README.md

---
language:
- en
- zh
- id
- th
- vi
- ms
- lo
datasets:
- cerebras/SlimPajama-627B
- Skywork/SkyPile-150B
- allenai/MADLAD-400
- cc100
tags:
- multilingual
- sea
- sailor
license: apache-2.0
base_model: Qwen/Qwen1.5-4B
inference: false
model-index:
- name: Sailor-4B
  results:
  - task:
      type: text-generation
    dataset:
      name: XQuAD-Thai
      type: XQuAD-Thai
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 46.82
    - name: F1 (3-Shot)
      type: F1 (3-Shot)
      value: 63.34
  - task:
      type: text-generation
    dataset:
      name: TyDiQA-Indonesian
      type: TyDiQA-Indonesian
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 53.98
    - name: F1 (3-Shot)
      type: F1 (3-Shot)
      value: 73.48
  - task:
      type: text-generation
    dataset:
      name: XQuAD-Vietnamese
      type: XQuAD-Vietnamese
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 47.65
    - name: F1 (3-Shot)
      type: F1 (3-Shot)
      value: 67.09
  - task:
      type: text-generation
    dataset:
      name: XCOPA-Thai
      type: XCOPA-Thai
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 53.4
  - task:
      type: text-generation
    dataset:
      name: XCOPA-Indonesian
      type: XCOPA-Indonesian
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 69.20
  - task:
      type: text-generation
    dataset:
      name: XCOPA-Vietnamese
      type: XCOPA-Vietnamese
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 68.20
  - task:
      type: text-generation
    dataset:
      name: M3Exam-Thai
      type: M3Exam-Thai
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 27.88
  - task:
      type: text-generation
    dataset:
      name: M3Exam-Indonesian
      type: M3Exam-Indonesian
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 31.27
  - task:
      type: text-generation
    dataset:
      name: M3Exam-Vietnamese
      type: M3Exam-Vietnamese
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 40.69
  - task:
      type: text-generation
    dataset:
      name: BELEBELE-Thai
      type: BELEBELE-Thai
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 36.11
  - task:
      type: text-generation
    dataset:
      name: BELEBELE-Indonesian
      type: BELEBELE-Indonesian
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 41.33
  - task:
      type: text-generation
    dataset:
      name: BELEBELE-Vietnamese
      type: BELEBELE-Vietnamese
    metrics:
    - name: EM (3-Shot)
      type: EM (3-Shot)
      value: 38.89
---

<div align="center">
  <img src="banner_sailor.jpg" width="700"/>
</div>

Sailor is a suite of Open Language Models tailored for South-East Asia (SEA), focusing on languages such as 🇮🇩Indonesian, 🇹🇭Thai, 🇻🇳Vietnamese, 🇲🇾Malay, and 🇱🇦Lao. 
Developed with careful data curation, Sailor models are designed to understand and generate text across diverse linguistic landscapes of SEA region. 
Built from [Qwen 1.5](https://huggingface.co/collections/Qwen/qwen15-65c0a2f577b1ecb76d786524) , Sailor encompasses models of varying sizes, spanning from 0.5B to 14B versions for different requirements. 
We further fine-tune the base model with open-source datasets to get instruction-tuned models, namedly Sailor-Chat. 
Benchmarking results demonstrate Sailor's proficiency in tasks such as question answering, commonsense reasoning, and other tasks in SEA languages.

> The logo was generated by MidJourney

## Model Summary
- **Model Collections:** [Base Model & Chat Model](https://huggingface.co/collections/sail/sailor-65e19a749f978976f1959825)
- **Project Website:** [sailorllm.github.io](https://sailorllm.github.io/)
- **Codebase:** [github.com/sail-sg/sailor-llm](https://github.com/sail-sg/sailor-llm)
- **Technical Report:** [arxiv.org/pdf/2404.03608.pdf](https://arxiv.org/pdf/2404.03608.pdf) 


## Training details
Sailor is crafted by continually pre-training from language models like the remarkable Qwen 1.5 models, which already has a great performance on SEA languages. 
The pre-training corpus heavily leverages the publicly available corpus, including 
[SlimPajama](https://huggingface.co/datasets/cerebras/SlimPajama-627B), 
[SkyPile](https://huggingface.co/datasets/Skywork/SkyPile-150B), 
[CC100](https://huggingface.co/datasets/cc100) and [MADLAD-400](https://huggingface.co/datasets/allenai/MADLAD-400).

By employing aggressive data deduplication and careful data cleaning on the collected corpus, we have attained a high-quality dataset spanning various languages. 
Through systematic experiments to determine the weights of different languages, Sailor models undergo training from 200B to 400B tokens, tailored to different model sizes. 
The approach boosts their performance on SEA languages while maintaining proficiency in English and Chinese without significant compromise. 
Finally, we continually pre-train the Qwen1.5-0.5B model with 400 Billion tokens, and other models with 200 Billion tokens to obtain the Sailor models.

## Requirements
The code of Sailor has been in the latest Hugging face transformers and we advise you to install `transformers>=4.37.0`.

## Quickstart

Here provides a code snippet to show you how to load the tokenizer and model and how to generate contents.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model

model = AutoModelForCausalLM.from_pretrained("sail/Sailor-4B", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("sail/Sailor-4B")

input_message = "Model bahasa adalah model probabilistik" 
### The given Indonesian input translates to 'A language model is a probabilistic model of.'

model_inputs = tokenizer([input_message], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=64
)

generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)
```

# License

Sailor is distributed under the terms of the Apache License 2.0. 
No restrict on the research and the commercial use, but should comply with the [Qwen License](https://huggingface.co/Qwen/Qwen1.5-1.8B/blob/main/LICENSE).

## Citation

If you find sailor useful, please cite our work as follows:

```
@article{dou2024sailor,
  title={Sailor: Open Language Models for South-East Asia},
  author={Dou, Longxu and Liu, Qian and Zeng, Guangtao and Guo, Jia and Zhou, Jiahui and Lu, Wei and Lin, Min},
  journal={arXiv preprint arXiv:2404.03608},
  year={2024}
}
```

# Contact Us

If you have any questions, please raise an issue or contact us at [doulx@sea.com](mailto:doulx@sea.com) or [liuqian@sea.com](mailto:liuqian@sea.com).