diffusers 0.23.0__py3-none-any.whl → 0.24.0__py3-none-any.whl

diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py

@@ -20,12 +20,23 @@ import numpy as np
 import PIL.Image
 import torch
 import torch.nn.functional as F
-from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer
+from transformers import (
+    CLIPImageProcessor,
+    CLIPTextModel,
+    CLIPTextModelWithProjection,
+    CLIPTokenizer,
+    CLIPVisionModelWithProjection,
+)
 
 from diffusers.utils.import_utils import is_invisible_watermark_available
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FromSingleFileMixin, StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import (
+    FromSingleFileMixin,
+    IPAdapterMixin,
+    StableDiffusionXLLoraLoaderMixin,
+    TextualInversionLoaderMixin,
+)
 from ...models import AutoencoderKL, ControlNetModel, UNet2DConditionModel
 from ...models.attention_processor import (
     AttnProcessor2_0,
@@ -35,7 +46,14 @@ from ...models.attention_processor import (
 )
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
-from ...utils import USE_PEFT_BACKEND, logging, replace_example_docstring, scale_lora_layers, unscale_lora_layers
+from ...utils import (
+    USE_PEFT_BACKEND,
+    deprecate,
+    logging,
+    replace_example_docstring,
+    scale_lora_layers,
+    unscale_lora_layers,
+)
 from ...utils.torch_utils import is_compiled_module, is_torch_version, randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from ..stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
@@ -97,7 +115,11 @@ EXAMPLE_DOC_STRING = """
 
 
 class StableDiffusionXLControlNetPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionXLLoraLoaderMixin, FromSingleFileMixin
+    DiffusionPipeline,
+    TextualInversionLoaderMixin,
+    StableDiffusionXLLoraLoaderMixin,
+    IPAdapterMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion XL with ControlNet guidance.
@@ -139,9 +161,18 @@ class StableDiffusionXLControlNetPipeline(
             watermark output images. If not defined, it defaults to `True` if the package is installed; otherwise no
             watermarker is used.
     """
+
     # leave controlnet out on purpose because it iterates with unet
     model_cpu_offload_seq = "text_encoder->text_encoder_2->unet->vae"
-    _optional_components = ["tokenizer", "tokenizer_2", "text_encoder", "text_encoder_2"]
+    _optional_components = [
+        "tokenizer",
+        "tokenizer_2",
+        "text_encoder",
+        "text_encoder_2",
+        "feature_extractor",
+        "image_encoder",
+    ]
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
 
     def __init__(
         self,
@@ -155,6 +186,8 @@ class StableDiffusionXLControlNetPipeline(
         scheduler: KarrasDiffusionSchedulers,
         force_zeros_for_empty_prompt: bool = True,
         add_watermarker: Optional[bool] = None,
+        feature_extractor: CLIPImageProcessor = None,
+        image_encoder: CLIPVisionModelWithProjection = None,
     ):
         super().__init__()
 
@@ -170,6 +203,8 @@ class StableDiffusionXLControlNetPipeline(
             unet=unet,
             controlnet=controlnet,
             scheduler=scheduler,
+            feature_extractor=feature_extractor,
+            image_encoder=image_encoder,
         )
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
@@ -453,6 +488,20 @@ class StableDiffusionXLControlNetPipeline(
 
         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):
+        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)
+        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
@@ -486,15 +535,21 @@ class StableDiffusionXLControlNetPipeline(
         controlnet_conditioning_scale=1.0,
         control_guidance_start=0.0,
         control_guidance_end=1.0,
+        callback_on_step_end_tensor_inputs=None,
     ):
-        if (callback_steps is None) or (
-            callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0)
-        ):
+        if callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0):
             raise ValueError(
                 f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
                 f" {type(callback_steps)}."
             )
 
+        if callback_on_step_end_tensor_inputs is not None and not all(
+            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
+        ):
+            raise ValueError(
+                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
+            )
+
         if prompt is not None and prompt_embeds is not None:
             raise ValueError(
                 f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
@@ -791,6 +846,58 @@ class StableDiffusionXLControlNetPipeline(
         """Disables the FreeU mechanism if enabled."""
         self.unet.disable_freeu()
 
