PixArt-alpha / PixArt-sigma

PixArt-Σ: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation
https://pixart-alpha.github.io/PixArt-sigma-project/
GNU Affero General Public License v3.0
1.63k stars 77 forks source link

In some scenarios, there are poor performance issues in 2K checkpoint. #77

Closed kelisiya closed 5 months ago

kelisiya commented 5 months ago

. When I tested with 2k weights, I noticed a lot of graininess. I would like to know if this is caused by insufficient training. If I want to fine-tune for this scenario, do you have any suggestions? image

lawrence-cj commented 5 months ago

Correct. We only use 8A100, 200K images(4K steps not finishing one epoch during training.) You can further improve it with more training data and time.