
"loss": 1.352,
|
| 699 |
+
"rewards/accuracies": 0.6875,
|
| 700 |
+
"rewards/chosen": -0.06627814471721649,
|
| 701 |
+
"rewards/margins": 0.22981509566307068,
|
| 702 |
+
"rewards/rejected": -0.296093225479126,
|
| 703 |
+
"step": 420
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.28132155708210665,
|
| 707 |
+
"grad_norm": 0.55859375,
|
| 708 |
+
"learning_rate": 4.515806549820084e-06,
|
| 709 |
+
"logits/chosen": -2.3839383125305176,
|
| 710 |
+
"logits/rejected": -2.2792294025421143,
|
| 711 |
+
"logps/chosen": -1.236488938331604,
|
| 712 |
+
"logps/rejected": -1.9642620086669922,
|
| 713 |
+
"loss": 1.3659,
|
| 714 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 715 |
+
"rewards/chosen": -0.09444905072450638,
|
| 716 |
+
"rewards/margins": 0.17058061063289642,
|
| 717 |
+
"rewards/rejected": -0.2650296688079834,
|
| 718 |
+
"step": 430
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.2878639188747138,
|
| 722 |
+
"grad_norm": 1.4453125,
|
| 723 |
+
"learning_rate": 4.48149832883101e-06,
|
| 724 |
+
"logits/chosen": -2.4349424839019775,
|
| 725 |
+
"logits/rejected": -2.303603410720825,
|
| 726 |
+
"logps/chosen": -1.3925915956497192,
|
| 727 |
+
"logps/rejected": -2.356269359588623,
|
| 728 |
+
"loss": 1.3655,
|
| 729 |
+
"rewards/accuracies": 0.699999988079071,
|
| 730 |
+
"rewards/chosen": -0.12983675301074982,
|
| 731 |
+
"rewards/margins": 0.2615430951118469,
|
| 732 |
+
"rewards/rejected": -0.39137983322143555,
|
| 733 |
+
"step": 440
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.2944062806673209,
|
| 737 |
+
"grad_norm": 0.37890625,
|
| 738 |
+
"learning_rate": 4.446155753917559e-06,
|
| 739 |
+
"logits/chosen": -2.427264928817749,
|
| 740 |
+
"logits/rejected": -2.3634347915649414,
|
| 741 |
+
"logps/chosen": -1.017136573791504,
|
| 742 |
+
"logps/rejected": -1.411401629447937,
|
| 743 |
+
"loss": 1.3677,
|
| 744 |
+
"rewards/accuracies": 0.612500011920929,
|
| 745 |
+
"rewards/chosen": -0.04918839782476425,
|
| 746 |
+
"rewards/margins": 0.07388140261173248,
|
| 747 |
+
"rewards/rejected": -0.12306980043649673,
|
| 748 |
+
"step": 450
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.30094864245992803,
|
| 752 |
+
"grad_norm": 0.70703125,
|
| 753 |
+
"learning_rate": 4.409797274114245e-06,
|
| 754 |
+
"logits/chosen": -2.504045009613037,
|
| 755 |
+
"logits/rejected": -2.3919787406921387,
|
| 756 |
+
"logps/chosen": -1.056522250175476,
|
| 757 |
+
"logps/rejected": -1.699597716331482,
|
| 758 |
+
"loss": 1.3575,
|
| 759 |
+
"rewards/accuracies": 0.800000011920929,
|
| 760 |
+
"rewards/chosen": -0.033085618168115616,
|
| 761 |
+
"rewards/margins": 0.1550474911928177,
|
| 762 |
+
"rewards/rejected": -0.1881330907344818,
|
| 763 |
+
"step": 460
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.30749100425253517,
|
| 767 |
+
"grad_norm": 1.21875,
|
| 768 |
+
"learning_rate": 4.372441868763981e-06,
|
| 769 |
+
"logits/chosen": -2.3571760654449463,
|
| 770 |
+
"logits/rejected": -2.317436933517456,
|
| 771 |
+
"logps/chosen": -1.1195136308670044,
|
| 772 |
+
"logps/rejected": -2.229264497756958,
|
| 773 |
+
"loss": 1.3748,
|
| 774 |
+
"rewards/accuracies": 0.8125,
|
| 775 |
+
"rewards/chosen": -0.07059238851070404,
|
| 776 |
+
"rewards/margins": 0.27098971605300903,
|
| 777 |
+
"rewards/rejected": -0.34158211946487427,
|
| 778 |
+
"step": 470
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.3140333660451423,
|
| 782 |
+
"grad_norm": 1.0859375,
|
| 783 |
+
"learning_rate": 4.334109037610757e-06,
|
| 784 |
+
"logits/chosen": -2.4400782585144043,
|
| 785 |
+
"logits/rejected": -2.4532837867736816,
|
| 786 |
+
"logps/chosen": -1.236473798751831,
|
| 787 |
+
"logps/rejected": -1.9734086990356445,
|
| 788 |
+
"loss": 1.3618,
|
| 789 |
+
"rewards/accuracies": 0.6875,
|
| 790 |
+
"rewards/chosen": -0.08498811721801758,
|
| 791 |
+
"rewards/margins": 0.18319377303123474,
|
| 792 |
+
"rewards/rejected": -0.2681818902492523,
|
| 793 |
+
"step": 480
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.3205757278377494,
|
| 797 |
+
"grad_norm": 0.51953125,
|
| 798 |
+
"learning_rate": 4.294818790620644e-06,
|
| 799 |
+
"logits/chosen": -2.3443775177001953,
|
| 800 |
+
"logits/rejected": -2.243708372116089,
|
| 801 |
+
"logps/chosen": -0.9792125821113586,
|
| 802 |
+
"logps/rejected": -2.1587882041931152,
|
| 803 |
+
"loss": 1.3573,
|
| 804 |
+
"rewards/accuracies": 0.762499988079071,
|
| 805 |
+
"rewards/chosen": -0.030305195599794388,
|
| 806 |
+
"rewards/margins": 0.2875385582447052,
|
| 807 |
+
"rewards/rejected": -0.3178437352180481,
|
| 808 |
+
"step": 490
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.32711808963035655,
|
| 812 |
+
"grad_norm": 0.8125,
|
| 813 |
+
"learning_rate": 4.2545916375364835e-06,
|
| 814 |
+
"logits/chosen": -2.3785207271575928,
|
| 815 |
+
"logits/rejected": -2.3217720985412598,
|
| 816 |
+
"logps/chosen": -1.1460835933685303,
|
| 817 |
+
"logps/rejected": -2.0593409538269043,
|
| 818 |
+
"loss": 1.3642,
|
| 819 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 820 |
+
"rewards/chosen": -0.06508897244930267,
|
| 821 |
+
"rewards/margins": 0.23457971215248108,
|
| 822 |
+
"rewards/rejected": -0.29966866970062256,
|
| 823 |
+
"step": 500
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 0.32711808963035655,
|
| 827 |
+
"eval_logits/chosen": -2.2533767223358154,
|
| 828 |
+
"eval_logits/rejected": -2.184062957763672,
|
| 829 |
+
"eval_logps/chosen": -1.1733927726745605,
|
| 830 |
+
"eval_logps/rejected": -2.094301223754883,
|
| 831 |
+
"eval_loss": 1.3624868392944336,
|
| 832 |
+
"eval_rewards/accuracies": 0.6875,
|
| 833 |
+
"eval_rewards/chosen": -0.07435852289199829,
|
| 834 |
+
"eval_rewards/margins": 0.23656095564365387,
|
| 835 |
+
"eval_rewards/rejected": -0.31091946363449097,
|
| 836 |
+
"eval_runtime": 192.5008,
|
| 837 |
+
"eval_samples_per_second": 10.39,
|
| 838 |
+
"eval_steps_per_second": 0.519,
|
| 839 |
+
"step": 500
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 0.3336604514229637,
|
| 843 |
+
"grad_norm": 1.0390625,
|
| 844 |
+
"learning_rate": 4.213448577171676e-06,
|
| 845 |
+
"logits/chosen": -2.344510316848755,
|
| 846 |
+
"logits/rejected": -2.1984267234802246,
|
| 847 |
+
"logps/chosen": -1.0696611404418945,
|
| 848 |
+
"logps/rejected": -1.9805517196655273,
|
| 849 |
+
"loss": 1.3504,
|
| 850 |
+
"rewards/accuracies": 0.699999988079071,
|
| 851 |
+
"rewards/chosen": -0.06186025217175484,
|
| 852 |
+
"rewards/margins": 0.2206924855709076,
|
| 853 |
+
"rewards/rejected": -0.2825527489185333,
|
| 854 |
+
"step": 510
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 0.34020281321557083,
|
| 858 |
+
"grad_norm": 1.609375,
|
| 859 |
+
"learning_rate": 4.171411086448674e-06,
|
| 860 |
+
"logits/chosen": -2.3313636779785156,
|
| 861 |
+
"logits/rejected": -2.265958786010742,
|
| 862 |
+
"logps/chosen": -1.5542428493499756,
|
| 863 |
+
"logps/rejected": -2.0602355003356934,
|
| 864 |
+
"loss": 1.369,
|
| 865 |
+
"rewards/accuracies": 0.737500011920929,
|
| 866 |
+
"rewards/chosen": -0.15848460793495178,
|
| 867 |
+
"rewards/margins": 0.12949112057685852,
|
| 868 |
+
"rewards/rejected": -0.2879757285118103,
|
| 869 |
+
"step": 520
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.346745175008178,
|
| 873 |
+
"grad_norm": 1.5703125,
|
| 874 |
+
"learning_rate": 4.128501109187903e-06,
|
| 875 |
+
"logits/chosen": -2.389492988586426,
|
| 876 |
+
"logits/rejected": -2.257643222808838,
|
| 877 |
+
"logps/chosen": -1.029513955116272,
|
| 878 |
+
"logps/rejected": -2.194140911102295,
|
| 879 |
+
"loss": 1.3562,
|
| 880 |
+
"rewards/accuracies": 0.75,
|
| 881 |
+
"rewards/chosen": -0.043850500136613846,
|
| 882 |
+
"rewards/margins": 0.27712133526802063,
|
| 883 |
+
"rewards/rejected": -0.3209717869758606,
|
| 884 |
+
"step": 530
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.35328753680078506,
|
| 888 |
+
"grad_norm": 0.59765625,
|
| 889 |
+
"learning_rate": 4.084741044652956e-06,
|
| 890 |
+
"logits/chosen": -2.425668239593506,
|
| 891 |
+
"logits/rejected": -2.2258124351501465,
|
| 892 |
+
"logps/chosen": -1.0867255926132202,
|
| 893 |
+
"logps/rejected": -2.128763437271118,
|
| 894 |
+
"loss": 1.3618,
|
| 895 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 896 |
+
"rewards/chosen": -0.06123261898756027,
|
| 897 |
+
"rewards/margins": 0.2616799473762512,
|
| 898 |
+
"rewards/rejected": -0.3229126036167145,
|
| 899 |
+
"step": 540
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.3598298985933922,
|
| 903 |
+
"grad_norm": 1.21875,
|
| 904 |
+
"learning_rate": 4.040153735858041e-06,
|
| 905 |
+
"logits/chosen": -2.3242132663726807,
|
| 906 |
+
"logits/rejected": -2.2058465480804443,
|
| 907 |
+
"logps/chosen": -1.2005125284194946,
|
| 908 |
+
"logps/rejected": -1.8793357610702515,
|
| 909 |
+
"loss": 1.3602,
|
| 910 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 911 |
+
"rewards/chosen": -0.08682914078235626,
|
| 912 |
+
"rewards/margins": 0.17635366320610046,
|
| 913 |
+
"rewards/rejected": -0.26318278908729553,
|
| 914 |
+
"step": 550
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 0.36637226038599935,
|
| 918 |
+
"grad_norm": 1.015625,
|
| 919 |
+
"learning_rate": 3.9947624576437975e-06,
|
| 920 |
+
"logits/chosen": -2.3776307106018066,
|
| 921 |
+
"logits/rejected": -2.244612455368042,
|
| 922 |
+
"logps/chosen": -1.2981702089309692,
|
| 923 |
+
"logps/rejected": -2.359877824783325,
|
| 924 |
+
"loss": 1.3495,
|
| 925 |
+
"rewards/accuracies": 0.6875,
|
| 926 |
+
"rewards/chosen": -0.12080104649066925,
|
| 927 |
+
"rewards/margins": 0.25152236223220825,
|
| 928 |
+
"rewards/rejected": -0.3723233938217163,
|
| 929 |
+
"step": 560
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.3729146221786065,
|
| 933 |
+
"grad_norm": 2.03125,
|
| 934 |
+
"learning_rate": 3.948590904527689e-06,
|
| 935 |
+
"logits/chosen": -2.256986141204834,
|
| 936 |
+
"logits/rejected": -2.216099977493286,
