78 lines
3.3 KiB
Markdown
78 lines
3.3 KiB
Markdown
---
|
|
frameworks:
|
|
- Pytorch
|
|
license: Apache License 2.0
|
|
tasks:
|
|
- text-generation
|
|
datasets:
|
|
train:
|
|
- Data-Juicer/redpajama-wiki-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-arxiv-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-c4-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-book-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-cc-2019-30-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-cc-2020-05-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-cc-2021-04-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-cc-2022-05-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-cc-2023-06-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-pile-stackexchange-refined-by-data-juicer
|
|
- Data-Juicer/redpajama-stack-code-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-nih-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-europarl-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-philpaper-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-pubmed-abstracts-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-pubmed-central-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-freelaw-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-hackernews-refined-by-data-juicer
|
|
- Data-Juicer/the-pile-uspto-refined-by-data-juicer
|
|
tags:
|
|
- data-juicer
|
|
- arxiv:2309.02033
|
|
---
|
|
## News
|
|
Our first data-centric LLM competition begins! Please visit the competition's official websites, **FT-Data Ranker** ([1B Track](https://tianchi.aliyun.com/competition/entrance/532157), [7B Track](https://tianchi.aliyun.com/competition/entrance/532158)), for more information.
|
|
## Introduction
|
|
This is a reference LLM from [Data-Juicer](https://github.com/alibaba/data-juicer).
|
|
|
|
The model architecture is LLaMA-1.3B and we adopt the [OpenLLaMA](https://github.com/openlm-research/open_llama) implementation.
|
|
The model is pre-trained on 100B tokens of Data-Juicer's refined RedPajama and Pile.
|
|
It achieves an average score of 33.07 over 16 HELM tasks, beating LLMs trained on original RedPajama and Pile datasets.
|
|
|
|
For more details, please refer to our [paper](https://arxiv.org/abs/2309.02033).
|
|
|
|
|
|
|
|
## 使用
|
|
```python
|
|
|
|
from modelscope import (
|
|
AutoModelForCausalLM, AutoTokenizer, GenerationConfig, snapshot_download
|
|
)
|
|
model_dir = 'Data-Juicer/LLaMA-1B-dj-refine-100B'
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_dir)
|
|
model = AutoModelForCausalLM.from_pretrained(model_dir).eval()
|
|
|
|
inputs = tokenizer('How are you?', return_tensors='pt').to(model.device)
|
|
response = model.generate(inputs.input_ids, max_length=128)
|
|
print(tokenizer.decode(response.cpu()[0], skip_special_tokens=True))
|
|
```
|
|
|
|
## 参考
|
|
If you find our work useful for your research or development, please kindly cite the following [paper](https://arxiv.org/abs/2309.02033).
|
|
```
|
|
@misc{chen2023datajuicer,
|
|
title={Data-Juicer: A One-Stop Data Processing System for Large Language Models},
|
|
author={Daoyuan Chen and Yilun Huang and Zhijian Ma and Hesen Chen and Xuchen Pan and Ce Ge and Dawei Gao and Yuexiang Xie and Zhaoyang Liu and Jinyang Gao and Yaliang Li and Bolin Ding and Jingren Zhou},
|
|
year={2023},
|
|
eprint={2309.02033},
|
|
archivePrefix={arXiv},
|
|
primaryClass={cs.LG}
|
|
}
|
|
```
|
|
|
|
#### Clone with HTTP
|
|
```bash
|
|
git clone https://www.modelscope.cn/Data-Juicer/LLaMA-1B-dj-refine-100B.git
|
|
```
|