conda create --name ddxtutor python=3.11 -y
conda activate ddxtutor
conda install pytorch pytorch-cuda=12.1 -c pytorch -c nvidia -y
conda install xformers -c xformers -y
pip install unsloth
pip install omegaconf
pip install beautifulsoup4
pip install llama-index
pip install llama-index-llms-openai
pip install vllm
pip install wandbSet the following environment variable to specify your model path:
export MODEL_PATH="/path/to/your/model"If not set, the default path will be ./models/phi-4.
The configuration file Config/phi4_14b_lora.yaml contains all the training parameters. You can modify the model path in the config file or use the MODEL_PATH environment variable.
refer to Finetune.py and inference.py
https://drive.google.com/drive/folders/1SnArTQy3rkQbBD-f0xPObcsLyf_j4AIf?usp=drive_link