|
| 937 |
+
"logps/chosen": -1.3303107023239136,
|
| 938 |
+
"logps/rejected": -2.433795213699341,
|
| 939 |
+
"loss": 1.3479,
|
| 940 |
+
"rewards/accuracies": 0.675000011920929,
|
| 941 |
+
"rewards/chosen": -0.11694888770580292,
|
| 942 |
+
"rewards/margins": 0.29369375109672546,
|
| 943 |
+
"rewards/rejected": -0.4106426239013672,
|
| 944 |
+
"step": 570
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 0.37945698397121363,
|
| 948 |
+
"grad_norm": 0.70703125,
|
| 949 |
+
"learning_rate": 3.901663178335318e-06,
|
| 950 |
+
"logits/chosen": -2.440453052520752,
|
| 951 |
+
"logits/rejected": -2.309227228164673,
|
| 952 |
+
"logps/chosen": -1.2955634593963623,
|
| 953 |
+
"logps/rejected": -2.1659696102142334,
|
| 954 |
+
"loss": 1.3523,
|
| 955 |
+
"rewards/accuracies": 0.762499988079071,
|
| 956 |
+
"rewards/chosen": -0.09721510112285614,
|
| 957 |
+
"rewards/margins": 0.2235115021467209,
|
| 958 |
+
"rewards/rejected": -0.32072657346725464,
|
| 959 |
+
"step": 580
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.3859993457638207,
|
| 963 |
+
"grad_norm": 2.109375,
|
| 964 |
+
"learning_rate": 3.854003775619142e-06,
|
| 965 |
+
"logits/chosen": -2.2439610958099365,
|
| 966 |
+
"logits/rejected": -2.22778582572937,
|
| 967 |
+
"logps/chosen": -1.3622268438339233,
|
| 968 |
+
"logps/rejected": -2.140564441680908,
|
| 969 |
+
"loss": 1.357,
|
| 970 |
+
"rewards/accuracies": 0.699999988079071,
|
| 971 |
+
"rewards/chosen": -0.10588552057743073,
|
| 972 |
+
"rewards/margins": 0.22728273272514343,
|
| 973 |
+
"rewards/rejected": -0.33316826820373535,
|
| 974 |
+
"step": 590
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.39254170755642787,
|
| 978 |
+
"grad_norm": 1.3828125,
|
| 979 |
+
"learning_rate": 3.805637574871115e-06,
|
| 980 |
+
"logits/chosen": -2.2500593662261963,
|
| 981 |
+
"logits/rejected": -2.1690380573272705,
|
| 982 |
+
"logps/chosen": -1.1624057292938232,
|
| 983 |
+
"logps/rejected": -2.2276828289031982,
|
| 984 |
+
"loss": 1.3489,
|
| 985 |
+
"rewards/accuracies": 0.75,
|
| 986 |
+
"rewards/chosen": -0.07329526543617249,
|
| 987 |
+
"rewards/margins": 0.307359516620636,
|
| 988 |
+
"rewards/rejected": -0.38065481185913086,
|
| 989 |
+
"step": 600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.39254170755642787,
|
| 993 |
+
"eval_logits/chosen": -2.103938341140747,
|
| 994 |
+
"eval_logits/rejected": -2.0262529850006104,
|
| 995 |
+
"eval_logps/chosen": -1.2905856370925903,
|
| 996 |
+
"eval_logps/rejected": -2.456772565841675,
|
| 997 |
+
"eval_loss": 1.3648583889007568,
|
| 998 |
+
"eval_rewards/accuracies": 0.6850000023841858,
|
| 999 |
+
"eval_rewards/chosen": -0.10951638966798782,
|
| 1000 |
+
"eval_rewards/margins": 0.3101446032524109,
|
| 1001 |
+
"eval_rewards/rejected": -0.4196610152721405,
|
| 1002 |
+
"eval_runtime": 192.5629,
|
| 1003 |
+
"eval_samples_per_second": 10.386,
|
| 1004 |
+
"eval_steps_per_second": 0.519,
|
| 1005 |
+
"step": 600
|
| 1006 |
+
},
|
| 1007 |
+
{
|
| 1008 |
+
"epoch": 0.399084069349035,
|
| 1009 |
+
"grad_norm": 1.046875,
|
| 1010 |
+
"learning_rate": 3.7565898235359717e-06,
|
| 1011 |
+
"logits/chosen": -2.273394823074341,
|
| 1012 |
+
"logits/rejected": -2.1735503673553467,
|
| 1013 |
+
"logps/chosen": -1.2579439878463745,
|
| 1014 |
+
"logps/rejected": -2.1365649700164795,
|
| 1015 |
+
"loss": 1.3734,
|
| 1016 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1017 |
+
"rewards/chosen": -0.10619225353002548,
|
| 1018 |
+
"rewards/margins": 0.23582622408866882,
|
| 1019 |
+
"rewards/rejected": -0.3420184850692749,
|
| 1020 |
+
"step": 610
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.40562643114164215,
|
| 1024 |
+
"grad_norm": 0.7578125,
|
| 1025 |
+
"learning_rate": 3.7068861248319127e-06,
|
| 1026 |
+
"logits/chosen": -2.228017807006836,
|
| 1027 |
+
"logits/rejected": -2.165989398956299,
|
| 1028 |
+
"logps/chosen": -1.0932655334472656,
|
| 1029 |
+
"logps/rejected": -2.0412039756774902,
|
| 1030 |
+
"loss": 1.364,
|
| 1031 |
+
"rewards/accuracies": 0.6875,
|
| 1032 |
+
"rewards/chosen": -0.05144476890563965,
|
| 1033 |
+
"rewards/margins": 0.2414330542087555,
|
| 1034 |
+
"rewards/rejected": -0.29287782311439514,
|
| 1035 |
+
"step": 620
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 0.41216879293424924,
|
| 1039 |
+
"grad_norm": 0.671875,
|
| 1040 |
+
"learning_rate": 3.6565524243855695e-06,
|
| 1041 |
+
"logits/chosen": -2.2673258781433105,
|
| 1042 |
+
"logits/rejected": -2.1727194786071777,
|
| 1043 |
+
"logps/chosen": -1.0849422216415405,
|
| 1044 |
+
"logps/rejected": -2.769263744354248,
|
| 1045 |
+
"loss": 1.354,
|
| 1046 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1047 |
+
"rewards/chosen": -0.04491913691163063,
|
| 1048 |
+
"rewards/margins": 0.44512462615966797,
|
| 1049 |
+
"rewards/rejected": -0.4900437295436859,
|
| 1050 |
+
"step": 630
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"epoch": 0.4187111547268564,
|
| 1054 |
+
"grad_norm": 0.92578125,
|
| 1055 |
+
"learning_rate": 3.6056149966882325e-06,
|
| 1056 |
+
"logits/chosen": -2.2358603477478027,
|
| 1057 |
+
"logits/rejected": -2.202164888381958,
|
| 1058 |
+
"logps/chosen": -1.0952130556106567,
|
| 1059 |
+
"logps/rejected": -2.2472681999206543,
|
| 1060 |
+
"loss": 1.3624,
|
| 1061 |
+
"rewards/accuracies": 0.6875,
|
| 1062 |
+
"rewards/chosen": -0.06672855466604233,
|
| 1063 |
+
"rewards/margins": 0.2949059009552002,
|
| 1064 |
+
"rewards/rejected": -0.3616344630718231,
|
| 1065 |
+
"step": 640
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 0.4252535165194635,
|
| 1069 |
+
"grad_norm": 0.6953125,
|
| 1070 |
+
"learning_rate": 3.554100431380414e-06,
|
| 1071 |
+
"logits/chosen": -2.2916007041931152,
|
| 1072 |
+
"logits/rejected": -2.186967134475708,
|
| 1073 |
+
"logps/chosen": -1.1993496417999268,
|
| 1074 |
+
"logps/rejected": -2.332792043685913,
|
| 1075 |
+
"loss": 1.3578,
|
| 1076 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1077 |
+
"rewards/chosen": -0.07057426869869232,
|
| 1078 |
+
"rewards/margins": 0.2939244210720062,
|
| 1079 |
+
"rewards/rejected": -0.36449870467185974,
|
| 1080 |
+
"step": 650
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 0.43179587831207067,
|
| 1084 |
+
"grad_norm": 0.875,
|
| 1085 |
+
"learning_rate": 3.5020356193718934e-06,
|
| 1086 |
+
"logits/chosen": -2.22035813331604,
|
| 1087 |
+
"logits/rejected": -2.168062686920166,
|
| 1088 |
+
"logps/chosen": -1.1497726440429688,
|
| 1089 |
+
"logps/rejected": -2.155555248260498,
|
| 1090 |
+
"loss": 1.3671,
|
| 1091 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1092 |
+
"rewards/chosen": -0.07245108485221863,
|
| 1093 |
+
"rewards/margins": 0.2514120042324066,
|
| 1094 |
+
"rewards/rejected": -0.32386311888694763,
|
| 1095 |
+
"step": 660
|
| 1096 |
+
},
|
| 1097 |
+
{
|
| 1098 |
+
"epoch": 0.4383382401046778,
|
| 1099 |
+
"grad_norm": 1.0390625,
|
| 1100 |
+
"learning_rate": 3.4494477388045035e-06,
|
| 1101 |
+
"logits/chosen": -2.310704469680786,
|
| 1102 |
+
"logits/rejected": -2.0709216594696045,
|
| 1103 |
+
"logps/chosen": -1.172662377357483,
|
| 1104 |
+
"logps/rejected": -1.904979944229126,
|
| 1105 |
+
"loss": 1.3603,
|
| 1106 |
+
"rewards/accuracies": 0.75,
|
| 1107 |
+
"rewards/chosen": -0.05323634669184685,
|
| 1108 |
+
"rewards/margins": 0.19738993048667908,
|
| 1109 |
+
"rewards/rejected": -0.25062626600265503,
|
| 1110 |
+
"step": 670
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"epoch": 0.4448806018972849,
|
| 1114 |
+
"grad_norm": 2.875,
|
| 1115 |
+
"learning_rate": 3.3963642408649783e-06,
|
| 1116 |
+
"logits/chosen": -2.2430570125579834,
|
| 1117 |
+
"logits/rejected": -2.1614744663238525,
|
| 1118 |
+
"logps/chosen": -1.2116143703460693,
|
| 1119 |
+
"logps/rejected": -2.23248028755188,
|
| 1120 |
+
"loss": 1.354,
|
| 1121 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1122 |
+
"rewards/chosen": -0.07912047952413559,
|
| 1123 |
+
"rewards/margins": 0.2703726291656494,
|
| 1124 |
+
"rewards/rejected": -0.3494931161403656,
|
| 1125 |
+
"step": 680
|
| 1126 |
+
},
|
| 1127 |
+
{
|
| 1128 |
+
"epoch": 0.45142296368989204,
|
| 1129 |
+
"grad_norm": 1.3671875,
|
| 1130 |
+
"learning_rate": 3.3428128354552727e-06,
|
| 1131 |
+
"logits/chosen": -2.129528522491455,
|
| 1132 |
+
"logits/rejected": -2.0921757221221924,
|
| 1133 |
+
"logps/chosen": -1.3085339069366455,
|
| 1134 |
+
"logps/rejected": -2.421966314315796,
|
| 1135 |
+
"loss": 1.3511,
|
| 1136 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1137 |
+
"rewards/chosen": -0.10610251128673553,
|
| 1138 |
+
"rewards/margins": 0.3096415400505066,
|
| 1139 |
+
"rewards/rejected": -0.4157440662384033,
|
| 1140 |
+
"step": 690
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"epoch": 0.4579653254824992,
|
| 1144 |
+
"grad_norm": 3.203125,
|
| 1145 |
+
"learning_rate": 3.2888214767278246e-06,
|
| 1146 |
+
"logits/chosen": -2.199361562728882,
|
| 1147 |
+
"logits/rejected": -2.0280404090881348,
|
| 1148 |
+
"logps/chosen": -1.3939224481582642,
|
| 1149 |
+
"logps/rejected": -2.0829360485076904,
|
| 1150 |
+
"loss": 1.3735,
|
| 1151 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1152 |
+
"rewards/chosen": -0.14213696122169495,
|
| 1153 |
+
"rewards/margins": 0.17512670159339905,
|
| 1154 |
+
"rewards/rejected": -0.3172636330127716,
|
| 1155 |
+
"step": 700
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 0.4579653254824992,
|
| 1159 |
+
"eval_logits/chosen": -2.015493392944336,
|
| 1160 |
+
"eval_logits/rejected": -1.9237298965454102,
|
| 1161 |
+
"eval_logps/chosen": -1.274337887763977,
|
| 1162 |
+
"eval_logps/rejected": -2.438903570175171,
|
| 1163 |
+
"eval_loss": 1.365309715270996,
|
| 1164 |
+
"eval_rewards/accuracies": 0.699999988079071,
|
| 1165 |
+
"eval_rewards/chosen": -0.10464204847812653,
|
| 1166 |
+
"eval_rewards/margins": 0.3096581697463989,
|
| 1167 |
+
"eval_rewards/rejected": -0.41430023312568665,
|
