update for image batch processing

pipeline.py

@@ -856,7 +856,8 @@ class StableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline, TextualInversi
         )
 
         # 4. Prepare image, and controlnet_conditioning_image
-        image = prepare_image(image)
+        # image = prepare_image(image)
+        images = [prepare_image(img) for img in image]
 
         # condition image(s)
         if isinstance(self.controlnet, ControlNetModel):
@@ -897,15 +898,27 @@ class StableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline, TextualInversi
         latent_timestep = timesteps[:1].repeat(batch_size * num_images_per_prompt)
 
         # 6. Prepare latent variables
-        latents = self.prepare_latents(
-            image,
+        # latents = self.prepare_latents(
+        #     image,
+        #     latent_timestep,
+        #     batch_size,
+        #     num_images_per_prompt,
+        #     prompt_embeds.dtype,
+        #     device,
+        #     generator,
+        # )
+
+        latents = [self.prepare_latents(
+            img,
             latent_timestep,
             batch_size,
             num_images_per_prompt,
             prompt_embeds.dtype,
             device,
             generator,
-        )
+        ) for img in images]
+        latents = torch.cat(latents)
+
 
         # 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
         extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
@@ -915,7 +928,9 @@ class StableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline, TextualInversi
         with self.progress_bar(total=num_inference_steps) as progress_bar:
             for i, t in enumerate(timesteps):
                 # expand the latents if we are doing classifier free guidance
-                latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
+                # latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
+                latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents.clone()
+
 
                 latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)