ComfyUI node & LyCORIS support

[bghira]

hello, i've created a ComfyUI node for T-LoRA loading / sampling: https://github.com/bghira/ComfyUI-T-LoRA

it currently supports SDXL LoRAs trained using this repo only, though I just noticed you've added Flux. it doesn't yet work with that.

the LyCORIS project has an initial pass at T-LoRA; it uses SVD instead of orthogonal init. simpletuner is using LyCORIS to train T-LoRA for Flux, SDXL, Chroma and numerous other image models.

i've been working over a few weeks on extending T-LoRA to have framewise handling for video, but it's tricky and costly to develop.

anyway, thank you for the neat research repository. it would be good to know if you have a roadmap for future development.

[V-Soboleva]

Hello @bghira,

Thank you so much for reaching out and sharing this!

We are incredibly grateful to see T-LoRA being implemented on popular platforms like ComfyUI and Simpletuner. Having community support and integration is very important to us, and we really appreciate the work you’ve put into this node.

Regarding our roadmap: we are planning to update the repository in the near future. Our main goal is to refactor the code to be more "PEFT-style" making it universal and easier to apply to a wider range of base models.

We also have plans to extend T-LoRA to video models, however, it might take some time to get it right, but it is definitely on our to-do list.

Thanks again for your interest and contribution!