| 1168 |
+
"eval_runtime": 192.5749,
|
| 1169 |
+
"eval_samples_per_second": 10.386,
|
| 1170 |
+
"eval_steps_per_second": 0.519,
|
| 1171 |
+
"step": 700
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 0.46450768727510633,
|
| 1175 |
+
"grad_norm": 0.6640625,
|
| 1176 |
+
"learning_rate": 3.2344183484933247e-06,
|
| 1177 |
+
"logits/chosen": -2.1935503482818604,
|
| 1178 |
+
"logits/rejected": -2.140709400177002,
|
| 1179 |
+
"logps/chosen": -1.2663156986236572,
|
| 1180 |
+
"logps/rejected": -2.3020918369293213,
|
| 1181 |
+
"loss": 1.3563,
|
| 1182 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1183 |
+
"rewards/chosen": -0.09797848761081696,
|
| 1184 |
+
"rewards/margins": 0.29972079396247864,
|
| 1185 |
+
"rewards/rejected": -0.3976992964744568,
|
| 1186 |
+
"step": 710
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"epoch": 0.47105004906771347,
|
| 1190 |
+
"grad_norm": 0.69921875,
|
| 1191 |
+
"learning_rate": 3.179631849508597e-06,
|
| 1192 |
+
"logits/chosen": -2.1977057456970215,
|
| 1193 |
+
"logits/rejected": -2.2053167819976807,
|
| 1194 |
+
"logps/chosen": -1.2915914058685303,
|
| 1195 |
+
"logps/rejected": -1.9569549560546875,
|
| 1196 |
+
"loss": 1.3619,
|
| 1197 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1198 |
+
"rewards/chosen": -0.10860061645507812,
|
| 1199 |
+
"rewards/margins": 0.14782245457172394,
|
| 1200 |
+
"rewards/rejected": -0.25642308592796326,
|
| 1201 |
+
"step": 720
|
| 1202 |
+
},
|
| 1203 |
+
{
|
| 1204 |
+
"epoch": 0.47759241086032056,
|
| 1205 |
+
"grad_norm": 0.67578125,
|
| 1206 |
+
"learning_rate": 3.1244905786522796e-06,
|
| 1207 |
+
"logits/chosen": -2.2671637535095215,
|
| 1208 |
+
"logits/rejected": -2.1876721382141113,
|
| 1209 |
+
"logps/chosen": -1.1692320108413696,
|
| 1210 |
+
"logps/rejected": -1.852294683456421,
|
| 1211 |
+
"loss": 1.3602,
|
| 1212 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1213 |
+
"rewards/chosen": -0.07541967183351517,
|
| 1214 |
+
"rewards/margins": 0.19343288242816925,
|
| 1215 |
+
"rewards/rejected": -0.268852561712265,
|
| 1216 |
+
"step": 730
|
| 1217 |
+
},
|
| 1218 |
+
{
|
| 1219 |
+
"epoch": 0.4841347726529277,
|
| 1220 |
+
"grad_norm": 2.296875,
|
| 1221 |
+
"learning_rate": 3.0690233199960393e-06,
|
| 1222 |
+
"logits/chosen": -2.04817271232605,
|
| 1223 |
+
"logits/rejected": -1.990860939025879,
|
| 1224 |
+
"logps/chosen": -1.218370795249939,
|
| 1225 |
+
"logps/rejected": -1.9844430685043335,
|
| 1226 |
+
"loss": 1.3607,
|
| 1227 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1228 |
+
"rewards/chosen": -0.0848841518163681,
|
| 1229 |
+
"rewards/margins": 0.20134691894054413,
|
| 1230 |
+
"rewards/rejected": -0.2862311005592346,
|
| 1231 |
+
"step": 740
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"epoch": 0.49067713444553485,
|
| 1235 |
+
"grad_norm": 0.92578125,
|
| 1236 |
+
"learning_rate": 3.0132590277791163e-06,
|
| 1237 |
+
"logits/chosen": -2.258011817932129,
|
| 1238 |
+
"logits/rejected": -2.1347296237945557,
|
| 1239 |
+
"logps/chosen": -1.089874029159546,
|
| 1240 |
+
"logps/rejected": -2.269296646118164,
|
| 1241 |
+
"loss": 1.3491,
|
| 1242 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1243 |
+
"rewards/chosen": -0.06120484322309494,
|
| 1244 |
+
"rewards/margins": 0.3166188597679138,
|
| 1245 |
+
"rewards/rejected": -0.37782374024391174,
|
| 1246 |
+
"step": 750
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"epoch": 0.497219496238142,
|
| 1250 |
+
"grad_norm": 3.25,
|
| 1251 |
+
"learning_rate": 2.9572268112940354e-06,
|
| 1252 |
+
"logits/chosen": -2.220092296600342,
|
| 1253 |
+
"logits/rejected": -2.2047886848449707,
|
| 1254 |
+
"logps/chosen": -1.5497848987579346,
|
| 1255 |
+
"logps/rejected": -2.784870147705078,
|
| 1256 |
+
"loss": 1.3614,
|
| 1257 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1258 |
+
"rewards/chosen": -0.18973752856254578,
|
| 1259 |
+
"rewards/margins": 0.32363444566726685,
|
| 1260 |
+
"rewards/rejected": -0.5133720636367798,
|
| 1261 |
+
"step": 760
|
| 1262 |
+
},
|
| 1263 |
+
{
|
| 1264 |
+
"epoch": 0.5037618580307491,
|
| 1265 |
+
"grad_norm": 0.984375,
|
| 1266 |
+
"learning_rate": 2.9009559196913882e-06,
|
| 1267 |
+
"logits/chosen": -2.261192798614502,
|
| 1268 |
+
"logits/rejected": -2.1291980743408203,
|
| 1269 |
+
"logps/chosen": -1.2792052030563354,
|
| 1270 |
+
"logps/rejected": -2.2746200561523438,
|
| 1271 |
+
"loss": 1.3559,
|
| 1272 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1273 |
+
"rewards/chosen": -0.08147454261779785,
|
| 1274 |
+
"rewards/margins": 0.27496951818466187,
|
| 1275 |
+
"rewards/rejected": -0.3564440906047821,
|
| 1276 |
+
"step": 770
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"epoch": 0.5103042198233563,
|
| 1280 |
+
"grad_norm": 0.74609375,
|
| 1281 |
+
"learning_rate": 2.844475726711595e-06,
|
| 1282 |
+
"logits/chosen": -2.2254199981689453,
|
| 1283 |
+
"logits/rejected": -2.0712227821350098,
|
| 1284 |
+
"logps/chosen": -1.2435195446014404,
|
| 1285 |
+
"logps/rejected": -2.187425136566162,
|
| 1286 |
+
"loss": 1.3536,
|
| 1287 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 1288 |
+
"rewards/chosen": -0.09367978572845459,
|
| 1289 |
+
"rewards/margins": 0.2134389877319336,
|
| 1290 |
+
"rewards/rejected": -0.30711880326271057,
|
| 1291 |
+
"step": 780
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"epoch": 0.5168465816159633,
|
| 1295 |
+
"grad_norm": 1.703125,
|
| 1296 |
+
"learning_rate": 2.7878157153516446e-06,
|
| 1297 |
+
"logits/chosen": -2.119149684906006,
|
| 1298 |
+
"logits/rejected": -2.021897554397583,
|
| 1299 |
+
"logps/chosen": -1.083522081375122,
|
| 1300 |
+
"logps/rejected": -2.1679904460906982,
|
| 1301 |
+
"loss": 1.3483,
|
| 1302 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1303 |
+
"rewards/chosen": -0.054320644587278366,
|
| 1304 |
+
"rewards/margins": 0.27394142746925354,
|
| 1305 |
+
"rewards/rejected": -0.328262060880661,
|
| 1306 |
+
"step": 790
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 0.5233889434085705,
|
| 1310 |
+
"grad_norm": 1.0625,
|
| 1311 |
+
"learning_rate": 2.731005462474787e-06,
|
| 1312 |
+
"logits/chosen": -2.2120959758758545,
|
| 1313 |
+
"logits/rejected": -2.121436595916748,
|
| 1314 |
+
"logps/chosen": -1.0082324743270874,
|
| 1315 |
+
"logps/rejected": -2.022320508956909,
|
| 1316 |
+
"loss": 1.3606,
|
| 1317 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1318 |
+
"rewards/chosen": -0.05436345189809799,
|
| 1319 |
+
"rewards/margins": 0.24595093727111816,
|
| 1320 |
+
"rewards/rejected": -0.30031436681747437,
|
| 1321 |
+
"step": 800
|
| 1322 |
+
},
|
| 1323 |
+
{
|
| 1324 |
+
"epoch": 0.5233889434085705,
|
| 1325 |
+
"eval_logits/chosen": -2.035632371902466,
|
| 1326 |
+
"eval_logits/rejected": -1.9493231773376465,
|
| 1327 |
+
"eval_logps/chosen": -1.1739249229431152,
|
| 1328 |
+
"eval_logps/rejected": -2.291534900665283,
|
| 1329 |
+
"eval_loss": 1.3592066764831543,
|
| 1330 |
+
"eval_rewards/accuracies": 0.6949999928474426,
|
| 1331 |
+
"eval_rewards/chosen": -0.074518121778965,
|
| 1332 |
+
"eval_rewards/margins": 0.2955714762210846,
|
| 1333 |
+
"eval_rewards/rejected": -0.3700896203517914,
|
| 1334 |
+
"eval_runtime": 192.6116,
|
| 1335 |
+
"eval_samples_per_second": 10.384,
|
| 1336 |
+
"eval_steps_per_second": 0.519,
|
| 1337 |
+
"step": 800
|
| 1338 |
+
},
|
| 1339 |
+
{
|
| 1340 |
+
"epoch": 0.5299313052011776,
|
| 1341 |
+
"grad_norm": 1.3671875,
|
| 1342 |
+
"learning_rate": 2.67407462337124e-06,
|
| 1343 |
+
"logits/chosen": -2.2169487476348877,
|
| 1344 |
+
"logits/rejected": -2.0953001976013184,
|
| 1345 |
+
"logps/chosen": -1.4013360738754272,
|
| 1346 |
+
"logps/rejected": -2.4634041786193848,
|
| 1347 |
+
"loss": 1.3603,
|
| 1348 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1349 |
+
"rewards/chosen": -0.1341012716293335,
|
| 1350 |
+
"rewards/margins": 0.2964847683906555,
|
| 1351 |
+
"rewards/rejected": -0.430586040019989,
|
| 1352 |
+
"step": 810
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 0.5364736669937847,
|
| 1356 |
+
"grad_norm": 1.9140625,
|
| 1357 |
+
"learning_rate": 2.617052916277952e-06,
|
| 1358 |
+
"logits/chosen": -2.237308979034424,
|
| 1359 |
+
"logits/rejected": -2.043175458908081,
|
| 1360 |
+
"logps/chosen": -1.1022270917892456,
|
| 1361 |
+
"logps/rejected": -2.4109721183776855,
|
| 1362 |
+
"loss": 1.3607,
|
| 1363 |
+
"rewards/accuracies": 0.75,
|
| 1364 |
+
"rewards/chosen": -0.0516006276011467,
|
| 1365 |
+
"rewards/margins": 0.34269508719444275,
|
| 1366 |
+
"rewards/rejected": -0.39429575204849243,
|
| 1367 |
+
"step": 820
|
| 1368 |
+
},
|
| 1369 |
+
{
|
| 1370 |
+
"epoch": 0.5430160287863919,
|
| 1371 |
+
"grad_norm": 1.3828125,
|
| 1372 |
+
"learning_rate": 2.5599701068654985e-06,
|
| 1373 |
+
"logits/chosen": -2.218151092529297,
|
| 1374 |
+
"logits/rejected": -2.1844322681427,
|
| 1375 |
+
"logps/chosen": -1.1159567832946777,
|
| 1376 |
+
"logps/rejected": -2.1058316230773926,
|
| 1377 |
+
"loss": 1.355,
|
| 1378 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1379 |
+
"rewards/chosen": -0.053866881877183914,
|
| 1380 |
+
"rewards/margins": 0.2691417634487152,
|
| 1381 |
+
"rewards/rejected": -0.32300865650177,
|
| 1382 |
+
"step": 830
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"epoch": 0.549558390578999,
|
| 1386 |
+
"grad_norm": 0.875,
|
| 1387 |
+
"learning_rate": 2.5028559927002326e-06,
|
| 1388 |
+
"logits/chosen": -2.2796895503997803,
|
| 1389 |
+
"logits/rejected": -2.186133861541748,
|
| 1390 |
+
"logps/chosen": -0.9579225778579712,
|
| 1391 |
+
"logps/rejected": -1.9913718700408936,
|
| 1392 |
+
"loss": 1.3578,
|
| 1393 |
+
"rewards/accuracies": 0.6875,
|
| 1394 |
+
"rewards/chosen": -0.025524040684103966,
|
| 1395 |
+
"rewards/margins": 0.27697938680648804,
|
| 1396 |
+
"rewards/rejected": -0.30250340700149536,
|
| 1397 |
+
"step": 840
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 0.5561007523716062,
