2024-11-02 16:18:09 +08:00
|
|
|
|
方案:
|
2024-11-02 16:34:20 +08:00
|
|
|
|
全参数微调,使用不同数据集训练多个模型和推理时增强进行融合。
|
2024-11-02 16:18:09 +08:00
|
|
|
|
|
|
|
|
|
|
2024-11-02 13:53:14 +08:00
|
|
|
|
训练代码:
|
2024-11-02 16:34:20 +08:00
|
|
|
|
LLaMA-Factory.zip 解压后使用,可参照https://github.com/hiyouga/LLaMA-Factory配置环境,或将代码映射到docker中使用。
|
2024-11-02 16:54:13 +08:00
|
|
|
|
训练:train.sh。将数据集放到LLaMA-Factory/data文件夹下,将train.sh放到LLaMA-Factory下使用。
|
2024-11-02 16:34:20 +08:00
|
|
|
|
推理: python inference.py(需在inference.py中修改好模型路径。) test_case.json是从题目中提取出来的测试用例。
|
2024-11-02 14:49:03 +08:00
|
|
|
|
|
2024-11-02 17:00:02 +08:00
|
|
|
|
百度网盘需要收费,使用阿里云盘
|
2024-11-02 14:49:03 +08:00
|
|
|
|
model_wight:通过百度网盘分享的文件:
|
2024-11-02 17:48:43 +08:00
|
|
|
|
链接:https://pan.baidu.com/s/1paYNO7d5OYESuyw3BVo7Ew
|
|
|
|
|
提取码:6666
|
2024-11-02 17:00:02 +08:00
|
|
|
|
https://www.alipan.com/s/FTPWUSBuz7s
|
2024-11-02 14:49:03 +08:00
|
|
|
|
|
|
|
|
|
docker:
|
2024-11-02 17:48:43 +08:00
|
|
|
|
链接:https://pan.baidu.com/s/1paYNO7d5OYESuyw3BVo7Ew
|
|
|
|
|
提取码:6666
|
2024-11-02 17:00:02 +08:00
|
|
|
|
https://www.alipan.com/s/FTPWUSBuz7s
|
2024-11-02 16:09:56 +08:00
|
|
|
|
|
|
|
|
|
train_data:
|
2024-11-02 17:48:43 +08:00
|
|
|
|
链接:https://pan.baidu.com/s/1paYNO7d5OYESuyw3BVo7Ew
|
|
|
|
|
提取码:6666
|
2024-11-02 17:00:02 +08:00
|
|
|
|
https://www.alipan.com/s/FTPWUSBuz7s
|