[IP Adapters] introduce `ip_adapter_image_embeds` in the SD pipeline call (#6868)

* add: support for passing ip adapter image embeddings

* debugging

* make feature_extractor unloading conditioned on safety_checker

* better condition

* type annotation

* index to look into value slices

* more debugging

* debugging

* serialize embeddings dict

* better conditioning

* remove unnecessary prints.

* Update src/diffusers/loaders/ip_adapter.py
Co-authored-by: YiYi Xu <yixu310@gmail.com>

* make fix-copies and styling.

* styling and further copy fixing.

* fix: check_inputs call in controlnet sdxl img2img pipeline

---------
Co-authored-by: YiYi Xu <yixu310@gmail.com>

src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_k_diffusion.py

@@ -386,7 +386,6 @@ class StableDiffusionKDiffusionPipeline(DiffusionPipeline, TextualInversionLoade
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
     def check_inputs(
         self,
         prompt,

src/diffusers/pipelines/stable_diffusion_ldm3d/pipeline_stable_diffusion_ldm3d.py

@@ -438,33 +438,38 @@ class StableDiffusionLDM3DPipeline(
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):
-        if not isinstance(ip_adapter_image, list):
-            ip_adapter_image = [ip_adapter_image]
-
-        if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-            raise ValueError(
-                f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-            )
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, do_classifier_free_guidance, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
 
-        image_embeds = []
-        for single_ip_adapter_image, image_proj_layer in zip(
-            ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-        ):
-            output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-            single_image_embeds, single_negative_image_embeds = self.encode_image(
-                single_ip_adapter_image, device, 1, output_hidden_state
-            )
-            single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
 
-            if self.do_classifier_free_guidance:
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                single_image_embeds = single_image_embeds.to(device)
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
 
-            image_embeds.append(single_image_embeds)
+                if do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
 
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
         return image_embeds
 
     def run_safety_checker(self, image, device, dtype):
@@ -510,6 +515,8 @@ class StableDiffusionLDM3DPipeline(
         negative_prompt=None,
         prompt_embeds=None,
         negative_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
         callback_on_step_end_tensor_inputs=None,
     ):
         if height % 8 != 0 or width % 8 != 0:
@@ -553,6 +560,11 @@ class StableDiffusionLDM3DPipeline(
                     f" {negative_prompt_embeds.shape}."
                 )
 
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
         shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
         if isinstance(generator, list) and len(generator) != batch_size:
@@ -587,6 +599,7 @@ class StableDiffusionLDM3DPipeline(
         prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -633,6 +646,9 @@ class StableDiffusionLDM3DPipeline(
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*):
                 Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generated image. Choose between `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
@@ -665,7 +681,15 @@ class StableDiffusionLDM3DPipeline(
 
         # 1. Check inputs. Raise error if not correct
         self.check_inputs(
-            prompt, height, width, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds
+            prompt,
+            height,
+            width,
+            callback_steps,
+            negative_prompt,
+            prompt_embeds,
+            negative_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
         )
 
         # 2. Define call parameters
@@ -682,9 +706,13 @@ class StableDiffusionLDM3DPipeline(
         # corresponds to doing no classifier free guidance.
         do_classifier_free_guidance = guidance_scale > 1.0
 
-        if ip_adapter_image is not None:
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
             image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image, device, batch_size * num_images_per_prompt
+                ip_adapter_image,
+                ip_adapter_image_embeds,
+                do_classifier_free_guidance,
+                device,
+                batch_size * num_images_per_prompt,
             )
 
         # 3. Encode input prompt

src/diffusers/pipelines/stable_diffusion_panorama/pipeline_stable_diffusion_panorama.py

@@ -397,32 +397,38 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
             return image_embeds, uncond_image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):
-        if not isinstance(ip_adapter_image, list):
-            ip_adapter_image = [ip_adapter_image]
-
-        if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-            raise ValueError(
-                f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-            )
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
 
-        image_embeds = []
-        for single_ip_adapter_image, image_proj_layer in zip(
-            ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-        ):
-            output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-            single_image_embeds, single_negative_image_embeds = self.encode_image(
-                single_ip_adapter_image, device, 1, output_hidden_state
-            )
-            single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
 
-            if self.do_classifier_free_guidance:
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                single_image_embeds = single_image_embeds.to(device)
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
 