|
| 1401 |
+
"grad_norm": 0.796875,
|
| 1402 |
+
"learning_rate": 2.4457403876897756e-06,
|
| 1403 |
+
"logits/chosen": -2.087404251098633,
|
| 1404 |
+
"logits/rejected": -2.0218749046325684,
|
| 1405 |
+
"logps/chosen": -1.2815109491348267,
|
| 1406 |
+
"logps/rejected": -1.9342288970947266,
|
| 1407 |
+
"loss": 1.3532,
|
| 1408 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1409 |
+
"rewards/chosen": -0.09889832884073257,
|
| 1410 |
+
"rewards/margins": 0.20040933787822723,
|
| 1411 |
+
"rewards/rejected": -0.2993076741695404,
|
| 1412 |
+
"step": 850
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 0.5626431141642133,
|
| 1416 |
+
"grad_norm": 0.9609375,
|
| 1417 |
+
"learning_rate": 2.388653106519975e-06,
|
| 1418 |
+
"logits/chosen": -2.2493090629577637,
|
| 1419 |
+
"logits/rejected": -2.053457498550415,
|
| 1420 |
+
"logps/chosen": -1.3042584657669067,
|
| 1421 |
+
"logps/rejected": -2.256798028945923,
|
| 1422 |
+
"loss": 1.3593,
|
| 1423 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1424 |
+
"rewards/chosen": -0.10398749262094498,
|
| 1425 |
+
"rewards/margins": 0.2570543885231018,
|
| 1426 |
+
"rewards/rejected": -0.36104193329811096,
|
| 1427 |
+
"step": 860
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 0.5691854759568205,
|
| 1431 |
+
"grad_norm": 2.09375,
|
| 1432 |
+
"learning_rate": 2.331623949091467e-06,
|
| 1433 |
+
"logits/chosen": -2.2236335277557373,
|
| 1434 |
+
"logits/rejected": -2.198726177215576,
|
| 1435 |
+
"logps/chosen": -1.1993200778961182,
|
| 1436 |
+
"logps/rejected": -1.8928115367889404,
|
| 1437 |
+
"loss": 1.3601,
|
| 1438 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1439 |
+
"rewards/chosen": -0.09070239961147308,
|
| 1440 |
+
"rewards/margins": 0.17054910957813263,
|
| 1441 |
+
"rewards/rejected": -0.2612515091896057,
|
| 1442 |
+
"step": 870
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 0.5757278377494276,
|
| 1446 |
+
"grad_norm": 1.296875,
|
| 1447 |
+
"learning_rate": 2.2746826849639513e-06,
|
| 1448 |
+
"logits/chosen": -2.2221107482910156,
|
| 1449 |
+
"logits/rejected": -2.101364850997925,
|
| 1450 |
+
"logps/chosen": -1.217944860458374,
|
| 1451 |
+
"logps/rejected": -2.0411202907562256,
|
| 1452 |
+
"loss": 1.3569,
|
| 1453 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1454 |
+
"rewards/chosen": -0.08954662829637527,
|
| 1455 |
+
"rewards/margins": 0.22928118705749512,
|
| 1456 |
+
"rewards/rejected": -0.3188278079032898,
|
| 1457 |
+
"step": 880
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 0.5822701995420346,
|
| 1461 |
+
"grad_norm": 1.1015625,
|
| 1462 |
+
"learning_rate": 2.2178590378162957e-06,
|
| 1463 |
+
"logits/chosen": -2.2392585277557373,
|
| 1464 |
+
"logits/rejected": -2.123332977294922,
|
| 1465 |
+
"logps/chosen": -1.3043301105499268,
|
| 1466 |
+
"logps/rejected": -2.3116343021392822,
|
| 1467 |
+
"loss": 1.361,
|
| 1468 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1469 |
+
"rewards/chosen": -0.10152602195739746,
|
| 1470 |
+
"rewards/margins": 0.2721943259239197,
|
| 1471 |
+
"rewards/rejected": -0.37372034788131714,
|
| 1472 |
+
"step": 890
|
| 1473 |
+
},
|
| 1474 |
+
{
|
| 1475 |
+
"epoch": 0.5888125613346418,
|
| 1476 |
+
"grad_norm": 1.7109375,
|
| 1477 |
+
"learning_rate": 2.1611826699306104e-06,
|
| 1478 |
+
"logits/chosen": -2.129063844680786,
|
| 1479 |
+
"logits/rejected": -2.0367393493652344,
|
| 1480 |
+
"logps/chosen": -1.134124755859375,
|
| 1481 |
+
"logps/rejected": -2.7047417163848877,
|
| 1482 |
+
"loss": 1.3462,
|
| 1483 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1484 |
+
"rewards/chosen": -0.05923975631594658,
|
| 1485 |
+
"rewards/margins": 0.4080350399017334,
|
| 1486 |
+
"rewards/rejected": -0.46727481484413147,
|
| 1487 |
+
"step": 900
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 0.5888125613346418,
|
| 1491 |
+
"eval_logits/chosen": -2.060875415802002,
|
| 1492 |
+
"eval_logits/rejected": -1.9784667491912842,
|
| 1493 |
+
"eval_logps/chosen": -1.210047960281372,
|
| 1494 |
+
"eval_logps/rejected": -2.2806875705718994,
|
| 1495 |
+
"eval_loss": 1.3567863702774048,
|
| 1496 |
+
"eval_rewards/accuracies": 0.7049999833106995,
|
| 1497 |
+
"eval_rewards/chosen": -0.08535508811473846,
|
| 1498 |
+
"eval_rewards/margins": 0.2814803123474121,
|
| 1499 |
+
"eval_rewards/rejected": -0.36683544516563416,
|
| 1500 |
+
"eval_runtime": 192.5258,
|
| 1501 |
+
"eval_samples_per_second": 10.388,
|
| 1502 |
+
"eval_steps_per_second": 0.519,
|
| 1503 |
+
"step": 900
|
| 1504 |
+
},
|
| 1505 |
+
{
|
| 1506 |
+
"epoch": 0.5953549231272489,
|
| 1507 |
+
"grad_norm": 1.2109375,
|
| 1508 |
+
"learning_rate": 2.1046831667083483e-06,
|
| 1509 |
+
"logits/chosen": -2.20284104347229,
|
| 1510 |
+
"logits/rejected": -2.077505111694336,
|
| 1511 |
+
"logps/chosen": -1.4609758853912354,
|
| 1512 |
+
"logps/rejected": -2.2469534873962402,
|
| 1513 |
+
"loss": 1.3637,
|
| 1514 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1515 |
+
"rewards/chosen": -0.123102568089962,
|
| 1516 |
+
"rewards/margins": 0.23405058681964874,
|
| 1517 |
+
"rewards/rejected": -0.35715317726135254,
|
| 1518 |
+
"step": 910
|
| 1519 |
+
},
|
| 1520 |
+
{
|
| 1521 |
+
"epoch": 0.6018972849198561,
|
| 1522 |
+
"grad_norm": 1.25,
|
| 1523 |
+
"learning_rate": 2.048390021226559e-06,
|
| 1524 |
+
"logits/chosen": -2.1845784187316895,
|
| 1525 |
+
"logits/rejected": -2.092951536178589,
|
| 1526 |
+
"logps/chosen": -1.1394016742706299,
|
| 1527 |
+
"logps/rejected": -2.110645055770874,
|
| 1528 |
+
"loss": 1.3573,
|
| 1529 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1530 |
+
"rewards/chosen": -0.05954326316714287,
|
| 1531 |
+
"rewards/margins": 0.2537994980812073,
|
| 1532 |
+
"rewards/rejected": -0.31334277987480164,
|
| 1533 |
+
"step": 920
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 0.6084396467124632,
|
| 1537 |
+
"grad_norm": 1.828125,
|
| 1538 |
+
"learning_rate": 1.9923326188423212e-06,
|
| 1539 |
+
"logits/chosen": -2.264650821685791,
|
| 1540 |
+
"logits/rejected": -2.1612370014190674,
|
| 1541 |
+
"logps/chosen": -1.2139930725097656,
|
| 1542 |
+
"logps/rejected": -1.9334100484848022,
|
| 1543 |
+
"loss": 1.3628,
|
| 1544 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1545 |
+
"rewards/chosen": -0.07819408178329468,
|
| 1546 |
+
"rewards/margins": 0.19942335784435272,
|
| 1547 |
+
"rewards/rejected": -0.2776174545288086,
|
| 1548 |
+
"step": 930
|
| 1549 |
+
},
|
| 1550 |
+
{
|
| 1551 |
+
"epoch": 0.6149820085050703,
|
| 1552 |
+
"grad_norm": 1.25,
|
| 1553 |
+
"learning_rate": 1.936540221853415e-06,
|
| 1554 |
+
"logits/chosen": -2.1965224742889404,
|
| 1555 |
+
"logits/rejected": -2.0824341773986816,
|
| 1556 |
+
"logps/chosen": -1.2189127206802368,
|
| 1557 |
+
"logps/rejected": -2.891991376876831,
|
| 1558 |
+
"loss": 1.3409,
|
| 1559 |
+
"rewards/accuracies": 0.75,
|
| 1560 |
+
"rewards/chosen": -0.08479885011911392,
|
| 1561 |
+
"rewards/margins": 0.45503416657447815,
|
| 1562 |
+
"rewards/rejected": -0.5398330092430115,
|
| 1563 |
+
"step": 940
|
| 1564 |
+
},
|
| 1565 |
+
{
|
| 1566 |
+
"epoch": 0.6215243702976775,
|
| 1567 |
+
"grad_norm": 1.1796875,
|
| 1568 |
+
"learning_rate": 1.8810419542232245e-06,
|
| 1569 |
+
"logits/chosen": -2.2407994270324707,
|
| 1570 |
+
"logits/rejected": -2.135453939437866,
|
| 1571 |
+
"logps/chosen": -1.1888017654418945,
|
| 1572 |
+
"logps/rejected": -1.9870593547821045,
|
| 1573 |
+
"loss": 1.3479,
|
| 1574 |
+
"rewards/accuracies": 0.75,
|
| 1575 |
+
"rewards/chosen": -0.0872371569275856,
|
| 1576 |
+
"rewards/margins": 0.1756531447172165,
|
| 1577 |
+
"rewards/rejected": -0.2628903090953827,
|
| 1578 |
+
"step": 950
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"epoch": 0.6280667320902846,
|
| 1582 |
+
"grad_norm": 1.5546875,
|
| 1583 |
+
"learning_rate": 1.8258667863778573e-06,
|
| 1584 |
+
"logits/chosen": -2.12614107131958,
|
| 1585 |
+
"logits/rejected": -2.054868221282959,
|
| 1586 |
+
"logps/chosen": -1.2383328676223755,
|
| 1587 |
+
"logps/rejected": -2.1294407844543457,
|
| 1588 |
+
"loss": 1.3472,
|
| 1589 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 1590 |
+
"rewards/chosen": -0.10099540650844574,
|
| 1591 |
+
"rewards/margins": 0.2554927170276642,
|
| 1592 |
+
"rewards/rejected": -0.3564881384372711,
|
| 1593 |
+
"step": 960
|
| 1594 |
+
},
|
| 1595 |
+
{
|
| 1596 |
+
"epoch": 0.6346090938828918,
|
| 1597 |
+
"grad_norm": 0.64453125,
|
| 1598 |
+
"learning_rate": 1.7710435200834126e-06,
|
| 1599 |
+
"logits/chosen": -2.1406524181365967,
|
| 1600 |
+
"logits/rejected": -2.126784086227417,
|
| 1601 |
+
"logps/chosen": -1.0479339361190796,
|
| 1602 |
+
"logps/rejected": -2.2236762046813965,
|
| 1603 |
+
"loss": 1.3411,
|
| 1604 |
+
"rewards/accuracies": 0.8125,
|
| 1605 |
+
"rewards/chosen": -0.047475624829530716,
|
| 1606 |
+
"rewards/margins": 0.30173081159591675,
|
| 1607 |
+
"rewards/rejected": -0.34920641779899597,
|
| 1608 |
+
"step": 970
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"epoch": 0.6411514556754988,
|
| 1612 |
+
"grad_norm": 3.921875,
|
| 1613 |
+
"learning_rate": 1.7166007734112808e-06,
|
| 1614 |
+
"logits/chosen": -2.194304943084717,
|
| 1615 |
+
"logits/rejected": -2.0795140266418457,
|
| 1616 |
+
"logps/chosen": -1.2003707885742188,
|
| 1617 |
+
"logps/rejected": -2.1898512840270996,
|
| 1618 |
+
"loss": 1.3494,
|
| 1619 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1620 |
+
"rewards/chosen": -0.07285211980342865,
|
| 1621 |
+
"rewards/margins": 0.28994783759117126,
|
| 1622 |
+
"rewards/rejected": -0.3627999424934387,
|
| 1623 |
+
"step": 980
|
| 1624 |
+
},
|
| 1625 |
+
{
|
| 1626 |
+
"epoch": 0.647693817468106,
|
| 1627 |
+
"grad_norm": 1.40625,
|
| 1628 |
+
"learning_rate": 1.6625669657993483e-06,
|
| 1629 |
+
"logits/chosen": -2.308607578277588,
|
| 1630 |
+
"logits/rejected": -2.228633403778076,
|
| 1631 |
+
