README.md

---
language:
- th
license: apache-2.0
library_name: transformers
tags:
- pretrained
pipeline_tag: text-generation
model-index:
- name: typhoon-7b
  results:
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: AI2 Reasoning Challenge (25-Shot)
      type: ai2_arc
      config: ARC-Challenge
      split: test
      args:
        num_few_shot: 25
    metrics:
    - type: acc_norm
      value: 58.53
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=scb10x/typhoon-7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: HellaSwag (10-Shot)
      type: hellaswag
      split: validation
      args:
        num_few_shot: 10
    metrics:
    - type: acc_norm
      value: 81.55
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=scb10x/typhoon-7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU (5-Shot)
      type: cais/mmlu
      config: all
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 59.54
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=scb10x/typhoon-7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: TruthfulQA (0-shot)
      type: truthful_qa
      config: multiple_choice
      split: validation
      args:
        num_few_shot: 0
    metrics:
    - type: mc2
      value: 40.52
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=scb10x/typhoon-7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: Winogrande (5-shot)
      type: winogrande
      config: winogrande_xl
      split: validation
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 76.56
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=scb10x/typhoon-7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GSM8k (5-shot)
      type: gsm8k
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 31.61
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=scb10x/typhoon-7b
      name: Open LLM Leaderboard
---
# Typhoon-7B: Thai Large Language Model (Pretrained)

**Typhoon-7B** is a *pretrained* Thai 🇹🇭 large language model with 7 billion parameters, and it is based on Mistral-7B.

**Typhoon-7B** outperforms all open-source Thai language models at the time of writing as evaluated on Thai examination benchmarks, and its instruction-tuned variant achieves the best results in instruction-following tasks. Also, its performance in Thai is on par with GPT-3.5 while being 2.62 times more efficient in tokenizing Thai text.

**This is not an instruction-tuned model** - It may not be able to follow  human instructions without using one/few-shot learning or instruction fine-tuning. The model does not have any moderation mechanisms, and may generate harmful or inappropriate responses. 

The Instruct model (chat-model) will be released soon. The beta version register is open at https://opentyphoon.ai/.

<div align="center">
<img src="https://storage.googleapis.com/scb10x-ai-lab-public/assets/typhoon_benchmark.png" alt="Typhoon benchmark" width="100%" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
</div>

For full details of this model, please read our [paper](https://arxiv.org/abs/2312.13951).


## Model Description
- **Model type**: A 7B pretrained decoder-only model
- **Requirement**: transformers 4.34.0 or newer.
- **Primary Language(s)**: Thai 🇹🇭 and English 🇬🇧
- **License**: Apache-2.0 (Commercial)

## Performance on Thai Benchmark

| **Model**           | **ONET** | **IC** | **TGAT** | **TPAT-1** | **A-Level** |
|---------------------|----------|--------|----------|------------|-------------|
| Typhoon-7B          | 0.379    | 0.393  | 0.700    | 0.414      | 0.324       |
| SeaLLM-7B           | 0.342    | 0.256  | 0.589    | 0.336      | 0.305       |
| OpenThaiGPT-beta-7B | 0.180    | 0.278  | 0.411    | 0.319      | 0.243       |
| WangChanGLM         | 0.192    | 0.271  | 0.167    | 0.172      | 0.175       |
| SEA-LION-7B         | 0.179    | 0.290  | 0.244    | 0.198      | 0.175       |
| Avg. Human          | 0.318    | -      | 0.472    | 0.406      | -           |

## Intended Uses & Limitations

This model is a pretrained base model. Thus, it may not be able to follow  human instructions without using one/few-shot learning or instruction fine-tuning. The model does not have any moderation mechanisms, and may generate harmful or inappropriate responses.


## SCB10X AI Team
 
- Kunat Pipatanakul, Phatrasek Jirabovonvisut, Potsawee Manakul, Sittipong Sripaisarnmongkol, Ruangsak Patomwong, Pathomporn Chokchainant, Kasima Tharnpipitchai
- If you find Typhoon-7B useful for your work, please cite it using:
```
@article{pipatanakul2023typhoon,
    title={Typhoon: Thai Large Language Models}, 
    author={Kunat Pipatanakul and Phatrasek Jirabovonvisut and Potsawee Manakul and Sittipong Sripaisarnmongkol and Ruangsak Patomwong and Pathomporn Chokchainant and Kasima Tharnpipitchai},
    year={2023},
    journal={arXiv preprint arXiv:2312.13951},
    url={https://arxiv.org/abs/2312.13951}
}
```

## Contact Us

- General & Collaboration: [email protected], [email protected]
- Technical: [email protected]

# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_scb10x__typhoon-7b)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |58.05|
|AI2 Reasoning Challenge (25-Shot)|58.53|
|HellaSwag (10-Shot)              |81.55|
|MMLU (5-Shot)                    |59.54|
|TruthfulQA (0-shot)              |40.52|
|Winogrande (5-shot)              |76.56|
|GSM8k (5-shot)                   |31.61|