Bagheera

### Background and motivation It would be nice to support other platforms than cuda, eg. MPS ### API Proposal Same as Pytorch. ### API Usage _No response_ ### Alternatives MLX...

Hunyuan-DiT feature branch plan

16

We're going to add Hunyuan DiT to its own feature branch, and this issue encompasses the plan to integrate this model. If you wish to take one of the components,...

when training using EMA, the validations follow upstream Diffusers, and we temporarily overwrite the unet / transformer parameters in the base model with the EMA weights, before running inference. however,...

[sdxl refiner] allow manual import of aesthetic scores

1

a user might want to manually score their dataset and then use those values instead of the single faked value. otoh it can also be useful to import a score...

it would be nice as an optional dependency to include the civitai api client as a way to export the final checkpoint straight to the site. it's a low priority...

merge: reorder model casting, fix kolors on newer diffusers, update cuda, refactor save hooks

1

this will wait a bit longer to shake issues out before merging

dockerfile is outdated and uses wrong cuda dependencies

2

@komninoschatzipapas can you look into this one?

resuming training from LoRA checkpoint fails when base model is quantised

2

workaround is to continue training it without the base being quantised but obviously that's difficult-to-impossible. the bug is seemingly upstream in PEFT.

PEFT implementations of adapters are outdated and languishing

7

### System Info Latest PEFT, transformers, diffusers codebase. ### Who can help? @benjaminbossan @sayakpaul ### Information - [X] The official example scripts - [ ] My own modified scripts ###...

Reported from Reddit: https://www.reddit.com/r/StableDiffusion/comments/1epd0bc/comment/lhlnx5r/?context=3 Starting training with Quanto enabled results in each GPU process trying to quantise simultaneously, which hurts a lot when it comes to system memory use. Can...