search

songweige / TATS

Official PyTorch implementation of TATS: A Long Video Generation Framework with Time-Agnostic VQGAN and Time-Sensitive Transformer (ECCV 2022)
MIT License
263 stars 17 forks source link

Checkpoints for TATS-Hierarchical #9

Closed Ugness closed 2 years ago

Ugness commented 2 years ago

Hi, thanks for the great work. Do you have any plan to share the checkpoints and the inference code for TATS-Hierarchical?

If it's hard to share, may I get some metrics or values to monitor about TATS-Hierarchical so that I can reproduce them?

songweige commented 2 years ago

Hi Yoo, thank you for your interests in our work. Please find the updated checkpoints and inference code with the hierarchical models under the synthesis section: https://github.com/SongweiGe/TATS#synthesis.

For the metrics, I usually calculated the FVD manually from time to time to track the training progress, since I found that sometimes even the loss is flattened, the quality can still improve.

Ugness commented 2 years ago

Thanks a lot.