AVC-Reward: Event-Centric Audio-Video Consistency Evaluation
Paper • 2610.22321 • Published • 17
AVC-Reward
Event-Centric Audio-Video Consistency Evaluation · Full reward-model checkpoint.
📄 arXiv · 🌐 Project Page · 💻 GitHub · 🤗 Benchmark · 🤗 Generation LoRA
Download
hf download Tsuki0512/AVC-Reward --local-dir checkpoints/AVC-Reward
Inference
From the GitHub repository, after setting up the environment:
export MODEL_DIR="$PWD/checkpoints/AVC-Reward"
export CUDA_VISIBLE_DEVICES=0
export TENSOR_PARALLEL_SIZE=1
export ALLOWED_LOCAL_MEDIA_PATH="/absolute/path/to/media"
bash reward_model/code/start_reward_vllm.sh
In another terminal:
python reward_model/code/reward_model_inference.py \
--input /absolute/path/to/manifest.jsonl \
--output-jsonl outputs/reward_scores.jsonl \
--endpoint http://127.0.0.1:8901/v1/chat/completions \
--model reward_model
- Downloads last month
- 73
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Tsuki0512/AVC-Reward
Base model
Qwen/Qwen3-Omni-30B-A3B-Thinking