[Bug]: How to input a mask image when predicting image after masked training?

[qingqinggu]

What happened?

When masked training STABLE_DIFFUSION_XL_10_BASE, the black and white masked images are named as" xxx-masklabel.png" as mentioned in QuickStartGuide.md.

1. how and where to input that masked image when predicting image after masked training? No python file like "predict_mask.py" is found in this project. The masked image needs to be converted into [0, 1] firstly.

2. Can STABLE_DIFFUSION_XL_10_BASE be used in masked training? Or only inpainting model?

In BaseStableDiffusionXLSetup.py Line 358, as follows, only inpainting_model can use batch['latent_mask'] ''' if config.model_type.has_mask_input() and config.model_type.has_conditioning_image_input(): latent_input = torch.concat( [scaled_noisy_latent_image, batch['latent_mask'], scaled_latent_conditioning_image], 1 ) else: latent_input = scaled_noisy_latent_image '''

What did you expect would happen?

Please provide a python file to tell me how and where to input that masked image when predicting image after masked training. Thank you!

Relevant log output

No response

Output of pip freeze

No response