enable basic img2img

CHANGELOG.md

@@ -2,12 +2,16 @@
 
 ## Update for 07/10/2023
 
+Service release with some fixes and enhancements:
+
 - diffusers:
   - option to move base and/or refiner model to cpu to free up vram  
   - model downloader options to specify model variant / revision / mirror  
   - now you can download `fp16` variant directly for reduced memory footprint  
+  - basic **img2img** workflow (*sketch* and *inpaint* are not supported yet)  
+    note that **sd-xl** img2img workflows are architecturaly different so it will take longer to implement  
   - updated hints for settings  
-- extra network:
+- extra networks:
   - fix corrupt display on refesh when new extra network type found  
   - additional ui tweaks  
   - generate thumbnails from previews only if preview resolution is above 1k

TODO.md

@@ -62,3 +62,4 @@ Tech that can be integrated as part of the core workflow...
 - upate `gradio`
 - extra network refresh breaks if new extra network type found
 - [sd-xl lora](https://civitai.com/models/104913/fcstyledxl)
+- sd-xl img2img with configurable steps

modules/processing.py

@@ -225,14 +225,13 @@ class StableDiffusionProcessing:
         # identify itself with a field common to all models. The conditioning_key is also hybrid.
         if backend == Backend.DIFFUSERS:
             log.warning('Diffusers not implemented: img2img_image_conditioning')
-            return None
         if isinstance(self.sd_model, LatentDepth2ImageDiffusion):
             return self.depth2img_image_conditioning(source_image)
-        if self.sd_model.cond_stage_key == "edit":
+        if hasattr(self.sd_model, 'cond_stage_key') and self.sd_model.cond_stage_key == "edit":
             return self.edit_image_conditioning(source_image)
-        if self.sampler.conditioning_key in {'hybrid', 'concat'}:
+        if hasattr(self.sampler, 'conditioning_key') and self.sampler.conditioning_key in {'hybrid', 'concat'}:
             return self.inpainting_image_conditioning(source_image, latent_image, image_mask=image_mask)
-        if self.sampler.conditioning_key == "crossattn-adm":
+        if hasattr(self.sampler, 'conditioning_key') and self.sampler.conditioning_key == "crossattn-adm":
             return self.unclip_image_conditioning(source_image)
         # Dummy zero conditioning if we're not using inpainting or depth model.
         return latent_image.new_zeros(latent_image.shape[0], 5, 1, 1)
@@ -752,7 +751,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
 
                 x_samples_ddim = output.images
 
-                if p.enable_hr:
+                if p.is_hr_pass:
                     log.warning('Diffusers not implemented: hires fix')
 
                 if lora_state['active']:
@@ -1130,8 +1129,8 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
         if backend == Backend.ORIGINAL:
             self.init_latent = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image))
         else:
-            # we don't pre-encode the latents for diffusers to allow the UI to stay general for different model types
-            self.init_latent = None
+            # TODO Diffusers don't pre-encode the latents for diffusers to allow the UI to stay general for different model types
+            self.init_latent = torch.Tensor(1)
 
         if self.resize_mode == 3:
             self.init_latent = torch.nn.functional.interpolate(self.init_latent, size=(self.height // opt_f, self.width // opt_f), mode="bilinear")

modules/sd_models.py

@@ -806,12 +806,21 @@ def set_diffuser_pipe(pipe, new_pipe_type):
 
     pipe_name = pipe.__class__.__name__
     pipe_name = pipe_name.replace("Img2Img", "").replace("Inpaint", "")
+    new_pipe_cls_str = None
     if new_pipe_type == DiffusersTaskType.TEXT_2_IMAGE:
         new_pipe_cls_str = pipe_name
     elif new_pipe_type == DiffusersTaskType.IMAGE_2_IMAGE:
-        new_pipe_cls_str = pipe_name.replace("Pipeline", "Img2ImgPipeline")
+        tmp_pipe_name = pipe_name.replace("Pipeline", "Img2ImgPipeline")
+        if hasattr(diffusers, tmp_pipe_name):
+            new_pipe_cls_str = pipe_name.replace("Pipeline", "Img2ImgPipeline")
     elif new_pipe_type == DiffusersTaskType.INPAINTING:
-        new_pipe_cls_str = pipe_name.replace("Pipeline", "InpaintPipeline")
+        tmp_pipe_name = pipe_name.replace("Pipeline", "InpaintPipeline")
+        if hasattr(diffusers, tmp_pipe_name):
+            new_pipe_cls_str = pipe_name.replace("Pipeline", "InpaintPipeline")
+
+    if new_pipe_cls_str is None:
+        shared.log.warning(f'Diffusers unknown pipeline: {tmp_pipe_name}')
+        new_pipe_cls_str = pipe_name
 
     new_pipe_cls = getattr(diffusers, new_pipe_cls_str)