-            image_embeds.append(single_image_embeds)
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
 
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
         return image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
@@ -493,6 +499,8 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
         negative_prompt=None,
         prompt_embeds=None,
         negative_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
         callback_on_step_end_tensor_inputs=None,
     ):
         if height % 8 != 0 or width % 8 != 0:
@@ -536,6 +544,11 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
                     f" {negative_prompt_embeds.shape}."
                 )
 
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
         shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
@@ -592,6 +605,7 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
         prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -643,6 +657,9 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*):
                 Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generated image. Choose between `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
@@ -680,7 +697,15 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
 
         # 1. Check inputs. Raise error if not correct
         self.check_inputs(
-            prompt, height, width, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds
+            prompt,
+            height,
+            width,
+            callback_steps,
+            negative_prompt,
+            prompt_embeds,
+            negative_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
         )
 
         # 2. Define call parameters
@@ -697,9 +722,9 @@ class StableDiffusionPanoramaPipeline(DiffusionPipeline, TextualInversionLoaderM
         # corresponds to doing no classifier free guidance.
         do_classifier_free_guidance = guidance_scale > 1.0
 
-        if ip_adapter_image is not None:
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
             image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image, device, batch_size * num_images_per_prompt
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_images_per_prompt
             )
 
         # 3. Encode input prompt

src/diffusers/pipelines/stable_diffusion_safe/pipeline_stable_diffusion_safe.py

@@ -361,7 +361,6 @@ class StableDiffusionPipelineSafe(DiffusionPipeline, IPAdapterMixin):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
     def check_inputs(
         self,
         prompt,

src/diffusers/pipelines/stable_diffusion_sag/pipeline_stable_diffusion_sag.py

@@ -462,7 +462,6 @@ class StableDiffusionSAGPipeline(DiffusionPipeline, TextualInversionLoaderMixin,
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
     def check_inputs(
         self,
         prompt,

src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py

@@ -550,32 +550,38 @@ class StableDiffusionXLPipeline(
             return image_embeds, uncond_image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):
-        if not isinstance(ip_adapter_image, list):
-            ip_adapter_image = [ip_adapter_image]
-
-        if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-            raise ValueError(
-                f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-            )
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
 
-        image_embeds = []
-        for single_ip_adapter_image, image_proj_layer in zip(
-            ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-        ):
-            output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-            single_image_embeds, single_negative_image_embeds = self.encode_image(
-                single_ip_adapter_image, device, 1, output_hidden_state
-            )
-            single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
 
-            if self.do_classifier_free_guidance:
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                single_image_embeds = single_image_embeds.to(device)
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
 
-            image_embeds.append(single_image_embeds)
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
 
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
         return image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
@@ -609,6 +615,8 @@ class StableDiffusionXLPipeline(
         negative_prompt_embeds=None,
         pooled_prompt_embeds=None,
         negative_pooled_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
         callback_on_step_end_tensor_inputs=None,
     ):
         if height % 8 != 0 or width % 8 != 0:
@@ -675,6 +683,11 @@ class StableDiffusionXLPipeline(
                 "If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
             )
 
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
         shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
@@ -905,6 +918,7 @@ class StableDiffusionXLPipeline(
         pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
@@ -994,6 +1008,9 @@ class StableDiffusionXLPipeline(
                 weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
                 input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generate image. Choose between
                 [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
@@ -1092,6 +1109,8 @@ class StableDiffusionXLPipeline(
             negative_prompt_embeds,
             pooled_prompt_embeds,
             negative_pooled_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
             callback_on_step_end_tensor_inputs,
         )
 
@@ -1191,9 +1210,9 @@ class StableDiffusionXLPipeline(
         add_text_embeds = add_text_embeds.to(device)
         add_time_ids = add_time_ids.to(device).repeat(batch_size * num_images_per_prompt, 1)
 
-        if ip_adapter_image is not None:
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
             image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image, device, batch_size * num_images_per_prompt
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_images_per_prompt
             )
 
