代码：[colab code](https://colab.research.google.com/drive/1SksjvgRbfpxNQUtYdr2mKxn-OXKHuSov?usp=sharing)
数据集：导出chatgpt数据并使用 [脚本程序整理出可训练的规范数据](https://huggingface.co/wlhb/Llama-3.1-8B-bnb-4bit-Chtagpt/blob/main/origin2trainDatasets.py)

导出Chatgpt的历史聊天记录后使用origin2trainDatasets.py清洗为符合微调模型的数据集，并通过[unsloth](https://unsloth.ai/)进行微调训练

基础模型：unsloth/Meta-Llama-3.1-8B-bnb-4bit
训练方式：lora

效果评价待确定

Code: [colab code](https://colab.research.google.com/drive/1SksjvgRbfpxNQUtYdr2mKxn-OXKHuSov?usp=sharing) Dataset: export chatgpt data and use [script program to organize trainable canonical data](https://huggingface.co/wlhb/Llama-3.1-8B-bnb-4bit-Chtagpt/blob/main/origin2trainDatasets.py)
Export Chatgpt's history chats and use origin2trainDatasets.py to clean them into datasets that match the fine-tuned model and train them with [unsloth](https://unsloth.ai/) for fine-tuning.
Base model: unsloth/Meta-Llama-3.1-8B-bnb-4bit Training method: lora
Effectiveness evaluation to be determined