search

crowsonkb / k-diffusion

Karras et al. (2022) diffusion models for PyTorch
MIT License
2.26k stars 372 forks source link

Fine-tuning/compatible pre-trained models #16

Open kjhenner opened 2 years ago

kjhenner commented 2 years ago

Is there currently a way to fine-tune an existing model with k-diffusion? And/or are there any existing large pre-trained models that would be compatible with the patch feature?

neverix commented 2 years ago

Ideally there would be a converter + ignoring mismatching inputs/outputs in case the model has a different amount of channels

crowsonkb commented 1 year ago

The wrappers have the loss function implemented which should let you fine-tune but you will have to modify train.py for it.