Created a size agnostic feature encoder (SAFE) model to be trained in replace of CLIP for ip adapters. It is mostly conv layers so will hopefully be able to handle facial features better than clip can. Also bug fixes

toolkit/stable_diffusion_model.py

@@ -1447,7 +1447,7 @@ class StableDiffusion:
             }
         if self.adapter is not None:
             if isinstance(self.adapter, IPAdapter):
-                requires_grad = self.adapter.adapter_modules.training
+                requires_grad = self.adapter.image_proj_model.training
                 adapter_device = self.unet.device
             elif isinstance(self.adapter, T2IAdapter):
                 requires_grad = self.adapter.adapter.conv_in.weight.requires_grad