Reward-Forcing-T2V-1.3B

<div align="center"> <div align="center"> <h1>Reward Forcing: <br> Efficient Streaming Video Generation with <br> Rewarded Distribution Matching Distillation</h1> <div> <a href="#" target="_blank">Yunhong Lu</a>^1,2, <a href="https://zengyh1900.github.io/" target="_blank">Yanhong Zeng</a>², <a href="#" target="_blank">Haobo Li</a>^2,4, <a href="https://ken-ouyang.github.io/" target="_blank">Hao Ouyang</a>², <a href="https://github.com/qiuyu96" target="_blank">Qiuyu Wang</a>², <a href="https://felixcheng97.github.io/" target="_blank">Ka Leong Cheng</a>², <br> <a href="#" target="_blank">Jiapeng Zhu</a>², <a href="#" target="_blank">Hengyuan Cao</a>¹, <a href="https://zhipengzhang.cn/" target="_blank">Zhipeng Zhang</a>⁵, <a href="https://openreview.net/profile?id=%7EXing_Zhu2" target="_blank">Xing Zhu</a>², <a href="https://shenyujun.github.io/" target="_blank">Yujun Shen</a>², <a href="#" target="_blank">Min Zhang</a>^1,3 </div> <br> <div> ¹ZJU, ²Ant Group, ³SIAS-ZJU, ⁴HUST, ⁵SJTU </div> </div> <br>

</div>

🚀 Progress

• 📝 Technical Report / Paper
• 🌐 Project Homepage
• 💻 Training & Inference Code
• 🤗 Pretrained Model: T2V-1.3B
• 🔜 Pretrained Model: T2V-14B (In progress)

🎯 Overview

<div align="center"> <img src="assets/teaser.png" width="800px"> </div>

TL;DR: We propose Reward Forcing to distill a bidirectional video diffusion model into a 4-step autoregressive student model that enables real-time (23.1 FPS) streaming video generation. Instead of using vanilla distribution matching distillation (DMD), Reward Forcing adopts a novel rewarded distribution matching distillation (Re-DMD) that prioritizes matching towards high-reward regions, leading to enhanced object motion dynamics and immersive scene navigation dynamics in generated videos.

📋 Table of Contents

• Requirements
• Installation
• Pretrained Checkpoints
• Inference
• Training
• Results
• Citation
• Acknowledgements
• Contact

🔧 Requirements

• GPU: NVIDIA GPU with at least 24GB memory for inference, 80GB memory for training.
• RAM: 64GB or more recommended.
• Linux operating system.

🛠️ Installation

Step 1: Clone the repository

git clone https://github.com/JaydenLyh/Reward-Forcing.git
cd Reward-Forcing

Step 2: Create conda environment

conda create -n reward_forcing python=3.10
conda activate reward_forcing

Step 3: Install dependencies

pip install -r requirements.txt
pip install flash-attn --no-build-isolation

Step 4: Install the package

pip install -e .

📦 Pretrained Checkpoints

Download Links

File Structure

After downloading, organize the checkpoints as follows:

checkpoints/
├── Videoreward/
│   ├── checkpoint-11352/
│   └── model_config.json
├── Wan2.1-T2V-1.3B/
├── Wan2.1-T2V-14B/
├── Reward-Forcing-T2V-1.3B/
└── ode_init.pt

Quick Download Script

pip install "huggingface_hub[cli]"

# Download all checkpoints
bash download_checkpoints.sh

🚀 Inference

Quick Start

# 5-seconds video inference
python inference.py \
    --num_output_frames 21 \
    --config_path configs/reward_forcing.yaml \
    --checkpoint_path checkpoints/Reward-Forcing-T2V-1.3B/rewardforcing.pt \
    --output_folder videos/rewardforcing-5s \
    --data_path prompts/MovieGenVideoBench_extended.txt \
    --use_ema

# 30-seconds video inference
python inference.py \
    --num_output_frames 120 \
    --config_path configs/reward_forcing.yaml \
    --checkpoint_path checkpoints/Reward-Forcing-T2V-1.3B/rewardforcing.pt \
    --output_folder videos/rewardforcing-30s \
    --data_path prompts/MovieGenVideoBench_extended.txt \
    --use_ema

🏋️ Training

Multi-GPU Training

# bash train.sh
torchrun --nnodes=1 --nproc_per_node=8 --rdzv_id=5235 --rdzv_backend=c10d  \
    --rdzv_endpoint=$MASTER_PORT train.py  --config_path configs/reward_forcing.yaml \
    --logdir logs/reward_forcing \
    --disable-wandb

Multi-Node Training

torchrun --nnodes=$NODE_SIZE --nproc_per_node=8 --node-rank=$NODE_RANK --rdzv_id=5235 --rdzv_backend=c10d  \
    --rdzv_endpoint=$MASTER_IP:$MASTER_PORT train.py  --config_path configs/reward_forcing.yaml \
    --logdir logs/reward_forcing \
    --disable-wandb

Configuration Files

Training configurations are in configs/:

• default_config.yaml: Default configuration
• reward_forcing.yaml: Reward Forcing configuration

📊 Results

Quantitative Results

Performance on VBench

Qualitative Results

Visualizations can be found in our Project Page.

📄 Citation

If you find this work useful, please consider citing:

@misc{lu2025rewardforcingefficientstreaming,
      title={Reward Forcing: Efficient Streaming Video Generation with Rewarded Distribution Matching Distillation}, 
      author={Yunhong Lu and Yanhong Zeng and Haobo Li and Hao Ouyang and Qiuyu Wang and Ka Leong Cheng and Jiapeng Zhu and Hengyuan Cao and Zhipeng Zhang and Xing Zhu and Yujun Shen and Min Zhang},
      year={2025},
      eprint={2512.04678},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2512.04678}, 
}

🙏 Acknowledgements

This project is built upon several excellent works: CausVid, Self Forcing, Infinite Forcing, Wan2.1, VideoAlign

We thank the authors for their great work and open-source contribution.

📧 Contact

For questions and discussions, please:

• Open an issue on GitHub Issues
• Contact us at: [email protected]