"logps/chosen": -1.0984961986541748,
|
| 1632 |
+
"logps/rejected": -2.0189504623413086,
|
| 1633 |
+
"loss": 1.3514,
|
| 1634 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1635 |
+
"rewards/chosen": -0.06574535369873047,
|
| 1636 |
+
"rewards/margins": 0.24618005752563477,
|
| 1637 |
+
"rewards/rejected": -0.31192541122436523,
|
| 1638 |
+
"step": 990
|
| 1639 |
+
},
|
| 1640 |
+
{
|
| 1641 |
+
"epoch": 0.6542361792607131,
|
| 1642 |
+
"grad_norm": 1.0859375,
|
| 1643 |
+
"learning_rate": 1.6089703032168736e-06,
|
| 1644 |
+
"logits/chosen": -2.1339077949523926,
|
| 1645 |
+
"logits/rejected": -2.12742018699646,
|
| 1646 |
+
"logps/chosen": -1.3830599784851074,
|
| 1647 |
+
"logps/rejected": -2.1155436038970947,
|
| 1648 |
+
"loss": 1.3527,
|
| 1649 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1650 |
+
"rewards/chosen": -0.12206210941076279,
|
| 1651 |
+
"rewards/margins": 0.2249002456665039,
|
| 1652 |
+
"rewards/rejected": -0.3469623625278473,
|
| 1653 |
+
"step": 1000
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 0.6542361792607131,
|
| 1657 |
+
"eval_logits/chosen": -2.0770561695098877,
|
| 1658 |
+
"eval_logits/rejected": -1.9978020191192627,
|
| 1659 |
+
"eval_logps/chosen": -1.134178638458252,
|
| 1660 |
+
"eval_logps/rejected": -2.2290825843811035,
|
| 1661 |
+
"eval_loss": 1.3548479080200195,
|
| 1662 |
+
"eval_rewards/accuracies": 0.7049999833106995,
|
| 1663 |
+
"eval_rewards/chosen": -0.06259430944919586,
|
| 1664 |
+
"eval_rewards/margins": 0.28875967860221863,
|
| 1665 |
+
"eval_rewards/rejected": -0.3513540029525757,
|
| 1666 |
+
"eval_runtime": 192.5548,
|
| 1667 |
+
"eval_samples_per_second": 10.387,
|
| 1668 |
+
"eval_steps_per_second": 0.519,
|
| 1669 |
+
"step": 1000
|
| 1670 |
+
},
|
| 1671 |
+
{
|
| 1672 |
+
"epoch": 0.6607785410533202,
|
| 1673 |
+
"grad_norm": 2.125,
|
| 1674 |
+
"learning_rate": 1.55583876344081e-06,
|
| 1675 |
+
"logits/chosen": -2.131939649581909,
|
| 1676 |
+
"logits/rejected": -1.9190185070037842,
|
| 1677 |
+
"logps/chosen": -1.2219233512878418,
|
| 1678 |
+
"logps/rejected": -2.523764133453369,
|
| 1679 |
+
"loss": 1.351,
|
| 1680 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1681 |
+
"rewards/chosen": -0.0856153592467308,
|
| 1682 |
+
"rewards/margins": 0.3270259499549866,
|
| 1683 |
+
"rewards/rejected": -0.4126412868499756,
|
| 1684 |
+
"step": 1010
|
| 1685 |
+
},
|
| 1686 |
+
{
|
| 1687 |
+
"epoch": 0.6673209028459274,
|
| 1688 |
+
"grad_norm": 1.234375,
|
| 1689 |
+
"learning_rate": 1.5032000814512372e-06,
|
| 1690 |
+
"logits/chosen": -2.2159934043884277,
|
| 1691 |
+
"logits/rejected": -2.1700668334960938,
|
| 1692 |
+
"logps/chosen": -1.3070309162139893,
|
| 1693 |
+
"logps/rejected": -2.2616512775421143,
|
| 1694 |
+
"loss": 1.3547,
|
| 1695 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1696 |
+
"rewards/chosen": -0.11680419743061066,
|
| 1697 |
+
"rewards/margins": 0.2500377595424652,
|
| 1698 |
+
"rewards/rejected": -0.36684197187423706,
|
| 1699 |
+
"step": 1020
|
| 1700 |
+
},
|
| 1701 |
+
{
|
| 1702 |
+
"epoch": 0.6738632646385345,
|
| 1703 |
+
"grad_norm": 0.8125,
|
| 1704 |
+
"learning_rate": 1.4510817349535323e-06,
|
| 1705 |
+
"logits/chosen": -2.184436082839966,
|
| 1706 |
+
"logits/rejected": -2.1190972328186035,
|
| 1707 |
+
"logps/chosen": -1.2869107723236084,
|
| 1708 |
+
"logps/rejected": -2.4080114364624023,
|
| 1709 |
+
"loss": 1.3565,
|
| 1710 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1711 |
+
"rewards/chosen": -0.09363288432359695,
|
| 1712 |
+
"rewards/margins": 0.29534098505973816,
|
| 1713 |
+
"rewards/rejected": -0.3889738917350769,
|
| 1714 |
+
"step": 1030
|
| 1715 |
+
},
|
| 1716 |
+
{
|
| 1717 |
+
"epoch": 0.6804056264311417,
|
| 1718 |
+
"grad_norm": 1.5234375,
|
| 1719 |
+
"learning_rate": 1.3995109300348537e-06,
|
| 1720 |
+
"logits/chosen": -2.2407493591308594,
|
| 1721 |
+
"logits/rejected": -2.1420841217041016,
|
| 1722 |
+
"logps/chosen": -1.2200353145599365,
|
| 1723 |
+
"logps/rejected": -2.453103542327881,
|
| 1724 |
+
"loss": 1.3542,
|
| 1725 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1726 |
+
"rewards/chosen": -0.08661921322345734,
|
| 1727 |
+
"rewards/margins": 0.32051774859428406,
|
| 1728 |
+
"rewards/rejected": -0.4071369767189026,
|
| 1729 |
+
"step": 1040
|
| 1730 |
+
},
|
| 1731 |
+
{
|
| 1732 |
+
"epoch": 0.6869479882237488,
|
| 1733 |
+
"grad_norm": 0.97265625,
|
| 1734 |
+
"learning_rate": 1.348514586962389e-06,
|
| 1735 |
+
"logits/chosen": -2.22872257232666,
|
| 1736 |
+
"logits/rejected": -2.131127119064331,
|
| 1737 |
+
"logps/chosen": -1.181624174118042,
|
| 1738 |
+
"logps/rejected": -2.3231348991394043,
|
| 1739 |
+
"loss": 1.3554,
|
| 1740 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1741 |
+
"rewards/chosen": -0.061110030859708786,
|
| 1742 |
+
"rewards/margins": 0.31904760003089905,
|
| 1743 |
+
"rewards/rejected": -0.38015761971473694,
|
| 1744 |
+
"step": 1050
|
| 1745 |
+
},
|
| 1746 |
+
{
|
| 1747 |
+
"epoch": 0.693490350016356,
|
| 1748 |
+
"grad_norm": 0.78125,
|
| 1749 |
+
"learning_rate": 1.2981193261308284e-06,
|
| 1750 |
+
"logits/chosen": -2.1978235244750977,
|
| 1751 |
+
"logits/rejected": -2.145724058151245,
|
| 1752 |
+
"logps/chosen": -1.0791642665863037,
|
| 1753 |
+
"logps/rejected": -2.3014321327209473,
|
| 1754 |
+
"loss": 1.3478,
|
| 1755 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1756 |
+
"rewards/chosen": -0.05117698386311531,
|
| 1757 |
+
"rewards/margins": 0.3088831603527069,
|
| 1758 |
+
"rewards/rejected": -0.3600601851940155,
|
| 1759 |
+
"step": 1060
|
| 1760 |
+
},
|
| 1761 |
+
{
|
| 1762 |
+
"epoch": 0.700032711808963,
|
| 1763 |
+
"grad_norm": 2.78125,
|
| 1764 |
+
"learning_rate": 1.2483514541663501e-06,
|
| 1765 |
+
"logits/chosen": -2.2404794692993164,
|
| 1766 |
+
"logits/rejected": -2.0921406745910645,
|
| 1767 |
+
"logps/chosen": -1.1699206829071045,
|
| 1768 |
+
"logps/rejected": -2.083617687225342,
|
| 1769 |
+
"loss": 1.3563,
|
| 1770 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1771 |
+
"rewards/chosen": -0.08945433795452118,
|
| 1772 |
+
"rewards/margins": 0.25244590640068054,
|
| 1773 |
+
"rewards/rejected": -0.3419002592563629,
|
| 1774 |
+
"step": 1070
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"epoch": 0.7065750736015701,
|
| 1778 |
+
"grad_norm": 0.98828125,
|
| 1779 |
+
"learning_rate": 1.1992369501944096e-06,
|
| 1780 |
+
"logits/chosen": -2.3222763538360596,
|
| 1781 |
+
"logits/rejected": -2.1950221061706543,
|
| 1782 |
+
"logps/chosen": -1.0542547702789307,
|
| 1783 |
+
"logps/rejected": -2.0735292434692383,
|
| 1784 |
+
"loss": 1.362,
|
| 1785 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1786 |
+
"rewards/chosen": -0.04106102138757706,
|
| 1787 |
+
"rewards/margins": 0.2844560146331787,
|
| 1788 |
+
"rewards/rejected": -0.32551708817481995,
|
| 1789 |
+
"step": 1080
|
| 1790 |
+
},
|
| 1791 |
+
{
|
| 1792 |
+
"epoch": 0.7131174353941773,
|
| 1793 |
+
"grad_norm": 1.6328125,
|
| 1794 |
+
"learning_rate": 1.1508014522784803e-06,
|
| 1795 |
+
"logits/chosen": -2.225466251373291,
|
| 1796 |
+
"logits/rejected": -2.114065408706665,
|
| 1797 |
+
"logps/chosen": -1.1339337825775146,
|
| 1798 |
+
"logps/rejected": -1.9321873188018799,
|
| 1799 |
+
"loss": 1.3505,
|
| 1800 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1801 |
+
"rewards/chosen": -0.07440061867237091,
|
| 1802 |
+
"rewards/margins": 0.20720815658569336,
|
| 1803 |
+
"rewards/rejected": -0.2816087603569031,
|
| 1804 |
+
"step": 1090
|
| 1805 |
+
},
|
| 1806 |
+
{
|
| 1807 |
+
"epoch": 0.7196597971867844,
|
| 1808 |
+
"grad_norm": 1.2109375,
|
| 1809 |
+
"learning_rate": 1.1030702440368319e-06,
|
| 1810 |
+
"logits/chosen": -2.1251111030578613,
|
| 1811 |
+
"logits/rejected": -2.109551429748535,
|
| 1812 |
+
"logps/chosen": -1.254265546798706,
|
| 1813 |
+
"logps/rejected": -2.5865750312805176,
|
| 1814 |
+
"loss": 1.3483,
|
| 1815 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1816 |
+
"rewards/chosen": -0.0920356959104538,
|
| 1817 |
+
"rewards/margins": 0.36417168378829956,
|
| 1818 |
+
"rewards/rejected": -0.4562074542045593,
|
| 1819 |
+
"step": 1100
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"epoch": 0.7196597971867844,
|
| 1823 |
+
"eval_logits/chosen": -2.059796094894409,
|
| 1824 |
+
"eval_logits/rejected": -1.9802155494689941,
|
| 1825 |
+
"eval_logps/chosen": -1.1471033096313477,
|
| 1826 |
+
"eval_logps/rejected": -2.3047683238983154,
|
| 1827 |
+
"eval_loss": 1.3557965755462646,
|
| 1828 |
+
"eval_rewards/accuracies": 0.7024999856948853,
|
| 1829 |
+
"eval_rewards/chosen": -0.06647168844938278,
|
| 1830 |
+
"eval_rewards/margins": 0.30758801102638245,
|
| 1831 |
+
"eval_rewards/rejected": -0.37405967712402344,
|
| 1832 |
+
"eval_runtime": 192.5505,
|
| 1833 |
+
"eval_samples_per_second": 10.387,
|
| 1834 |
+
"eval_steps_per_second": 0.519,
|
| 1835 |
+
"step": 1100
|
| 1836 |
+
},
|
| 1837 |
+
{
|
| 1838 |
+
"epoch": 0.7262021589793916,
|
| 1839 |
+
"grad_norm": 1.9609375,
|
| 1840 |
+
"learning_rate": 1.0560682414443315e-06,
|
| 1841 |
+
"logits/chosen": -2.192129135131836,
|
| 1842 |
+
"logits/rejected": -2.050372362136841,
|
| 1843 |
+
"logps/chosen": -1.2834129333496094,
|
| 1844 |
+
"logps/rejected": -2.1301047801971436,
|
| 1845 |
+
"loss": 1.3589,
|
| 1846 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1847 |
+
"rewards/chosen": -0.08568480610847473,
|
| 1848 |
+
"rewards/margins": 0.2297004908323288,
|
| 1849 |
+
"rewards/rejected": -0.31538528203964233,
|
| 1850 |
+
"step": 1110
|
| 1851 |
+
},
|
| 1852 |
+
{
|
| 1853 |
+
"epoch": 0.7327445207719987,
|
| 1854 |
+
"grad_norm": 1.03125,
|
| 1855 |
+
"learning_rate": 1.009819979826156e-06,
|
| 1856 |
+
"logits/chosen": -2.202493190765381,
|
| 1857 |
+
"logits/rejected": -2.130976438522339,
|
| 1858 |
+
"logps/chosen": -1.107555627822876,
|
| 1859 |
+
"logps/rejected": -2.4273107051849365,
|
| 1860 |
+
