---
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: 0.7456
- Rewards/chosen: 0.0514
- Rewards/rejected: 0.0332
- Rewards/accuracies: 0.5365
- Rewards/margins: 0.0182
- Logps/rejected: -37.4614
- Logps/chosen: -33.9489
- Logits/rejected: -2.2407
- Logits/chosen: -2.2456

## 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 |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
| 0.5903        | 0.26  | 100  | -2.2366       | -2.2318         | -34.0141     | -37.5087       | 0.7450          | 0.5241             | 0.0123         | 0.0075          | 0.0047           |
| 0.59          | 0.52  | 200  | -2.2332       | -2.2284         | -34.0691     | -37.5752       | 0.7500          | 0.5365             | -0.0208        | 0.0144          | -0.0352          |
| 0.3871        | 0.78  | 300  | -2.2333       | -2.2285         | -34.0622     | -37.5657       | 0.7655          | 0.5071             | -0.0166        | 0.0129          | -0.0295          |
| 0.3315        | 1.04  | 400  | 0.7479        | -0.0009         | -0.0296      | 0.5515         | 0.0287          | -37.5660           | -34.0361       | -2.2265         | -2.2314          |
| 0.3857        | 1.3   | 500  | 0.7601        | -0.0010         | -0.0248      | 0.5577         | 0.0238          | -37.5580           | -34.0362       | -2.2294         | -2.2343          |
| 0.5531        | 1.56  | 600  | 0.7435        | 0.0040          | -0.0183      | 0.5577         | 0.0223          | -37.5471           | -34.0279       | -2.2416         | -2.2465          |
| 0.4315        | 1.82  | 700  | 0.8141        | -0.0077         | -0.0028      | 0.4996         | -0.0049         | -37.5213           | -34.0474       | -2.2449         | -2.2498          |
| 0.3445        | 2.08  | 800  | 0.8104        | 0.0382          | 0.0369       | 0.5282         | 0.0012          | -37.4551           | -33.9710       | -2.2437         | -2.2485          |
| 0.1902        | 2.34  | 900  | 0.8123        | 0.0329          | 0.0364       | 0.5199         | -0.0035         | -37.4560           | -33.9797       | -2.2426         | -2.2475          |
| 0.1724        | 2.6   | 1000 | 0.7850        | 0.0479          | 0.0385       | 0.5365         | 0.0094          | -37.4525           | -33.9548       | -2.2433         | -2.2481          |
| 0.1864        | 2.86  | 1100 | 0.7625        | 0.0435          | 0.0297       | 0.5133         | 0.0137          | -37.4671           | -33.9621       | -2.2392         | -2.2441          |
| 0.1629        | 3.12  | 1200 | 0.7624        | 0.0479          | 0.0337       | 0.5341         | 0.0142          | -37.4604           | -33.9547       | -2.2387         | -2.2436          |
| 0.1881        | 3.38  | 1300 | 0.7945        | 0.0454          | 0.0440       | 0.4967         | 0.0014          | -37.4432           | -33.9588       | -2.2391         | -2.2440          |
| 0.1395        | 3.64  | 1400 | 0.7520        | 0.0496          | 0.0341       | 0.5453         | 0.0156          | -37.4598           | -33.9518       | -2.2408         | -2.2457          |
| 0.1436        | 3.9   | 1500 | 0.7352        | 0.0537          | 0.0336       | 0.5399         | 0.0201          | -37.4606           | -33.9451       | -2.2407         | -2.2456          |


### Framework versions

- PEFT 0.8.2
- Transformers 4.37.2
- Pytorch 2.1.2+cu121
- Datasets 2.17.0
- Tokenizers 0.15.1