File size: 4,906 Bytes
3829c7f 572c514 3829c7f 572c514 3829c7f 572c514 3829c7f 572c514 3829c7f 4d259a0 3829c7f 13707dc 3829c7f 13707dc 3829c7f 13707dc 3829c7f 13707dc 3829c7f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 |
---
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
base_model: NbAiLab/nb-gpt-j-6B-v2
datasets:
- hugodk-sch/aftonposten_title_prefs
model-index:
- name: aftonposten-6b-align-scan
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# aftonposten-6b-align-scan
This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4763
- Rewards/chosen: 0.2774
- Rewards/rejected: 0.1670
- Rewards/accuracies: 0.5685
- Rewards/margins: 0.1104
- Logps/rejected: -37.2781
- Logps/chosen: -33.6382
- Logits/rejected: -2.1561
- Logits/chosen: -2.1608
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-06
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 4
### Training results
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
| 0.4799 | 0.26 | 100 | -2.2381 | -2.2333 | -33.8607 | -37.3570 | 0.4978 | 0.5341 | 0.1217 | 0.0100 | 0.1117 |
| 0.4453 | 0.52 | 200 | -2.2347 | -2.2299 | -33.7685 | -37.2937 | 0.4928 | 0.5370 | 0.1862 | 0.0302 | 0.1561 |
| 0.3947 | 0.78 | 300 | -2.2322 | -2.2274 | -33.7551 | -37.2894 | 0.4910 | 0.5565 | 0.1956 | 0.0365 | 0.1591 |
| 0.3136 | 1.04 | 400 | 0.4857 | 0.2846 | 0.2244 | 0.5797 | 0.0602 | -37.1961 | -33.6280 | -2.2032 | -2.2080 |
| 0.2784 | 1.3 | 500 | 0.4891 | 0.2959 | 0.2519 | 0.5220 | 0.0439 | -37.1567 | -33.6119 | -2.2050 | -2.2098 |
| 0.2593 | 1.56 | 600 | 0.4795 | 0.3345 | 0.2439 | 0.5743 | 0.0906 | -37.1682 | -33.5567 | -2.1866 | -2.1914 |
| 0.2606 | 1.82 | 700 | 0.4764 | 0.3188 | 0.2158 | 0.6063 | 0.1031 | -37.2084 | -33.5791 | -2.1788 | -2.1836 |
| 0.1758 | 2.08 | 800 | 0.4767 | 0.2840 | 0.1749 | 0.5860 | 0.1091 | -37.2668 | -33.6289 | -2.1680 | -2.1727 |
| 0.1687 | 2.34 | 900 | 0.4770 | 0.2898 | 0.1833 | 0.5486 | 0.1065 | -37.2547 | -33.6205 | -2.1626 | -2.1674 |
| 0.1826 | 2.6 | 1000 | 0.4764 | 0.2700 | 0.1574 | 0.5831 | 0.1126 | -37.2917 | -33.6489 | -2.1578 | -2.1625 |
| 0.1541 | 2.86 | 1100 | 0.4751 | 0.2864 | 0.1692 | 0.5777 | 0.1171 | -37.2748 | -33.6254 | -2.1561 | -2.1608 |
| 0.194 | 3.12 | 1200 | 0.4748 | 0.2856 | 0.1654 | 0.5801 | 0.1202 | -37.2803 | -33.6265 | -2.1565 | -2.1612 |
| 0.1414 | 3.38 | 1300 | 0.4753 | 0.2859 | 0.1690 | 0.5831 | 0.1169 | -37.2751 | -33.6261 | -2.1558 | -2.1605 |
| 0.1492 | 3.64 | 1400 | 0.4744 | 0.2846 | 0.1627 | 0.5918 | 0.1220 | -37.2842 | -33.6279 | -2.1556 | -2.1603 |
| 0.1694 | 3.9 | 1500 | 0.4747 | 0.2822 | 0.1614 | 0.5569 | 0.1208 | -37.2860 | -33.6314 | -2.1560 | -2.1607 |
### Framework versions
- PEFT 0.10.0
- Transformers 4.39.0.dev0
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.15.1 |