"loss": 1.3471,
|
| 1861 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1862 |
+
"rewards/chosen": -0.048850033432245255,
|
| 1863 |
+
"rewards/margins": 0.3565898835659027,
|
| 1864 |
+
"rewards/rejected": -0.40544000267982483,
|
| 1865 |
+
"step": 1120
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 0.7392868825646058,
|
| 1869 |
+
"grad_norm": 1.5703125,
|
| 1870 |
+
"learning_rate": 9.643496010502054e-07,
|
| 1871 |
+
"logits/chosen": -2.088268756866455,
|
| 1872 |
+
"logits/rejected": -2.066437244415283,
|
| 1873 |
+
"logps/chosen": -1.4891324043273926,
|
| 1874 |
+
"logps/rejected": -2.3021399974823,
|
| 1875 |
+
"loss": 1.3614,
|
| 1876 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1877 |
+
"rewards/chosen": -0.12404866516590118,
|
| 1878 |
+
"rewards/margins": 0.20802974700927734,
|
| 1879 |
+
"rewards/rejected": -0.33207839727401733,
|
| 1880 |
+
"step": 1130
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 0.745829244357213,
|
| 1884 |
+
"grad_norm": 0.87109375,
|
| 1885 |
+
"learning_rate": 9.196808409249086e-07,
|
| 1886 |
+
"logits/chosen": -2.1621835231781006,
|
| 1887 |
+
"logits/rejected": -2.047921657562256,
|
| 1888 |
+
"logps/chosen": -1.2306039333343506,
|
| 1889 |
+
"logps/rejected": -2.0379228591918945,
|
| 1890 |
+
"loss": 1.3603,
|
| 1891 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1892 |
+
"rewards/chosen": -0.07890699058771133,
|
| 1893 |
+
"rewards/margins": 0.23432865738868713,
|
| 1894 |
+
"rewards/rejected": -0.3132356107234955,
|
| 1895 |
+
"step": 1140
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 0.7523716061498201,
|
| 1899 |
+
"grad_norm": 1.390625,
|
| 1900 |
+
"learning_rate": 8.758370168089797e-07,
|
| 1901 |
+
"logits/chosen": -2.1358230113983154,
|
| 1902 |
+
"logits/rejected": -2.1172406673431396,
|
| 1903 |
+
"logps/chosen": -1.055590271949768,
|
| 1904 |
+
"logps/rejected": -2.3818840980529785,
|
| 1905 |
+
"loss": 1.3584,
|
| 1906 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1907 |
+
"rewards/chosen": -0.07134061306715012,
|
| 1908 |
+
"rewards/margins": 0.3324960172176361,
|
| 1909 |
+
"rewards/rejected": -0.4038366377353668,
|
| 1910 |
+
"step": 1150
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"epoch": 0.7589139679424273,
|
| 1914 |
+
"grad_norm": 2.09375,
|
| 1915 |
+
"learning_rate": 8.328410154396318e-07,
|
| 1916 |
+
"logits/chosen": -2.1972057819366455,
|
| 1917 |
+
"logits/rejected": -2.175696849822998,
|
| 1918 |
+
"logps/chosen": -1.1727460622787476,
|
| 1919 |
+
"logps/rejected": -2.109738349914551,
|
| 1920 |
+
"loss": 1.3503,
|
| 1921 |
+
"rewards/accuracies": 0.6875,
|
| 1922 |
+
"rewards/chosen": -0.048242855817079544,
|
| 1923 |
+
"rewards/margins": 0.2853914201259613,
|
| 1924 |
+
"rewards/rejected": -0.33363425731658936,
|
| 1925 |
+
"step": 1160
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 0.7654563297350343,
|
| 1929 |
+
"grad_norm": 1.3203125,
|
| 1930 |
+
"learning_rate": 7.907152809855529e-07,
|
| 1931 |
+
"logits/chosen": -2.1163055896759033,
|
| 1932 |
+
"logits/rejected": -2.045872926712036,
|
| 1933 |
+
"logps/chosen": -1.347551941871643,
|
| 1934 |
+
"logps/rejected": -2.065335750579834,
|
| 1935 |
+
"loss": 1.3635,
|
| 1936 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1937 |
+
"rewards/chosen": -0.12098999321460724,
|
| 1938 |
+
"rewards/margins": 0.18405140936374664,
|
| 1939 |
+
"rewards/rejected": -0.3050413727760315,
|
| 1940 |
+
"step": 1170
|
| 1941 |
+
},
|
| 1942 |
+
{
|
| 1943 |
+
"epoch": 0.7719986915276414,
|
| 1944 |
+
"grad_norm": 1.875,
|
| 1945 |
+
"learning_rate": 7.494818033309207e-07,
|
| 1946 |
+
"logits/chosen": -2.158379554748535,
|
| 1947 |
+
"logits/rejected": -2.000858783721924,
|
| 1948 |
+
"logps/chosen": -1.1294301748275757,
|
| 1949 |
+
"logps/rejected": -1.8507928848266602,
|
| 1950 |
+
"loss": 1.355,
|
| 1951 |
+
"rewards/accuracies": 0.637499988079071,
|
| 1952 |
+
"rewards/chosen": -0.06103038787841797,
|
| 1953 |
+
"rewards/margins": 0.1857200264930725,
|
| 1954 |
+
"rewards/rejected": -0.24675039947032928,
|
| 1955 |
+
"step": 1180
|
| 1956 |
+
},
|
| 1957 |
+
{
|
| 1958 |
+
"epoch": 0.7785410533202486,
|
| 1959 |
+
"grad_norm": 0.6953125,
|
| 1960 |
+
"learning_rate": 7.091621065965521e-07,
|
| 1961 |
+
"logits/chosen": -2.1394495964050293,
|
| 1962 |
+
"logits/rejected": -2.115633726119995,
|
| 1963 |
+
"logps/chosen": -1.1215696334838867,
|
| 1964 |
+
"logps/rejected": -2.1693038940429688,
|
| 1965 |
+
"loss": 1.3518,
|
| 1966 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1967 |
+
"rewards/chosen": -0.060358185321092606,
|
| 1968 |
+
"rewards/margins": 0.26639682054519653,
|
| 1969 |
+
"rewards/rejected": -0.32675498723983765,
|
| 1970 |
+
"step": 1190
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"epoch": 0.7850834151128557,
|
| 1974 |
+
"grad_norm": 1.1328125,
|
| 1975 |
+
"learning_rate": 6.697772379041823e-07,
|
| 1976 |
+
"logits/chosen": -2.1604604721069336,
|
| 1977 |
+
"logits/rejected": -2.116100788116455,
|
| 1978 |
+
"logps/chosen": -1.2930102348327637,
|
| 1979 |
+
"logps/rejected": -2.0868096351623535,
|
| 1980 |
+
"loss": 1.3558,
|
| 1981 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1982 |
+
"rewards/chosen": -0.0996454581618309,
|
| 1983 |
+
"rewards/margins": 0.2223787009716034,
|
| 1984 |
+
"rewards/rejected": -0.3220241367816925,
|
| 1985 |
+
"step": 1200
|
| 1986 |
+
},
|
| 1987 |
+
{
|
| 1988 |
+
"epoch": 0.7850834151128557,
|
| 1989 |
+
"eval_logits/chosen": -2.052238702774048,
|
| 1990 |
+
"eval_logits/rejected": -1.9719146490097046,
|
| 1991 |
+
"eval_logps/chosen": -1.1348469257354736,
|
| 1992 |
+
"eval_logps/rejected": -2.2732903957366943,
|
| 1993 |
+
"eval_loss": 1.3541505336761475,
|
| 1994 |
+
"eval_rewards/accuracies": 0.7049999833106995,
|
| 1995 |
+
"eval_rewards/chosen": -0.0627947449684143,
|
| 1996 |
+
"eval_rewards/margins": 0.30182158946990967,
|
| 1997 |
+
"eval_rewards/rejected": -0.3646163046360016,
|
| 1998 |
+
"eval_runtime": 192.5468,
|
| 1999 |
+
"eval_samples_per_second": 10.387,
|
| 2000 |
+
"eval_steps_per_second": 0.519,
|
| 2001 |
+
"step": 1200
|
| 2002 |
+
},
|
| 2003 |
+
{
|
| 2004 |
+
"epoch": 0.7916257769054629,
|
| 2005 |
+
"grad_norm": 2.140625,
|
| 2006 |
+
"learning_rate": 6.313477563897466e-07,
|
| 2007 |
+
"logits/chosen": -2.198477268218994,
|
| 2008 |
+
"logits/rejected": -2.16310453414917,
|
| 2009 |
+
"logps/chosen": -1.0447008609771729,
|
| 2010 |
+
"logps/rejected": -2.3631505966186523,
|
| 2011 |
+
"loss": 1.3481,
|
| 2012 |
+
"rewards/accuracies": 0.762499988079071,
|
| 2013 |
+
"rewards/chosen": -0.05016545578837395,
|
| 2014 |
+
"rewards/margins": 0.3603706955909729,
|
| 2015 |
+
"rewards/rejected": -0.4105361998081207,
|
| 2016 |
+
"step": 1210
|
| 2017 |
+
},
|
| 2018 |
+
{
|
| 2019 |
+
"epoch": 0.79816813869807,
|
| 2020 |
+
"grad_norm": 0.7890625,
|
| 2021 |
+
"learning_rate": 5.9389372247138e-07,
|
| 2022 |
+
"logits/chosen": -2.1831748485565186,
|
| 2023 |
+
"logits/rejected": -2.049494504928589,
|
| 2024 |
+
"logps/chosen": -1.0988541841506958,
|
| 2025 |
+
"logps/rejected": -2.0942153930664062,
|
| 2026 |
+
"loss": 1.3416,
|
| 2027 |
+
"rewards/accuracies": 0.737500011920929,
|
| 2028 |
+
"rewards/chosen": -0.08485239744186401,
|
| 2029 |
+
"rewards/margins": 0.2300022840499878,
|
| 2030 |
+
"rewards/rejected": -0.3148546814918518,
|
| 2031 |
+
"step": 1220
|
| 2032 |
+
},
|
| 2033 |
+
{
|
| 2034 |
+
"epoch": 0.8047105004906772,
|
| 2035 |
+
"grad_norm": 1.578125,
|
| 2036 |
+
"learning_rate": 5.574346873777714e-07,
|
| 2037 |
+
"logits/chosen": -2.1483089923858643,
|
| 2038 |
+
"logits/rejected": -2.146624803543091,
|
| 2039 |
+
"logps/chosen": -1.1647026538848877,
|
| 2040 |
+
"logps/rejected": -2.1079747676849365,
|
| 2041 |
+
"loss": 1.3511,
|
| 2042 |
+
"rewards/accuracies": 0.762499988079071,
|
| 2043 |
+
"rewards/chosen": -0.05951924994587898,
|
| 2044 |
+
"rewards/margins": 0.2496817409992218,
|
| 2045 |
+
"rewards/rejected": -0.3092009723186493,
|
| 2046 |
+
"step": 1230
|
| 2047 |
+
},
|
| 2048 |
+
{
|
| 2049 |
+
"epoch": 0.8112528622832843,
|
| 2050 |
+
"grad_norm": 1.703125,
|
| 2051 |
+
"learning_rate": 5.219896829422927e-07,
|
| 2052 |
+
"logits/chosen": -2.1868844032287598,
|
| 2053 |
+
"logits/rejected": -2.071396589279175,
|
| 2054 |
+
"logps/chosen": -1.1010544300079346,
|
| 2055 |
+
"logps/rejected": -2.396719455718994,
|
| 2056 |
+
"loss": 1.3481,
|
| 2057 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 2058 |
+
"rewards/chosen": -0.06258542090654373,
|
| 2059 |
+
"rewards/margins": 0.3550402522087097,
|
| 2060 |
+
"rewards/rejected": -0.41762566566467285,
|
| 2061 |
+
"step": 1240
|
| 2062 |
+
},
|
| 2063 |
+
{
|
| 2064 |
+
"epoch": 0.8177952240758914,
|
| 2065 |
+
"grad_norm": 0.8984375,
|
| 2066 |
+
"learning_rate": 4.875772116682817e-07,
|
| 2067 |
+
"logits/chosen": -2.044847249984741,
|
| 2068 |
+
"logits/rejected": -1.9548499584197998,
|
| 2069 |
+
"logps/chosen": -1.1090381145477295,
|
| 2070 |
+
"logps/rejected": -2.3034088611602783,
|
| 2071 |
+
"loss": 1.3592,
|
| 2072 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 2073 |
+
"rewards/chosen": -0.07034337520599365,
|
| 2074 |
+
"rewards/margins": 0.2901187539100647,
|
| 2075 |
+
"rewards/rejected": -0.36046212911605835,
|
| 2076 |
+
"step": 1250
|
| 2077 |
+
},
|
| 2078 |
+
{
|
| 2079 |
+
"epoch": 0.8243375858684985,
|
| 2080 |
+
"grad_norm": 1.671875,
|
| 2081 |
+
"learning_rate": 4.542152370706149e-07,
|
| 2082 |
+
"logits/chosen": -2.072751522064209,
|
| 2083 |
+
"logits/rejected": -1.9732776880264282,
|
| 2084 |
+
"logps/chosen": -1.3094862699508667,
|
| 2085 |
+
"logps/rejected": -2.3592586517333984,
|
| 2086 |
+
"loss": 1.3592,
|
| 2087 |
+
"rewards/accuracies": 0.637499988079071,
|
| 2088 |
+
"rewards/chosen": -0.12541761994361877,
|
| 2089 |
+
"rewards/margins": 0.2671012580394745,
|
| 2090 |
+
"rewards/rejected": -0.3925188481807709,
|
