TeaCache/TeaCache4CogVideoX1.5

TeaCache4CogVideoX1.5

TeaCache can speedup CogVideoX1.5 1.8x without much visual quality degradation, in a training-free manner. The following video shows the results generated by TeaCache-ConsisID with various rel_l1_thresh values: 0 (original), 0.1 (1.3x speedup), 0.2 (1.8x speedup), and 0.3(2.1x speedup).Additionally, the image-to-video (i2v) results are also demonstrated, with the following speedups: 0.1 (1.5x speedup), 0.2 (2.2x speedup), and 0.3 (2.7x speedup).

https://github.com/user-attachments/assets/c444b850-3252-4b37-ad4a-122d389218d9

https://github.com/user-attachments/assets/5f181a57-d5e3-46db-b388-8591e50f98e2

📈 Inference Latency Comparisons on a Single H100 GPU

CogVideoX1.5-t2v	TeaCache (0.1)	TeaCache (0.2)	TeaCache (0.3)
~465 s	~372 s	~261 s	~223 s

CogVideoX1.5-i2v	TeaCache (0.1)	TeaCache (0.2)	TeaCache (0.3)
~475 s	~323 s	~218 s	~171 s

Installation

pip install --upgrade diffusers[torch] transformers protobuf tokenizers sentencepiece imageio imageio-ffmpeg

Usage

You can modify the rel_l1_thresh to obtain your desired trade-off between latency and visul quality, and change the ckpts_path, prompt, image_path to customize your identity-preserving video.

For T2V inference, you can use the following command:

cd TeaCache4CogVideoX1.5

python3 teacache_sample_video.py \
    --rel_l1_thresh 0.2 \
    --ckpts_path THUDM/CogVideoX1.5-5B \
    --prompt "A clear, turquoise river flows through a rocky canyon, cascading over a small waterfall and forming a pool of water at the bottom.The river is the main focus of the scene, with its clear water reflecting the surrounding trees and rocks. The canyon walls are steep and rocky, with some vegetation growing on them. The trees are mostly pine trees, with their green needles contrasting with the brown and gray rocks. The overall tone of the scene is one of peace and tranquility.", help='Description of the video for the model to generate." \
    --seed 42 \
    --num_inference_steps 50 \
    --output_path ./teacache_results

For I2V inference, you can use the following command:

cd TeaCache4CogVideoX1.5

python3 teacache_sample_video.py \
    --rel_l1_thresh 0.1 \
    --ckpts_path THUDM/CogVideoX1.5-5B-I2V \
    --prompt "A girl gazed at the camera and smiled, her hair drifting in the wind." \
    --seed 42 \
    --num_inference_steps 50 \
    --output_path ./teacache_results \
    --image_path ./image/path \

Citation

If you find TeaCache is useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@article{liu2024timestep,
  title={Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model},
  author={Liu, Feng and Zhang, Shiwei and Wang, Xiaofeng and Wei, Yujie and Qiu, Haonan and Zhao, Yuzhong and Zhang, Yingya and Ye, Qixiang and Wan, Fang},
  journal={arXiv preprint arXiv:2411.19108},
  year={2024}
}

Acknowledgements

We would like to thank the contributors to the CogVideoX and Diffusers.