.load_ip_adapter in StableDiffusionXLAdapterPipeline (#6246)

* Added testing notebook and .load_ip_adapter to XLAdapterPipeline

* Added annotations

* deleted testing notebook

* Update src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py

Co-authored-by: YiYi Xu <[email protected]>

* code clean up

* Add feature_extractor and image_encoder to components

---------

Co-authored-by: YiYi Xu <[email protected]>

Author: jquintanilla4

src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py

@@ -18,11 +18,22 @@
 import numpy as np
 import PIL.Image
 import torch
-from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer
+from transformers import (
+    CLIPImageProcessor,
+    CLIPTextModel,
+    CLIPTextModelWithProjection,
+    CLIPTokenizer,
+    CLIPVisionModelWithProjection,
+)
 
-from ...image_processor import VaeImageProcessor
-from ...loaders import FromSingleFileMixin, StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
-from ...models import AutoencoderKL, MultiAdapter, T2IAdapter, UNet2DConditionModel
+from ...image_processor import PipelineImageInput, VaeImageProcessor
+from ...loaders import (
+    FromSingleFileMixin,
+    IPAdapterMixin,
+    StableDiffusionXLLoraLoaderMixin,
+    TextualInversionLoaderMixin,
+)
+from ...models import AutoencoderKL, ImageProjection, MultiAdapter, T2IAdapter, UNet2DConditionModel
 from ...models.attention_processor import (
     AttnProcessor2_0,
     LoRAAttnProcessor2_0,
@@ -169,7 +180,11 @@ def retrieve_timesteps(
 
 
 class StableDiffusionXLAdapterPipeline(
-    DiffusionPipeline, FromSingleFileMixin, StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
+    DiffusionPipeline,
+    TextualInversionLoaderMixin,
+    StableDiffusionXLLoraLoaderMixin,
+    IPAdapterMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion augmented with T2I-Adapter
@@ -183,6 +198,7 @@ class StableDiffusionXLAdapterPipeline(
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
         - [`~loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
         - [`~loaders.StableDiffusionXLLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
     Args:
         adapter ([`T2IAdapter`] or [`MultiAdapter`] or `List[T2IAdapter]`):
@@ -211,8 +227,15 @@ class StableDiffusionXLAdapterPipeline(
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
 
-    model_cpu_offload_seq = "text_encoder->text_encoder_2->unet->vae"
-    _optional_components = ["tokenizer", "tokenizer_2", "text_encoder", "text_encoder_2"]
+    model_cpu_offload_seq = "text_encoder->text_encoder_2->image_encoder->unet->vae"
+    _optional_components = [
+        "tokenizer",
+        "tokenizer_2",
+        "text_encoder",
+        "text_encoder_2",
+        "feature_extractor",
+        "image_encoder",
+    ]
 
     def __init__(
         self,
@@ -225,6 +248,8 @@ def __init__(
         adapter: Union[T2IAdapter, MultiAdapter, List[T2IAdapter]],
         scheduler: KarrasDiffusionSchedulers,
         force_zeros_for_empty_prompt: bool = True,
+        feature_extractor: CLIPImageProcessor = None,
+        image_encoder: CLIPVisionModelWithProjection = None,
     ):
         super().__init__()
 
@@ -237,6 +262,8 @@ def __init__(
             unet=unet,
             adapter=adapter,
             scheduler=scheduler,
+            feature_extractor=feature_extractor,
+            image_encoder=image_encoder,
         )
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
@@ -511,6 +538,31 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
+    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
+        dtype = next(self.image_encoder.parameters()).dtype
+
+        if not isinstance(image, torch.Tensor):
+            image = self.feature_extractor(image, return_tensors="pt").pixel_values
+
+        image = image.to(device=device, dtype=dtype)
+        if output_hidden_states:
+            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
+            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
+            uncond_image_enc_hidden_states = self.image_encoder(
+                torch.zeros_like(image), output_hidden_states=True
+            ).hidden_states[-2]
+            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
+                num_images_per_prompt, dim=0
+            )
+            return image_enc_hidden_states, uncond_image_enc_hidden_states
+        else:
+            image_embeds = self.image_encoder(image).image_embeds
+            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
+            uncond_image_embeds = torch.zeros_like(image_embeds)
+
+            return image_embeds, uncond_image_embeds
+
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
@@ -768,7 +820,7 @@ def __call__(
         self,
         prompt: Union[str, List[str]] = None,
         prompt_2: Optional[Union[str, List[str]]] = None,
-        image: Union[torch.Tensor, PIL.Image.Image, List[PIL.Image.Image]] = None,
+        image: PipelineImageInput = None,
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 50,
@@ -785,6 +837,7 @@ def __call__(
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
         pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
+        ip_adapter_image: Optional[PipelineImageInput] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -876,6 +929,7 @@ def __call__(
                 Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 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.
             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`.
@@ -991,7 +1045,7 @@ def __call__(
 
