|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- frankminors123/chinese-shepherd-critic-dataset |
|
language: |
|
- zh |
|
pipeline_tag: question-answering |
|
--- |
|
We trained a Chinese version of Shepherd based on Chinese-LLaMA-2-7B, and we used 2 V100 GPUs with 32G for supervised fine-tuning based on LoRA. |
|
|
|
We designed the appropriate prompt template, and the dataset we used has been published in this HuggingFace repository: frankminors123/chinese-shepherd-critic-dataset, please go to the data page to view details. |
|
|
|
The prompt template used is as follows: |
|
```python |
|
PROMPT_TEMPLATE = ( |
|
"请试着评论下面问题的答案.\n" |
|
"### 问题:\n{question}\n### 答案:\n{answer}\n### 评论:\n" |
|
) |
|
``` |
