Add LoRA Method Support for Parameter-Efficient Fine-Tuning

[harrison-huan-liu]

Description

This PR introduces support for Low-Rank Adaptation (LoRA) method, a parameter-efficient fine-tuning technique that significantly reduces computational costs and memory usage during training while maintaining model performance.

Related Issue

#7

Key Changes

LoRA Implementation:

• Added LoRA configuration parameters to the Args class in arguments.py
• Implemented _apply_lora() method in model.py to apply LoRA adaptation using the PEFT library

Configuration Updates:

• Added a new configuration file configs/config_lora.json with example LoRA settings
• Updated requirements.txt to include the peft library dependency
• Added documentation for LoRA training in README.md