本项目是通过使用金融问答数据对Llama-3-8B-Instruct模型进行4bit量化微调,以提升模型在金融领域的咨询问答功能。
According to usage tips for Llama3 Training the model in float16 is not recommended and is known to produce nan; as such, the model should be trained in bfloat16.
申请Llama-3-8B-Instruct模型的使用权,
将huggingface token填入loftq_4bit.yaml
与benchmark.sh
文件中。
python train.py loftq_4bit.yaml
sh benchmark.sh