         # 8. Denoising loop
@@ -1236,7 +1255,7 @@ class StableDiffusionXLPipeline(
 
                 # predict the noise residual
                 added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
-                if ip_adapter_image is not None:
+                if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
                     added_cond_kwargs["image_embeds"] = image_embeds
                 noise_pred = self.unet(
                     latent_model_input,

src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py

@@ -575,6 +575,8 @@ class StableDiffusionXLImg2ImgPipeline(
         negative_prompt_2=None,
         prompt_embeds=None,
         negative_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
         callback_on_step_end_tensor_inputs=None,
     ):
         if strength < 0 or strength > 1:
@@ -637,6 +639,11 @@ class StableDiffusionXLImg2ImgPipeline(
                     f" {negative_prompt_embeds.shape}."
                 )
 
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
     def get_timesteps(self, num_inference_steps, strength, device, denoising_start=None):
         # get the original timestep using init_timestep
         if denoising_start is None:
@@ -767,32 +774,38 @@ class StableDiffusionXLImg2ImgPipeline(
             return image_embeds, uncond_image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):
-        if not isinstance(ip_adapter_image, list):
-            ip_adapter_image = [ip_adapter_image]
-
-        if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-            raise ValueError(
-                f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-            )
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
 
-        image_embeds = []
-        for single_ip_adapter_image, image_proj_layer in zip(
-            ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-        ):
-            output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-            single_image_embeds, single_negative_image_embeds = self.encode_image(
-                single_ip_adapter_image, device, 1, output_hidden_state
-            )
-            single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
 
-            if self.do_classifier_free_guidance:
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                single_image_embeds = single_image_embeds.to(device)
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
 
-            image_embeds.append(single_image_embeds)
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
 
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
         return image_embeds
 
     def _get_add_time_ids(
@@ -1047,6 +1060,7 @@ class StableDiffusionXLImg2ImgPipeline(
         pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
@@ -1145,6 +1159,9 @@ class StableDiffusionXLImg2ImgPipeline(
                 weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
                 input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generate image. Choose between
                 [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
@@ -1245,6 +1262,8 @@ class StableDiffusionXLImg2ImgPipeline(
             negative_prompt_2,
             prompt_embeds,
             negative_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
             callback_on_step_end_tensor_inputs,
         )
 
@@ -1365,9 +1384,9 @@ class StableDiffusionXLImg2ImgPipeline(
         add_text_embeds = add_text_embeds.to(device)
         add_time_ids = add_time_ids.to(device)
 
-        if ip_adapter_image is not None:
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
             image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image, device, batch_size * num_images_per_prompt
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_images_per_prompt
             )
 
         # 9. Denoising loop
@@ -1416,7 +1435,7 @@ class StableDiffusionXLImg2ImgPipeline(
 
                 # predict the noise residual
                 added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
-                if ip_adapter_image is not None:
+                if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
                     added_cond_kwargs["image_embeds"] = image_embeds
                 noise_pred = self.unet(
                     latent_model_input,

src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py

@@ -488,32 +488,38 @@ class StableDiffusionXLInpaintPipeline(
             return image_embeds, uncond_image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):
-        if not isinstance(ip_adapter_image, list):
-            ip_adapter_image = [ip_adapter_image]
-
-        if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-            raise ValueError(
-                f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-            )
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
 
-        image_embeds = []
-        for single_ip_adapter_image, image_proj_layer in zip(
-            ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-        ):
-            output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-            single_image_embeds, single_negative_image_embeds = self.encode_image(
-                single_ip_adapter_image, device, 1, output_hidden_state
-            )
-            single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
 
-            if self.do_classifier_free_guidance:
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                single_image_embeds = single_image_embeds.to(device)
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
 
-            image_embeds.append(single_image_embeds)
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
 
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
         return image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
@@ -784,6 +790,8 @@ class StableDiffusionXLInpaintPipeline(
         negative_prompt_2=None,
         prompt_embeds=None,
         negative_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
         callback_on_step_end_tensor_inputs=None,
         padding_mask_crop=None,
     ):
@@ -856,6 +864,11 @@ class StableDiffusionXLInpaintPipeline(
             if output_type != "pil":
                 raise ValueError(f"The output type should be PIL when inpainting mask crop, but is" f" {output_type}.")
 