| 2091 |
+
"step": 1260
|
| 2092 |
+
},
|
| 2093 |
+
{
|
| 2094 |
+
"epoch": 0.8308799476611056,
|
| 2095 |
+
"grad_norm": 0.98046875,
|
| 2096 |
+
"learning_rate": 4.2192117429865067e-07,
|
| 2097 |
+
"logits/chosen": -1.9732551574707031,
|
| 2098 |
+
"logits/rejected": -1.8982023000717163,
|
| 2099 |
+
"logps/chosen": -1.216827630996704,
|
| 2100 |
+
"logps/rejected": -2.2013864517211914,
|
| 2101 |
+
"loss": 1.3502,
|
| 2102 |
+
"rewards/accuracies": 0.675000011920929,
|
| 2103 |
+
"rewards/chosen": -0.07743276655673981,
|
| 2104 |
+
"rewards/margins": 0.25767821073532104,
|
| 2105 |
+
"rewards/rejected": -0.33511093258857727,
|
| 2106 |
+
"step": 1270
|
| 2107 |
+
},
|
| 2108 |
+
{
|
| 2109 |
+
"epoch": 0.8374223094537128,
|
| 2110 |
+
"grad_norm": 4.125,
|
| 2111 |
+
"learning_rate": 3.907118810454172e-07,
|
| 2112 |
+
"logits/chosen": -2.127220869064331,
|
| 2113 |
+
"logits/rejected": -2.117337703704834,
|
| 2114 |
+
"logps/chosen": -1.1793607473373413,
|
| 2115 |
+
"logps/rejected": -2.3696749210357666,
|
| 2116 |
+
"loss": 1.3544,
|
| 2117 |
+
"rewards/accuracies": 0.800000011920929,
|
| 2118 |
+
"rewards/chosen": -0.07967950403690338,
|
| 2119 |
+
"rewards/margins": 0.31056874990463257,
|
| 2120 |
+
"rewards/rejected": -0.39024823904037476,
|
| 2121 |
+
"step": 1280
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 0.8439646712463199,
|
| 2125 |
+
"grad_norm": 1.578125,
|
| 2126 |
+
"learning_rate": 3.6060364874779455e-07,
|
| 2127 |
+
"logits/chosen": -2.1260552406311035,
|
| 2128 |
+
"logits/rejected": -2.1024839878082275,
|
| 2129 |
+
"logps/chosen": -1.387900471687317,
|
| 2130 |
+
"logps/rejected": -2.1651806831359863,
|
| 2131 |
+
"loss": 1.3659,
|
| 2132 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 2133 |
+
"rewards/chosen": -0.10205866396427155,
|
| 2134 |
+
"rewards/margins": 0.2208395004272461,
|
| 2135 |
+
"rewards/rejected": -0.32289814949035645,
|
| 2136 |
+
"step": 1290
|
| 2137 |
+
},
|
| 2138 |
+
{
|
| 2139 |
+
"epoch": 0.850507033038927,
|
| 2140 |
+
"grad_norm": 0.76171875,
|
| 2141 |
+
"learning_rate": 3.3161219408229026e-07,
|
| 2142 |
+
"logits/chosen": -2.2469534873962402,
|
| 2143 |
+
"logits/rejected": -2.171238422393799,
|
| 2144 |
+
"logps/chosen": -1.35056471824646,
|
| 2145 |
+
"logps/rejected": -2.452230453491211,
|
| 2146 |
+
"loss": 1.3515,
|
| 2147 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 2148 |
+
"rewards/chosen": -0.1100931391119957,
|
| 2149 |
+
"rewards/margins": 0.2768345773220062,
|
| 2150 |
+
"rewards/rejected": -0.3869277536869049,
|
| 2151 |
+
"step": 1300
|
| 2152 |
+
},
|
| 2153 |
+
{
|
| 2154 |
+
"epoch": 0.850507033038927,
|
| 2155 |
+
"eval_logits/chosen": -2.0505053997039795,
|
| 2156 |
+
"eval_logits/rejected": -1.9694212675094604,
|
| 2157 |
+
"eval_logps/chosen": -1.1402031183242798,
|
| 2158 |
+
"eval_logps/rejected": -2.2917935848236084,
|
| 2159 |
+
"eval_loss": 1.3543208837509155,
|
| 2160 |
+
"eval_rewards/accuracies": 0.7049999833106995,
|
| 2161 |
+
"eval_rewards/chosen": -0.06440159678459167,
|
| 2162 |
+
"eval_rewards/margins": 0.30576565861701965,
|
| 2163 |
+
"eval_rewards/rejected": -0.3701672852039337,
|
| 2164 |
+
"eval_runtime": 192.512,
|
| 2165 |
+
"eval_samples_per_second": 10.389,
|
| 2166 |
+
"eval_steps_per_second": 0.519,
|
| 2167 |
+
"step": 1300
|
| 2168 |
+
},
|
| 2169 |
+
{
|
| 2170 |
+
"epoch": 0.8570493948315342,
|
| 2171 |
+
"grad_norm": 1.1640625,
|
| 2172 |
+
"learning_rate": 3.0375265076083796e-07,
|
| 2173 |
+
"logits/chosen": -2.238908529281616,
|
| 2174 |
+
"logits/rejected": -2.219423770904541,
|
| 2175 |
+
"logps/chosen": -1.4968189001083374,
|
| 2176 |
+
"logps/rejected": -1.8981587886810303,
|
| 2177 |
+
"loss": 1.3654,
|
| 2178 |
+
"rewards/accuracies": 0.6875,
|
| 2179 |
+
"rewards/chosen": -0.1579573005437851,
|
| 2180 |
+
"rewards/margins": 0.11951179802417755,
|
| 2181 |
+
"rewards/rejected": -0.27746909856796265,
|
| 2182 |
+
"step": 1310
|
| 2183 |
+
},
|
| 2184 |
+
{
|
| 2185 |
+
"epoch": 0.8635917566241413,
|
| 2186 |
+
"grad_norm": 1.234375,
|
| 2187 |
+
"learning_rate": 2.7703956163091153e-07,
|
| 2188 |
+
"logits/chosen": -2.2014071941375732,
|
| 2189 |
+
"logits/rejected": -2.1226069927215576,
|
| 2190 |
+
"logps/chosen": -1.133750081062317,
|
| 2191 |
+
"logps/rejected": -2.1364989280700684,
|
| 2192 |
+
"loss": 1.3546,
|
| 2193 |
+
"rewards/accuracies": 0.675000011920929,
|
| 2194 |
+
"rewards/chosen": -0.06972350180149078,
|
| 2195 |
+
"rewards/margins": 0.25297701358795166,
|
| 2196 |
+
"rewards/rejected": -0.32270053029060364,
|
| 2197 |
+
"step": 1320
|
| 2198 |
+
},
|
| 2199 |
+
{
|
| 2200 |
+
"epoch": 0.8701341184167485,
|
| 2201 |
+
"grad_norm": 0.58203125,
|
| 2202 |
+
"learning_rate": 2.514868710840723e-07,
|
| 2203 |
+
"logits/chosen": -2.192532777786255,
|
| 2204 |
+
"logits/rejected": -2.1196610927581787,
|
| 2205 |
+
"logps/chosen": -1.2276318073272705,
|
| 2206 |
+
"logps/rejected": -1.9988428354263306,
|
| 2207 |
+
"loss": 1.3551,
|
| 2208 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 2209 |
+
"rewards/chosen": -0.061108749359846115,
|
| 2210 |
+
"rewards/margins": 0.2282850742340088,
|
| 2211 |
+
"rewards/rejected": -0.2893938422203064,
|
| 2212 |
+
"step": 1330
|
| 2213 |
+
},
|
| 2214 |
+
{
|
| 2215 |
+
"epoch": 0.8766764802093556,
|
| 2216 |
+
"grad_norm": 1.65625,
|
| 2217 |
+
"learning_rate": 2.271079177769117e-07,
|
| 2218 |
+
"logits/chosen": -2.122554302215576,
|
| 2219 |
+
"logits/rejected": -2.1153998374938965,
|
| 2220 |
+
"logps/chosen": -1.290633201599121,
|
| 2221 |
+
"logps/rejected": -2.0642950534820557,
|
| 2222 |
+
"loss": 1.3459,
|
| 2223 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 2224 |
+
"rewards/chosen": -0.09484368562698364,
|
| 2225 |
+
"rewards/margins": 0.19020384550094604,
|
| 2226 |
+
"rewards/rejected": -0.2850475609302521,
|
| 2227 |
+
"step": 1340
|
| 2228 |
+
},
|
| 2229 |
+
{
|
| 2230 |
+
"epoch": 0.8832188420019627,
|
| 2231 |
+
"grad_norm": 1.3125,
|
| 2232 |
+
"learning_rate": 2.0391542766819456e-07,
|
| 2233 |
+
"logits/chosen": -2.1564154624938965,
|
| 2234 |
+
"logits/rejected": -1.9759187698364258,
|
| 2235 |
+
"logps/chosen": -1.3263826370239258,
|
| 2236 |
+
"logps/rejected": -2.2100300788879395,
|
| 2237 |
+
"loss": 1.3491,
|
| 2238 |
+
"rewards/accuracies": 0.675000011920929,
|
| 2239 |
+
"rewards/chosen": -0.10235595703125,
|
| 2240 |
+
"rewards/margins": 0.22581568360328674,
|
| 2241 |
+
"rewards/rejected": -0.32817164063453674,
|
| 2242 |
+
"step": 1350
|
| 2243 |
+
},
|
| 2244 |
+
{
|
| 2245 |
+
"epoch": 0.8897612037945698,
|
| 2246 |
+
"grad_norm": 1.515625,
|
| 2247 |
+
"learning_rate": 1.8192150737583264e-07,
|
| 2248 |
+
"logits/chosen": -2.180866241455078,
|
| 2249 |
+
"logits/rejected": -2.0520682334899902,
|
| 2250 |
+
"logps/chosen": -1.0884932279586792,
|
| 2251 |
+
"logps/rejected": -2.1588473320007324,
|
| 2252 |
+
"loss": 1.3511,
|
| 2253 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 2254 |
+
"rewards/chosen": -0.04868602380156517,
|
| 2255 |
+
"rewards/margins": 0.29052504897117615,
|
| 2256 |
+
"rewards/rejected": -0.33921104669570923,
|
| 2257 |
+
"step": 1360
|
| 2258 |
+
},
|
| 2259 |
+
{
|
| 2260 |
+
"epoch": 0.8963035655871769,
|
| 2261 |
+
"grad_norm": 1.7421875,
|
| 2262 |
+
"learning_rate": 1.61137637857158e-07,
|
| 2263 |
+
"logits/chosen": -2.253401517868042,
|
| 2264 |
+
"logits/rejected": -2.1362268924713135,
|
| 2265 |
+
"logps/chosen": -1.2787913084030151,
|
| 2266 |
+
"logps/rejected": -1.7992349863052368,
|
| 2267 |
+
"loss": 1.3467,
|
| 2268 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 2269 |
+
"rewards/chosen": -0.0960894376039505,
|
| 2270 |
+
"rewards/margins": 0.15551379323005676,
|
| 2271 |
+
"rewards/rejected": -0.25160321593284607,
|
| 2272 |
+
"step": 1370
|
| 2273 |
+
},
|
| 2274 |
+
{
|
| 2275 |
+
"epoch": 0.9028459273797841,
|
| 2276 |
+
"grad_norm": 0.8671875,
|
| 2277 |
+
"learning_rate": 1.415746684157951e-07,
|
| 2278 |
+
"logits/chosen": -2.173496723175049,
|
| 2279 |
+
"logits/rejected": -2.1127450466156006,
|
| 2280 |
+
"logps/chosen": -0.9452150464057922,
|
| 2281 |
+
"logps/rejected": -1.9824330806732178,
|
| 2282 |
+
"loss": 1.34,
|
| 2283 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 2284 |
+
"rewards/chosen": -0.03502793237566948,
|
| 2285 |
+
"rewards/margins": 0.2680448889732361,
|
| 2286 |
+
"rewards/rejected": -0.3030727803707123,
|
| 2287 |
+
"step": 1380
|
| 2288 |
+
},
|
| 2289 |
+
{
|
| 2290 |
+
"epoch": 0.9093882891723912,
|
| 2291 |
+
"grad_norm": 1.890625,
|
| 2292 |
+
"learning_rate": 1.232428110382586e-07,
|
| 2293 |
+
"logits/chosen": -2.2799248695373535,
|
| 2294 |
+
"logits/rejected": -2.1891300678253174,
|
| 2295 |
+
"logps/chosen": -1.111943006515503,
|
| 2296 |
+
"logps/rejected": -2.683452606201172,
|
| 2297 |
+
"loss": 1.3477,
|
| 2298 |
+
"rewards/accuracies": 0.75,
|
| 2299 |
+
"rewards/chosen": -0.05621108412742615,
|
| 2300 |
+
"rewards/margins": 0.4199764132499695,
|
| 2301 |
+
"rewards/rejected": -0.47618746757507324,
|
| 2302 |
+
"step": 1390
|
| 2303 |
+
},
|
| 2304 |
+
{
|
| 2305 |
+
"epoch": 0.9159306509649984,
|
| 2306 |
+
"grad_norm": 1.4921875,
|
| 2307 |
+
"learning_rate": 1.0615163506323856e-07,
|
| 2308 |
+
"logits/chosen": -2.187314510345459,
|
| 2309 |
+
"logits/rejected": -2.100705623626709,
|
| 2310 |
+
"logps/chosen": -1.0987522602081299,
|
| 2311 |
+
"logps/rejected": -2.2279646396636963,
|
| 2312 |
+
"loss": 1.3572,
|
| 2313 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 2314 |
+
"rewards/chosen": -0.07077629119157791,
|
| 2315 |
+
"rewards/margins": 0.282015860080719,
|
| 2316 |
+
"rewards/rejected": -0.3527921736240387,
|
| 2317 |
+
"step": 1400
|
| 2318 |
+
},
|
| 2319 |
+
{
|
| 2320 |
+
"epoch": 0.9159306509649984,
|
| 2321 |
+
"eval_logits/chosen": -2.0522220134735107,
|
| 2322 |
+
"eval_logits/rejected": -1.9716111421585083,
|
| 2323 |
+
"eval_logps/chosen": -1.1385436058044434,
