metadata
library_name: peft
tags:
- trl
- dpo
- alignment-handbook
- generated_from_trainer
base_model: NbAiLab/nb-gpt-j-6B-v2
model-index:
- name: aftonposten-6b-align-scan
results: []
aftonposten-6b-align-scan
This model is a fine-tuned version of NbAiLab/nb-gpt-j-6B-v2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4684
- Rewards/chosen: 0.3669
- Rewards/rejected: 0.2161
- Rewards/accuracies: 0.5743
- Rewards/margins: 0.1508
- Logps/rejected: -37.2465
- Logps/chosen: -33.5759
- Logits/rejected: -2.1622
- Logits/chosen: -2.1669
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.4749 | 0.26 | 100 | -2.2375 | -2.2327 | -33.8512 | -37.3537 | 0.4963 | 0.5336 | 0.1467 | 0.0164 | 0.1303 |
| 0.4376 | 0.52 | 200 | -2.2339 | -2.2291 | -33.7896 | -37.2955 | 0.4956 | 0.5486 | 0.1959 | 0.0191 | 0.1769 |
| 0.3835 | 0.78 | 300 | -2.2312 | -2.2264 | -33.7789 | -37.2872 | 0.4950 | 0.5245 | 0.2045 | 0.0210 | 0.1836 |
| 0.3117 | 1.04 | 400 | 0.4891 | 0.3054 | 0.2586 | 0.5652 | 0.0468 | -37.1934 | -33.6528 | -2.2112 | -2.2160 |
| 0.2459 | 1.3 | 500 | 0.4885 | 0.3186 | 0.2671 | 0.5623 | 0.0514 | -37.1827 | -33.6364 | -2.1858 | -2.1906 |
| 0.2639 | 1.56 | 600 | 0.4750 | 0.3623 | 0.2503 | 0.5855 | 0.1120 | -37.2038 | -33.5817 | -2.1784 | -2.1832 |
| 0.2437 | 1.82 | 700 | 0.4742 | 0.3483 | 0.2298 | 0.5748 | 0.1184 | -37.2294 | -33.5992 | -2.1739 | -2.1786 |
| 0.1567 | 2.08 | 800 | 0.4695 | 0.3879 | 0.2480 | 0.5826 | 0.1399 | -37.2066 | -33.5496 | -2.1755 | -2.1803 |
| 0.131 | 2.34 | 900 | 0.4716 | 0.3533 | 0.2206 | 0.5860 | 0.1326 | -37.2408 | -33.5930 | -2.1658 | -2.1705 |
| 0.1784 | 2.6 | 1000 | 0.4673 | 0.3677 | 0.2130 | 0.5860 | 0.1548 | -37.2504 | -33.5749 | -2.1646 | -2.1693 |
| 0.1956 | 2.86 | 1100 | 0.4706 | 0.3580 | 0.2180 | 0.5860 | 0.1400 | -37.2442 | -33.5871 | -2.1622 | -2.1669 |
| 0.137 | 3.12 | 1200 | 0.4680 | 0.3694 | 0.2182 | 0.6063 | 0.1511 | -37.2438 | -33.5728 | -2.1625 | -2.1672 |
| 0.1211 | 3.38 | 1300 | 0.4705 | 0.3633 | 0.2219 | 0.5918 | 0.1414 | -37.2393 | -33.5805 | -2.1622 | -2.1669 |
| 0.1553 | 3.64 | 1400 | 0.4654 | 0.3698 | 0.2068 | 0.6034 | 0.1630 | -37.2582 | -33.5723 | -2.1621 | -2.1668 |
| 0.1447 | 3.9 | 1500 | 0.4684 | 0.3669 | 0.2161 | 0.5743 | 0.1508 | -37.2465 | -33.5759 | -2.1622 | -2.1669 |
Framework versions
- PEFT 0.8.2
- Transformers 4.37.2
- Pytorch 2.1.2+cu121
- Datasets 2.17.0
- Tokenizers 0.15.1