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
     def prepare_latents(
         self,
         batch_size,
@@ -1288,6 +1301,7 @@ class StableDiffusionXLInpaintPipeline(
         pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
@@ -1397,6 +1411,9 @@ class StableDiffusionXLInpaintPipeline(
                 weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
                 input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             num_images_per_prompt (`int`, *optional*, defaults to 1):
                 The number of images to generate per prompt.
             eta (`float`, *optional*, defaults to 0.0):
@@ -1512,6 +1529,8 @@ class StableDiffusionXLInpaintPipeline(
             negative_prompt_2,
             prompt_embeds,
             negative_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
             callback_on_step_end_tensor_inputs,
             padding_mask_crop,
         )
@@ -1713,9 +1732,9 @@ class StableDiffusionXLInpaintPipeline(
         add_text_embeds = add_text_embeds.to(device)
         add_time_ids = add_time_ids.to(device)
 
-        if ip_adapter_image is not None:
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
             image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image, device, batch_size * num_images_per_prompt
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_images_per_prompt
             )
 
         # 11. Denoising loop
@@ -1766,7 +1785,7 @@ class StableDiffusionXLInpaintPipeline(
 
                 # predict the noise residual
                 added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
-                if ip_adapter_image is not None:
+                if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
                     added_cond_kwargs["image_embeds"] = image_embeds
                 noise_pred = self.unet(
                     latent_model_input,

src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py

@@ -564,32 +564,38 @@ class StableDiffusionXLAdapterPipeline(
             return image_embeds, uncond_image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):
-        if not isinstance(ip_adapter_image, list):
-            ip_adapter_image = [ip_adapter_image]
-
-        if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-            raise ValueError(
-                f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-            )
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
 
-        image_embeds = []
-        for single_ip_adapter_image, image_proj_layer in zip(
-            ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-        ):
-            output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-            single_image_embeds, single_negative_image_embeds = self.encode_image(
-                single_ip_adapter_image, device, 1, output_hidden_state
-            )
-            single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
 
-            if self.do_classifier_free_guidance:
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                single_image_embeds = single_image_embeds.to(device)
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
 
-            image_embeds.append(single_image_embeds)
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
 
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
         return image_embeds
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
@@ -624,6 +630,8 @@ class StableDiffusionXLAdapterPipeline(
         negative_prompt_embeds=None,
         pooled_prompt_embeds=None,
         negative_pooled_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
         callback_on_step_end_tensor_inputs=None,
     ):
         if height % 8 != 0 or width % 8 != 0:
@@ -690,6 +698,11 @@ class StableDiffusionXLAdapterPipeline(
                 "If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
             )
 
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
         shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
@@ -867,6 +880,7 @@ class StableDiffusionXLAdapterPipeline(
         pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -959,6 +973,9 @@ class StableDiffusionXLAdapterPipeline(
                 weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
                 input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generate image. Choose between
                 [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
@@ -1060,6 +1077,8 @@ class StableDiffusionXLAdapterPipeline(
             negative_prompt_embeds,
             pooled_prompt_embeds,
             negative_pooled_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
         )
 
         self._guidance_scale = guidance_scale
@@ -1096,9 +1115,9 @@ class StableDiffusionXLAdapterPipeline(
         )
 
         # 3.2 Encode ip_adapter_image
-        if ip_adapter_image is not None:
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
             image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image, device, batch_size * num_images_per_prompt
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_images_per_prompt
             )
 
         # 4. Prepare timesteps
@@ -1199,7 +1218,7 @@ class StableDiffusionXLAdapterPipeline(
 
                 added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
 
-                if ip_adapter_image is not None:
+                if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
                     added_cond_kwargs["image_embeds"] = image_embeds
 
                 # predict the noise residual

src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth.py

@@ -418,7 +418,6 @@ class TextToVideoSDPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lora
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
     def check_inputs(
         self,
         prompt,

src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth_img2img.py

@@ -495,7 +495,6 @@ class VideoToVideoSDPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.StableDiffusionImg2ImgPipeline.check_inputs
     def check_inputs(
         self,
         prompt,

src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero.py

@@ -447,7 +447,6 @@ class TextToVideoZeroPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
                         callback(step_idx, t, latents)
         return latents.clone().detach()
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
     def check_inputs(
         self,
         prompt,

src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero_sdxl.py

@@ -510,7 +510,6 @@ class TextToVideoZeroSDXLPipeline(
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.check_inputs
     def check_inputs(
         self,
         prompt,