search

ShivamShrirao / diffusers

🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch
https://huggingface.co/docs/diffusers
Apache License 2.0
1.89k stars 505 forks source link

Is the train_dreambooth.py only trains the VAE part or also including diffusers models? #198

Closed wyang22 closed 1 year ago

wyang22 commented 1 year ago

After training, only there are some scheduler, VAE, U net and Text related folders & bin files saved, and there is no bin file generated in the scheduler folder, so want to confirm this, thanks

ShivamShrirao commented 1 year ago

It trains the unet and optionally the text encoder. VAE isn't trained in this and scheduler is basically an algorithm, not a model, it is not/can't be trained.

wyang22 commented 1 year ago

Thanks @ShivamShrirao

How can I config to train the text encoder? Can the Unet and the text encoder represent the whole stable diffusion model? If not, will the model capacity too small? Thanks