         device = self._execution_device
 
-        # 3. Encode input prompt
+        # 3.1 Encode input prompt
         (
             prompt_embeds,
             negative_prompt_embeds,
@@ -1012,6 +1066,15 @@ def __call__(
             clip_skip=clip_skip,
         )
 
+        # 3.2 Encode ip_adapter_image
+        if ip_adapter_image is not None:
+            output_hidden_state = False if isinstance(self.unet.encoder_hid_proj, ImageProjection) else True
+            image_embeds, negative_image_embeds = self.encode_image(
+                ip_adapter_image, device, num_images_per_prompt, output_hidden_state
+            )
+            if self.do_classifier_free_guidance:
+                image_embeds = torch.cat([negative_image_embeds, image_embeds])
+
         # 4. Prepare timesteps
         timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)
 
@@ -1028,10 +1091,10 @@ def __call__(
             latents,
         )
 
-        # 6. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
+        # 6.1 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)
 
-        # 6.5 Optionally get Guidance Scale Embedding
+        # 6.2 Optionally get Guidance Scale Embedding
         timestep_cond = None
         if self.unet.config.time_cond_proj_dim is not None:
             guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(batch_size * num_images_per_prompt)
@@ -1090,8 +1153,7 @@ def __call__(
 
         # 8. Denoising loop
         num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
-
-        # 7.1 Apply denoising_end
+        # Apply denoising_end
         if denoising_end is not None and isinstance(denoising_end, float) and denoising_end > 0 and denoising_end < 1:
             discrete_timestep_cutoff = int(
                 round(
@@ -1109,9 +1171,12 @@ def __call__(
 
                 latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
 
-                # predict the noise residual
                 added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
 
+                if ip_adapter_image is not None:
+                    added_cond_kwargs["image_embeds"] = image_embeds
+
+                # predict the noise residual
                 if i < int(num_inference_steps * adapter_conditioning_factor):
                     down_intrablock_additional_residuals = [state.clone() for state in adapter_state]
                 else:
@@ -1123,9 +1188,9 @@ def __call__(
                     encoder_hidden_states=prompt_embeds,
                     timestep_cond=timestep_cond,
                     cross_attention_kwargs=cross_attention_kwargs,
+                    down_intrablock_additional_residuals=down_intrablock_additional_residuals,
                     added_cond_kwargs=added_cond_kwargs,
                     return_dict=False,
-                    down_intrablock_additional_residuals=down_intrablock_additional_residuals,
                 )[0]
 
                 # perform guidance

tests/pipelines/stable_diffusion_xl/test_stable_diffusion_xl_adapter.py

@@ -159,7 +159,8 @@ def get_dummy_components(self, adapter_type="full_adapter_xl", time_cond_proj_di
             "text_encoder_2": text_encoder_2,
             "tokenizer_2": tokenizer_2,
             # "safety_checker": None,
-            # "feature_extractor": None,
+            "feature_extractor": None,
+            "image_encoder": None,
         }
         return components
 
@@ -265,7 +266,8 @@ def get_dummy_components_with_full_downscaling(self, adapter_type="full_adapter_
             "text_encoder_2": text_encoder_2,
             "tokenizer_2": tokenizer_2,
             # "safety_checker": None,
-            # "feature_extractor": None,
+            "feature_extractor": None,
+            "image_encoder": None,
         }
         return components