---
license: apache-2.0
datasets:
- shareAI/ShareGPT-Chinese-English-90k
language:
- zh
pipeline_tag: text-generation
tags:
- tiny
- small
- opt
---
## 这是什么 - Intro

中文 OPT-125m，在原 [OPT-125m](https://huggingface.co/facebook/opt-125m) 模型基础上[扩充了中文词表](https://github.com/stanleylsx/llms_tool)，并在 WebText, ShareGPT 等数据集上作了一定的继续预训练。

Chinese OPT-125m, expanding the Chinese word list based on the original [OPT-125m](https://huggingface.co/facebook/opt-125m) model, and making some continued pre-training on WebText, ShareGPT, etc.



## 快速试用 - Quick Use

```python
# 快速试用
# quick use
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name_or_path = 'Sholder/chinese-opt-125m'
device = 'cuda'

model = AutoModelForCausalLM.from_pretrained(model_name_or_path).to(device)
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)

def quick_use():
      text = input('User: ').strip()
      inputs = tokenizer(text, return_tensors="pt").to(device)

      print('Opt: ', end='', flush=True)
      tokens = model.generate(
          **inputs,
          max_new_tokens=64,
          do_sample=True,
      )

      response = tokenizer.decode(tokens[0])
      print(response)

if __name__ == '__main__':
    quick_use()
```