Llama2_LoRA_ReduceBias

Use LoRA to finetune Llama2 to reduce model bias. We use CoLab Environment generally.

Data Processing

We used data set BBQ and preprocessed it for training. The processed dataset was pushed to huggingface dataset.

Applying huggingface PEFT library to finetune Llama2-7b-hf with LoRA.

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
Dataset process.ipynb		Dataset process.ipynb
README.md		README.md
llama2_LoRA_Training.ipynb		llama2_LoRA_Training.ipynb
sample_evaluation.ipynb		sample_evaluation.ipynb