61 lines
2.0 KiB
Markdown
61 lines
2.0 KiB
Markdown
|
---
|
||
|
|
base_model: tanliboy/Llama-3.2-3B
|
||
|
|
datasets:
|
||
|
|
- wassname/ultrachat_200k_filtered
|
||
|
|
library_name: transformers
|
||
|
|
model_name: llama-3.2-3b-sft
|
||
|
|
tags:
|
||
|
|
- generated_from_trainer
|
||
|
|
- alignment-handbook
|
||
|
|
licence: license
|
||
|
|
---
|
||
|
|
|
||
|
|
# Model Card for llama-3.2-3b-sft
|
||
|
|
|
||
|
|
This model is a fine-tuned version of [tanliboy/Llama-3.2-3B](https://huggingface.co/tanliboy/Llama-3.2-3B) on the [['wassname/ultrachat_200k_filtered']](https://huggingface.co/datasets/['wassname/ultrachat_200k_filtered']) dataset.
|
||
|
|
It has been trained using [TRL](https://github.com/huggingface/trl).
|
||
|
|
|
||
|
|
|
||
|
|
Why? because experiments with DPO require a SFT model, this means you can't use base or instruction tuned you need an intermediate model such as https://huggingface.co/allenai/Olmo-3-7B-Think-SFT
|
||
|
|
|
||
|
|
## Quick start
|
||
|
|
|
||
|
|
```python
|
||
|
|
from transformers import pipeline
|
||
|
|
|
||
|
|
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
||
|
|
generator = pipeline("text-generation", model="None", device="cuda")
|
||
|
|
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
||
|
|
print(output["generated_text"])
|
||
|
|
```
|
||
|
|
|
||
|
|
## Training procedure
|
||
|
|
|
||
|
|
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/wassname/huggingface/runs/82mx7iry)
|
||
|
|
|
||
|
|
This model was trained with SFT.
|
||
|
|
|
||
|
|
### Framework versions
|
||
|
|
|
||
|
|
- TRL: 0.12.1
|
||
|
|
- Transformers: 4.52.4
|
||
|
|
- Pytorch: 2.7.0
|
||
|
|
- Datasets: 3.6.0
|
||
|
|
- Tokenizers: 0.21.1
|
||
|
|
|
||
|
|
## Citations
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
Cite TRL as:
|
||
|
|
|
||
|
|
```bibtex
|
||
|
|
@misc{vonwerra2022trl,
|
||
|
|
title = {{TRL: Transformer Reinforcement Learning}},
|
||
|
|
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
||
|
|
year = 2020,
|
||
|
|
journal = {GitHub repository},
|
||
|
|
publisher = {GitHub},
|
||
|
|
howpublished = {\url{https://github.com/huggingface/trl}}
|
||
|
|
}
|
||
|
|
```
|