Open-aligned models using Magpie datasets. • 11 items • Updated • 1
🐦 Llama-3.1-8B-Magpie-Align-SFT-v0.1
Project Web: https://magpie-align.github.io/
Arxiv Technical Report: https://arxiv.org/abs/2406.08464
Codes: https://github.com/magpie-align/magpie
Abstract
About This Model
This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B on
It achieves performance comparable with the official Llama-3.1-8B-Instruct Model with SFT only!
- Alpaca Eval 2 (GPT-4-Turbo-1106): 24.79 (LC), 25.05 (WR)
- Arena Hard: 21.0
Other Information
License: Please follow Meta Llama 3 Community License (Data) and Meta Llama 3.1 Community License (Model).
Conversation Template: Please use Llama 3 official chat template for the best performance.
Questions? Please contact Zhangchen by email.
Citation
If you find the model, data, or code useful, please cite our paper:
@article{xu2024magpie,
title={Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothing},
author={Zhangchen Xu and Fengqing Jiang and Luyao Niu and Yuntian Deng and Radha Poovendran and Yejin Choi and Bill Yuchen Lin},
year={2024},
eprint={2406.08464},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 79
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.7863 | 0.0024 | 1 | 0.7710 |
| 0.5422 | 0.2007 | 85 | 0.4937 |
| 0.476 | 0.4014 | 170 | 0.4382 |
| 0.4594 | 0.6021 | 255 | 0.4174 |
| 0.4383 | 0.8028 | 340 | 0.4057 |
| 0.4397 | 1.0035 | 425 | 0.3978 |
| 0.3927 | 1.1845 | 510 | 0.3956 |
| 0.3895 | 1.3852 | 595 | 0.3934 |
| 0.3832 | 1.5859 | 680 | 0.3925 |
| 0.3957 | 1.7866 | 765 | 0.3924 |
Framework versions
- Transformers 4.43.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 538
Safetensors
Model size
8B params
Tensor type
BF16
·
Model tree for Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.1
Base model
meta-llama/Llama-3.1-8B