MLLM-Finetuning-Demo

安装LLaMA-Factory

git clone https://github.com/hiyouga/LLaMA-Factory.git
cd LLaMA-Factory
pip install -e .[torch,metrics]
cd .. # 回到项目根目录

预训练

LLaVA中的特征对齐，冻结language_model和vision_tower，只微调multi_modal_projector。

CUDA_VISIBLE_DEVICES=0 llamafactory-cli train config/llava_pt.yaml

微调

CUDA_VISIBLE_DEVICES=0 llamafactory-cli train config/llava_lora_sft.yaml

网页聊天

CUDA_VISIBLE_DEVICES=0 llamafactory-cli webchat \
--model_name_or_path llava-hf/llava-1.5-7b-hf \
--adapter_name_or_path saves/llava1_5-7b/lora/sft \
--template llava

上传数据集到Huggingface

请在 upload_dataset.py 中替换您自己的key.

python3 upload_dataset.py

导出和上传模型到huggingface

请在 config/llava_lora_sft_export.yaml 中替换您自己的 export_hub_model_id 和 hf_hub_token.

CUDA_VISIBLE_DEVICES=0 llamafactory-cli export config/llava_lora_sft_export.yaml