Self-alignment of Large Video Language Models with Refined Regularized Preference Optimization
Paper • 2504.12083 • Published • 3
Clone the repository and navigate to the RRPO directory:
git clone https://github.com/pritamqu/RRPO
cd RRPO
conda create -n longvu python=3.10 -y
conda activate longvu
pip install -r longvu.txt
git clone git@hf.co:pritamqu/LongVU_Qwen2_7B-RRPO-16f
conda activate longvu
WEIGHTS_ROOT="./"
## using merged weights
python inference.py \
--base_model_name "longvu_qwen_7b" \
--model-path ${WEIGHTS_ROOT}"/LongVU_Qwen2_7B-RRPO-16f" \
--video_path "sample_video.mp4" \
--question "Describe this video." \
--model_max_length 1024
Base model
Qwen/Qwen2-7B