metadata
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: []
aftonposten-6b-align-scan
This model is a fine-tuned version of 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: 0.4730
- Rewards/chosen: 0.2464
- Rewards/rejected: 0.1175
- Rewards/accuracies: 0.5947
- Rewards/margins: 0.1288
- Logps/rejected: -37.3207
- Logps/chosen: -33.6239
- Logits/rejected: -2.1470
- Logits/chosen: -2.1517
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 | -2.2349 | -2.2301 | -33.9029 | -37.3943 | 0.4987 | 0.5104 | 0.0790 | 0.0056 | 0.0734 |
| 0.4548 | 0.52 | 200 | -2.2335 | -2.2287 | -33.7696 | -37.2846 | 0.4956 | 0.5341 | 0.1590 | 0.0198 | 0.1392 |
| 0.41 | 0.78 | 300 | -2.2309 | -2.2261 | -33.7614 | -37.2848 | 0.4937 | 0.5361 | 0.1639 | 0.0248 | 0.1391 |
| 0.3497 | 1.04 | 400 | 0.4927 | 0.2171 | 0.1863 | 0.5652 | 0.0309 | -37.2062 | -33.6727 | -2.2113 | -2.2162 |
| 0.2906 | 1.3 | 500 | 0.4870 | 0.2484 | 0.1921 | 0.5922 | 0.0563 | -37.1964 | -33.6205 | -2.1834 | -2.1881 |
| 0.3014 | 1.56 | 600 | 0.4796 | 0.2630 | 0.1719 | 0.5797 | 0.0911 | -37.2301 | -33.5962 | -2.1694 | -2.1741 |
| 0.2776 | 1.82 | 700 | 0.4825 | 0.2341 | 0.1554 | 0.5768 | 0.0787 | -37.2576 | -33.6444 | -2.1625 | -2.1672 |
| 0.201 | 2.08 | 800 | 0.4766 | 0.2639 | 0.1595 | 0.5914 | 0.1043 | -37.2507 | -33.5948 | -2.1641 | -2.1689 |
| 0.1721 | 2.34 | 900 | 0.4749 | 0.2446 | 0.1298 | 0.5860 | 0.1148 | -37.3003 | -33.6269 | -2.1516 | -2.1563 |
| 0.2259 | 2.6 | 1000 | 0.4736 | 0.2483 | 0.1257 | 0.5860 | 0.1226 | -37.3072 | -33.6207 | -2.1481 | -2.1528 |
| 0.2405 | 2.86 | 1100 | 0.4740 | 0.2438 | 0.1229 | 0.5860 | 0.1209 | -37.3118 | -33.6283 | -2.1475 | -2.1522 |
| 0.1793 | 3.12 | 1200 | 0.4746 | 0.2441 | 0.1249 | 0.5685 | 0.1192 | -37.3085 | -33.6277 | -2.1469 | -2.1516 |
| 0.1633 | 3.38 | 1300 | 0.4744 | 0.2433 | 0.1235 | 0.6009 | 0.1198 | -37.3107 | -33.6290 | -2.1471 | -2.1518 |
| 0.202 | 3.64 | 1400 | 0.4748 | 0.2450 | 0.1279 | 0.5831 | 0.1170 | -37.3034 | -33.6263 | -2.1472 | -2.1519 |
| 0.1889 | 3.9 | 1500 | 0.4727 | 0.2480 | 0.1188 | 0.6005 | 0.1292 | -37.3186 | -33.6212 | -2.1470 | -2.1517 |
Framework versions
- PEFT 0.8.2
- Transformers 4.37.2
- Pytorch 2.1.2+cu121
- Datasets 2.17.0
- Tokenizers 0.15.1