Skip to content

litwellchi/videocrafter-training-pytorch

Repository files navigation

TODO

  1. Load Videocrafter2 weight
  2. Longer slide attention
  3. Motion?

🤗 Acknowledgements

Our codebase builds on Stable Diffusion. Our codebase builds on VideoCrafter. Thanks the authors for sharing their awesome codebases!

Install

Already adjust requirements.txt to H800.

pip install -r requirements.txt

Data

/scratch/suptest/video_data/macvid/

数据还在这个位置下;有一些包的metadata还没ready

Training

source activate anaconda3/bin/activate
conda activate videocrafter
sh scripts/train_diffusion.sh

数据的config形式变换为在/home/xchiaa/MACVideoGen/configs/training_data设置要跑的文件。 细节懒得写了,下次一定. 目前由于集群问题,所以wandb需要手动同步一下wandb sync test_macvid_t2v_1024_20240207/wandb/offline-run-20240207_064351-test_macvid_t2v_1024_20240207

About

Training code for the videocrafter.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published