|
| 2324 |
+
"eval_logps/rejected": -2.282519817352295,
|
| 2325 |
+
"eval_loss": 1.3540371656417847,
|
| 2326 |
+
"eval_rewards/accuracies": 0.7074999809265137,
|
| 2327 |
+
"eval_rewards/chosen": -0.0639037936925888,
|
| 2328 |
+
"eval_rewards/margins": 0.3034813106060028,
|
| 2329 |
+
"eval_rewards/rejected": -0.3673850893974304,
|
| 2330 |
+
"eval_runtime": 192.5536,
|
| 2331 |
+
"eval_samples_per_second": 10.387,
|
| 2332 |
+
"eval_steps_per_second": 0.519,
|
| 2333 |
+
"step": 1400
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 0.9224730127576055,
|
| 2337 |
+
"grad_norm": 1.34375,
|
| 2338 |
+
"learning_rate": 9.031006218634342e-08,
|
| 2339 |
+
"logits/chosen": -2.1624093055725098,
|
| 2340 |
+
"logits/rejected": -2.102839708328247,
|
| 2341 |
+
"logps/chosen": -1.2127426862716675,
|
| 2342 |
+
"logps/rejected": -2.093308210372925,
|
| 2343 |
+
"loss": 1.3509,
|
| 2344 |
+
"rewards/accuracies": 0.737500011920929,
|
| 2345 |
+
"rewards/chosen": -0.07421986013650894,
|
| 2346 |
+
"rewards/margins": 0.24446694552898407,
|
| 2347 |
+
"rewards/rejected": -0.3186867833137512,
|
| 2348 |
+
"step": 1410
|
| 2349 |
+
},
|
| 2350 |
+
{
|
| 2351 |
+
"epoch": 0.9290153745502127,
|
| 2352 |
+
"grad_norm": 0.734375,
|
| 2353 |
+
"learning_rate": 7.572636180292831e-08,
|
| 2354 |
+
"logits/chosen": -2.111999273300171,
|
| 2355 |
+
"logits/rejected": -2.043356418609619,
|
| 2356 |
+
"logps/chosen": -1.2025669813156128,
|
| 2357 |
+
"logps/rejected": -2.113124370574951,
|
| 2358 |
+
"loss": 1.3489,
|
| 2359 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 2360 |
+
"rewards/chosen": -0.08289094269275665,
|
| 2361 |
+
"rewards/margins": 0.25174033641815186,
|
| 2362 |
+
"rewards/rejected": -0.3346312642097473,
|
| 2363 |
+
"step": 1420
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 0.9355577363428198,
|
| 2367 |
+
"grad_norm": 0.8984375,
|
| 2368 |
+
"learning_rate": 6.240814669141559e-08,
|
| 2369 |
+
"logits/chosen": -2.1452536582946777,
|
| 2370 |
+
"logits/rejected": -2.118813991546631,
|
| 2371 |
+
"logps/chosen": -1.1873977184295654,
|
| 2372 |
+
"logps/rejected": -2.4279136657714844,
|
| 2373 |
+
"loss": 1.3462,
|
| 2374 |
+
"rewards/accuracies": 0.6875,
|
| 2375 |
+
"rewards/chosen": -0.07043514400720596,
|
| 2376 |
+
"rewards/margins": 0.3395000398159027,
|
| 2377 |
+
"rewards/rejected": -0.4099351763725281,
|
| 2378 |
+
"step": 1430
|
| 2379 |
+
},
|
| 2380 |
+
{
|
| 2381 |
+
"epoch": 0.9421000981354269,
|
| 2382 |
+
"grad_norm": 2.34375,
|
| 2383 |
+
"learning_rate": 5.036236903938285e-08,
|
| 2384 |
+
"logits/chosen": -2.224623918533325,
|
| 2385 |
+
"logits/rejected": -2.127098560333252,
|
| 2386 |
+
"logps/chosen": -1.1137665510177612,
|
| 2387 |
+
"logps/rejected": -1.9331905841827393,
|
| 2388 |
+
"loss": 1.348,
|
| 2389 |
+
"rewards/accuracies": 0.675000011920929,
|
| 2390 |
+
"rewards/chosen": -0.05770399048924446,
|
| 2391 |
+
"rewards/margins": 0.2274278849363327,
|
| 2392 |
+
"rewards/rejected": -0.28513190150260925,
|
| 2393 |
+
"step": 1440
|
| 2394 |
+
},
|
| 2395 |
+
{
|
| 2396 |
+
"epoch": 0.948642459928034,
|
| 2397 |
+
"grad_norm": 0.80078125,
|
| 2398 |
+
"learning_rate": 3.959531681447859e-08,
|
| 2399 |
+
"logits/chosen": -2.0762100219726562,
|
| 2400 |
+
"logits/rejected": -2.0221409797668457,
|
| 2401 |
+
"logps/chosen": -1.2568187713623047,
|
| 2402 |
+
"logps/rejected": -2.275439739227295,
|
| 2403 |
+
"loss": 1.3529,
|
| 2404 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 2405 |
+
"rewards/chosen": -0.08186222612857819,
|
| 2406 |
+
"rewards/margins": 0.28715780377388,
|
| 2407 |
+
"rewards/rejected": -0.3690200448036194,
|
| 2408 |
+
"step": 1450
|
| 2409 |
+
},
|
| 2410 |
+
{
|
| 2411 |
+
"epoch": 0.9551848217206411,
|
| 2412 |
+
"grad_norm": 0.83203125,
|
| 2413 |
+
"learning_rate": 3.0112610482064544e-08,
|
| 2414 |
+
"logits/chosen": -2.2796406745910645,
|
| 2415 |
+
"logits/rejected": -2.173074722290039,
|
| 2416 |
+
"logps/chosen": -1.1611659526824951,
|
| 2417 |
+
"logps/rejected": -2.0751075744628906,
|
| 2418 |
+
"loss": 1.3478,
|
| 2419 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 2420 |
+
"rewards/chosen": -0.07703422009944916,
|
| 2421 |
+
"rewards/margins": 0.24533399939537048,
|
| 2422 |
+
"rewards/rejected": -0.32236820459365845,
|
| 2423 |
+
"step": 1460
|
| 2424 |
+
},
|
| 2425 |
+
{
|
| 2426 |
+
"epoch": 0.9617271835132483,
|
| 2427 |
+
"grad_norm": 0.77734375,
|
| 2428 |
+
"learning_rate": 2.1919200071301715e-08,
|
| 2429 |
+
"logits/chosen": -2.3272135257720947,
|
| 2430 |
+
"logits/rejected": -2.1798338890075684,
|
| 2431 |
+
"logps/chosen": -1.190741777420044,
|
| 2432 |
+
"logps/rejected": -2.2400176525115967,
|
| 2433 |
+
"loss": 1.3635,
|
| 2434 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 2435 |
+
"rewards/chosen": -0.07553628087043762,
|
| 2436 |
+
"rewards/margins": 0.28841260075569153,
|
| 2437 |
+
"rewards/rejected": -0.36394888162612915,
|
| 2438 |
+
"step": 1470
|
| 2439 |
+
},
|
| 2440 |
+
{
|
| 2441 |
+
"epoch": 0.9682695453058554,
|
| 2442 |
+
"grad_norm": 0.90625,
|
| 2443 |
+
"learning_rate": 1.50193625912029e-08,
|
| 2444 |
+
"logits/chosen": -2.059523105621338,
|
| 2445 |
+
"logits/rejected": -2.071715831756592,
|
| 2446 |
+
"logps/chosen": -1.2252368927001953,
|
| 2447 |
+
"logps/rejected": -2.461535692214966,
|
| 2448 |
+
"loss": 1.3431,
|
| 2449 |
+
"rewards/accuracies": 0.737500011920929,
|
| 2450 |
+
"rewards/chosen": -0.08799092471599579,
|
| 2451 |
+
"rewards/margins": 0.312566339969635,
|
| 2452 |
+
"rewards/rejected": -0.4005572199821472,
|
| 2453 |
+
"step": 1480
|
| 2454 |
+
},
|
| 2455 |
+
{
|
| 2456 |
+
"epoch": 0.9748119070984625,
|
| 2457 |
+
"grad_norm": 1.421875,
|
| 2458 |
+
"learning_rate": 9.416699798010521e-09,
|
| 2459 |
+
"logits/chosen": -2.070498466491699,
|
| 2460 |
+
"logits/rejected": -1.9673669338226318,
|
| 2461 |
+
"logps/chosen": -1.2495880126953125,
|
| 2462 |
+
"logps/rejected": -2.9194693565368652,
|
| 2463 |
+
"loss": 1.3353,
|
| 2464 |
+
"rewards/accuracies": 0.8125,
|
| 2465 |
+
"rewards/chosen": -0.07810261845588684,
|
| 2466 |
+
"rewards/margins": 0.4632912278175354,
|
| 2467 |
+
"rewards/rejected": -0.5413938164710999,
|
| 2468 |
+
"step": 1490
|
| 2469 |
+
},
|
| 2470 |
+
{
|
| 2471 |
+
"epoch": 0.9813542688910697,
|
| 2472 |
+
"grad_norm": 2.125,
|
| 2473 |
+
"learning_rate": 5.114136315058083e-09,
|
| 2474 |
+
"logits/chosen": -2.101090908050537,
|
| 2475 |
+
"logits/rejected": -2.0724215507507324,
|
| 2476 |
+
"logps/chosen": -1.1968806982040405,
|
| 2477 |
+
"logps/rejected": -2.266988754272461,
|
| 2478 |
+
"loss": 1.3527,
|
| 2479 |
+
"rewards/accuracies": 0.737500011920929,
|
| 2480 |
+
"rewards/chosen": -0.10394704341888428,
|
| 2481 |
+
"rewards/margins": 0.25496524572372437,
|
| 2482 |
+
"rewards/rejected": -0.35891225934028625,
|
| 2483 |
+
"step": 1500
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 0.9813542688910697,
|
| 2487 |
+
"eval_logits/chosen": -2.051304817199707,
|
| 2488 |
+
"eval_logits/rejected": -1.9704300165176392,
|
| 2489 |
+
"eval_logps/chosen": -1.1379503011703491,
|
| 2490 |
+
"eval_logps/rejected": -2.283447027206421,
|
| 2491 |
+
"eval_loss": 1.3540750741958618,
|
| 2492 |
+
"eval_rewards/accuracies": 0.7024999856948853,
|
| 2493 |
+
"eval_rewards/chosen": -0.06372576951980591,
|
| 2494 |
+
"eval_rewards/margins": 0.30393749475479126,
|
| 2495 |
+
"eval_rewards/rejected": -0.36766326427459717,
|
| 2496 |
+
"eval_runtime": 192.5377,
|
| 2497 |
+
"eval_samples_per_second": 10.388,
|
| 2498 |
+
"eval_steps_per_second": 0.519,
|
| 2499 |
+
"step": 1500
|
| 2500 |
+
},
|
| 2501 |
+
{
|
| 2502 |
+
"epoch": 0.9878966306836768,
|
| 2503 |
+
"grad_norm": 1.40625,
|
| 2504 |
+
"learning_rate": 2.113918106098345e-09,
|
| 2505 |
+
"logits/chosen": -2.1591796875,
|
| 2506 |
+
"logits/rejected": -2.158271312713623,
|
| 2507 |
+
"logps/chosen": -1.1788362264633179,
|
| 2508 |
+
"logps/rejected": -1.7396657466888428,
|
| 2509 |
+
"loss": 1.3552,
|
| 2510 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 2511 |
+
"rewards/chosen": -0.05985639616847038,
|
| 2512 |
+
"rewards/margins": 0.1703128218650818,
|
| 2513 |
+
"rewards/rejected": -0.23016922175884247,
|
| 2514 |
+
"step": 1510
|
| 2515 |
+
},
|
| 2516 |
+
{
|
| 2517 |
+
"epoch": 0.994438992476284,
|
| 2518 |
+
"grad_norm": 1.1640625,
|
| 2519 |
+
"learning_rate": 4.176113028983575e-10,
|
| 2520 |
+
"logits/chosen": -2.104264974594116,
|
| 2521 |
+
"logits/rejected": -2.0566983222961426,
|
| 2522 |
+
"logps/chosen": -1.284037709236145,
|
| 2523 |
+
"logps/rejected": -2.4648444652557373,
|
| 2524 |
+
"loss": 1.3592,
|
| 2525 |
+
"rewards/accuracies": 0.699999988079071,
|
| 2526 |
+
"rewards/chosen": -0.09095671027898788,
|
| 2527 |
+
"rewards/margins": 0.30293816328048706,
|
| 2528 |
+
"rewards/rejected": -0.39389482140541077,
|
| 2529 |
+
"step": 1520
|
| 2530 |
+
},
|
| 2531 |
+
{
|
| 2532 |
+
"epoch": 0.9996728819103696,
|
| 2533 |
+
"step": 1528,
|
| 2534 |
+
"total_flos": 0.0,
|
| 2535 |
+
"train_loss": 1.3600662709530735,
|
| 2536 |
+
"train_runtime": 13966.5548,
|
| 2537 |
+
"train_samples_per_second": 4.377,
|
| 2538 |
+
"train_steps_per_second": 0.109
|
| 2539 |
+
}
|
| 2540 |
+
],
|
| 2541 |
+
"logging_steps": 10,
|
| 2542 |
+
"max_steps": 1528,
|
| 2543 |
+
"num_input_tokens_seen": 0,
|
| 2544 |
+
"num_train_epochs": 1,
|
| 2545 |
+
"save_steps": 100,
|
| 2546 |
+
"stateful_callbacks": {
|
| 2547 |
+
"TrainerControl": {
|
| 2548 |
+
"args": {
|
| 2549 |
+
"should_epoch_stop": false,
|
| 2550 |
+
"should_evaluate": false,
|
| 2551 |
+
"should_log": false,
|
| 2552 |
+
"should_save": true,
|
| 2553 |
+
"should_training_stop": true
|
| 2554 |
+
},
|
| 2555 |
+
"attributes": {}
|
| 2556 |
+
}
|
| 2557 |
+
},
|
| 2558 |
+
"total_flos": 0.0,
|
| 2559 |
+
"train_batch_size": 2,
|
| 2560 |
+
"trial_name": null,
|
| 2561 |
+
"trial_params": null
|
| 2562 |
+
}
|