Added methods to the dataloader to automatically generate controls for line, mask, inpainting, depth, and pose.

toolkit/stable_diffusion_model.py

@@ -3009,6 +3009,8 @@ class StableDiffusion:
             active_modules = ['vae']
         if device_state_preset in ['cache_clip']:
             active_modules = ['clip']
+        if device_state_preset in ['unload']:
+            active_modules = []
         if device_state_preset in ['generate']:
             active_modules = ['vae', 'unet', 'text_encoder', 'adapter', 'refiner_unet']