README.md

---
language:
- nl
license: mit
tags:
- trl
- fietje
- alignment-handbook
- dpo
base_model: BramVanroy/fietje-2-instruct
datasets:
- BramVanroy/ultra_feedback_dutch_cleaned
- BramVanroy/orca_dpo_pairs_dutch_cleaned
pipeline_tag: text-generation
inference: false
model-index:
- name: fietje-2-chat
  results: []
---

<p align="center" style="margin:0;padding:0">
  <img src="https://huggingface.co/BramVanroy/fietje-2-chat/resolve/main/img/fietje-2b-banner-rounded.png" alt="Fietje banner" width="800" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
</p>

<div style="margin:auto; text-align:center">
  <h1 style="margin-bottom: 0">Fietje 2 Chat</h1>
  <em>An open and efficient LLM for Dutch</em>
</div>

<blockquote class="tip" style="padding: 1.5em; border: 0">
  <p align="center" style="text-align: center; margin: 0">
    <a href="https://huggingface.co/BramVanroy/fietje-2">👱‍♀️ Base version</a> -
    <a href="https://huggingface.co/BramVanroy/fietje-2-instruct">🤖 Instruct version</a> -
    <a href="https://huggingface.co/BramVanroy/fietje-2-chat">💬 Chat version</a> (this one) -
    <a href="https://huggingface.co/BramVanroy/fietje-2-chat-GGUF">🚀 GGUF of Chat</a>
  </p>
  <p align="center" style="text-align: center; margin: 0">
    <a href="https://huggingface.co/spaces/BramVanroy/fietje-2b"><strong>Chat with Fietje here!</strong></a>
  </p>
</blockquote>

This is the chat version of Fietje, a DPO-tuned (aligned) continuation on [the instruct version](https://huggingface.co/BramVanroy/fietje-2-instruct). Fietje is an adapated version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2), tailored to Dutch text generation by training on 28B tokens. It is small and efficient with a size of 2.7 billion parameters while performing almost on par with more powerful Dutch LLMs of twice its size like [GEITje 7B Ultra](https://huggingface.co/BramVanroy/GEITje-7B-ultra).

A thorough description of the creation and evaluation of Fietje as well as usage examples are available in [this Github repository](https://github.com/BramVanroy/fietje).

## Citation

If you use Fietje or the [CulturaX + Wikipedia filtered subset](https://huggingface.co/datasets/BramVanroy/wikipedia_culturax_dutch) in your work, please cite to the following paper:

```bibtex
@misc{vanroy2024fietjeopenefficientllm,
      title={Fietje: An open, efficient LLM for Dutch}, 
      author={Bram Vanroy},
      year={2024},
      eprint={2412.15450},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2412.15450}, 
}
```

## Intended uses & limitations

The same limitations as [phi-2](https://huggingface.co/microsoft/phi-2#limitations-of-phi-2), and LLMs in general, apply here. LLMs hallucinate, make mistakes, and should not be trusted. Use at your own risk!

## Training and evaluation data

Fietje 2 Chat was finetuned from [the instruct model](https://huggingface.co/BramVanroy/fietje-2-instruct) on the following datasets. Number of training samples per dataset given in brackets, totalling 18,653 samples.

- [BramVanroy/ultra_feedback_dutch_cleaned](https://huggingface.co/datasets/BramVanroy/ultra_feedback_dutch_cleaned) subset `dpo_hq`: a cleaned version of [BramVanroy/ultra_feedback_dutch](https://huggingface.co/datasets/BramVanroy/ultra_feedback_dutch) (9186)
- [BramVanroy/orca_dpo_pairs_dutch_cleaned](https://huggingface.co/datasets/BramVanroy/orca_dpo_pairs_dutch_cleaned) subset `dpo_all`: a cleaned version of [BramVanroy/orca_dpo_pairs_dutch](https://huggingface.co/datasets/BramVanroy/orca_dpo_pairs_dutch) (9467)

A lot of different learning rates, beta, en batch sizes were investigated in search of a converging combination. You can find them all in [the W&B runs](https://wandb.ai/bramvanroy/dpo-fietje-2b).

## Training procedure

I am thankful to the [Flemish Supercomputer Center](https://www.vscentrum.be/) (VSC) for providing the computational power to accomplish this project. Accounting for waiting for jobs, training a single run took around nine hours on one A100 80GB.

Training was done with the wonderful [alignment-handbook](https://github.com/huggingface/alignment-handbook), using DeepSpeed as a back-end. Exact training recipes and SLURM script are given in the [Github repository](https://github.com/BramVanroy/fietje).


### Training hyperparameters

The following hyperparameters were used during training:
- beta: 0.2
- learning_rate: 2e-06
- train_batch_size: 8
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-07
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1.0

### Training results

| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.2515        | 1.0   | 1166 | 0.2842          | -1.1549        | -3.6363          | 0.8867             | 2.4815          | -657.6813      | -451.3364    | -1.2868         | -1.3528       |


### Framework versions

- Transformers 4.39.1
- Pytorch 2.1.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2