Musawar Ali

did u find any solution.

You can use gradio and call the sampling function in your script

It is already controlling LDMs

You can you use stable diffusion inpaint for inference

Hello! First of all lock your SD. Which seems to to be unlocked. Second use increased batch size i.e 64, using the gradient accumulation. Increased batch size will cost you...

These are train images, you can generate validation images also by some modifications in logger.py

Yes you can have it, you need to have separate data loader for validation loss

You can generate the validation images also by changing the split to train in cldm/logger.py. You must load validation dataset at the start of training

I have quite similar question. Do they drop the 50% of the text prompts during the training process?

the dimensions should be multiple of 64..