+    # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
+    def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
+        """
+        See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
+
+        Args:
+            timesteps (`torch.Tensor`):
+                generate embedding vectors at these timesteps
+            embedding_dim (`int`, *optional*, defaults to 512):
+                dimension of the embeddings to generate
+            dtype:
+                data type of the generated embeddings
+
+        Returns:
+            `torch.FloatTensor`: Embedding vectors with shape `(len(timesteps), embedding_dim)`
+        """
+        assert len(w.shape) == 1
+        w = w * 1000.0
+
+        half_dim = embedding_dim // 2
+        emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
+        emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
+        emb = w.to(dtype)[:, None] * emb[None, :]
+        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
+        if embedding_dim % 2 == 1:  # zero pad
+            emb = torch.nn.functional.pad(emb, (0, 1))
+        assert emb.shape == (w.shape[0], embedding_dim)
+        return emb
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def clip_skip(self):
+        return self._clip_skip
+
+    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
+    # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
+    # corresponds to doing no classifier free guidance.
+    @property
+    def do_classifier_free_guidance(self):
+        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
+
+    @property
+    def cross_attention_kwargs(self):
+        return self._cross_attention_kwargs
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
@@ -812,10 +919,9 @@ class StableDiffusionXLControlNetPipeline(
         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,
-        callback_steps: int = 1,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
         guess_mode: bool = False,
@@ -828,6 +934,9 @@ class StableDiffusionXLControlNetPipeline(
         negative_crops_coords_top_left: Tuple[int, int] = (0, 0),
         negative_target_size: Optional[Tuple[int, int]] = None,
         clip_skip: Optional[int] = None,
+        callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
+        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
+        **kwargs,
     ):
         r"""
         The call function to the pipeline for generation.
@@ -891,17 +1000,12 @@ class StableDiffusionXLControlNetPipeline(
                 Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs (prompt
                 weighting). If not provided, pooled `negative_prompt_embeds` are 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 generated image. Choose between `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
                 Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
                 plain tuple.
-            callback (`Callable`, *optional*):
-                A function that calls every `callback_steps` steps during inference. The function is called with the
-                following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
-            callback_steps (`int`, *optional*, defaults to 1):
-                The frequency at which the `callback` function is called. If not specified, the callback is called at
-                every step.
             cross_attention_kwargs (`dict`, *optional*):
                 A kwargs dictionary that if specified is passed along to the [`AttentionProcessor`] as defined in
                 [`self.processor`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
@@ -948,6 +1052,15 @@ class StableDiffusionXLControlNetPipeline(
             clip_skip (`int`, *optional*):
                 Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
                 the output of the pre-final layer will be used for computing the prompt embeddings.
+            callback_on_step_end (`Callable`, *optional*):
+                A function that calls at the end of each denoising steps during the inference. The function is called
+                with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,
+                callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by
+                `callback_on_step_end_tensor_inputs`.
+            callback_on_step_end_tensor_inputs (`List`, *optional*):
+                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
+                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
+                `._callback_tensor_inputs` attribute of your pipeine class.
 
         Examples:
 
@@ -956,6 +1069,23 @@ class StableDiffusionXLControlNetPipeline(
                 If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] is returned,
                 otherwise a `tuple` is returned containing the output images.
         """
+
+        callback = kwargs.pop("callback", None)
+        callback_steps = kwargs.pop("callback_steps", None)
+
+        if callback is not None:
+            deprecate(
+                "callback",
+                "1.0.0",
+                "Passing `callback` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
+            )
+        if callback_steps is not None:
+            deprecate(
+                "callback_steps",
+                "1.0.0",
+                "Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
+            )
+
         controlnet = self.controlnet._orig_mod if is_compiled_module(self.controlnet) else self.controlnet
 
         # align format for control guidance
@@ -965,9 +1095,10 @@ class StableDiffusionXLControlNetPipeline(
             control_guidance_end = len(control_guidance_start) * [control_guidance_end]
         elif not isinstance(control_guidance_start, list) and not isinstance(control_guidance_end, list):
             mult = len(controlnet.nets) if isinstance(controlnet, MultiControlNetModel) else 1
-            control_guidance_start, control_guidance_end = mult * [control_guidance_start], mult * [
-                control_guidance_end
-            ]
+            control_guidance_start, control_guidance_end = (
+                mult * [control_guidance_start],
+                mult * [control_guidance_end],
+            )
 
         # 1. Check inputs. Raise error if not correct
         self.check_inputs(
@@ -984,8 +1115,13 @@ class StableDiffusionXLControlNetPipeline(
             controlnet_conditioning_scale,
             control_guidance_start,
             control_guidance_end,
+            callback_on_step_end_tensor_inputs,
         )
 
+        self._guidance_scale = guidance_scale
+        self._clip_skip = clip_skip
+        self._cross_attention_kwargs = cross_attention_kwargs
+
         # 2. Define call parameters
         if prompt is not None and isinstance(prompt, str):
             batch_size = 1
@@ -995,10 +1131,6 @@ class StableDiffusionXLControlNetPipeline(
             batch_size = prompt_embeds.shape[0]
 
         device = self._execution_device
-        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
-        # corresponds to doing no classifier free guidance.
-        do_classifier_free_guidance = guidance_scale > 1.0
 
         if isinstance(controlnet, MultiControlNetModel) and isinstance(controlnet_conditioning_scale, float):
             controlnet_conditioning_scale = [controlnet_conditioning_scale] * len(controlnet.nets)
@@ -1010,9 +1142,9 @@ class StableDiffusionXLControlNetPipeline(
         )
         guess_mode = guess_mode or global_pool_conditions
 
-        # 3. Encode input prompt
+        # 3.1 Encode input prompt
         text_encoder_lora_scale = (
-            cross_attention_kwargs.get("scale", None) if cross_attention_kwargs is not None else None
+            self.cross_attention_kwargs.get("scale", None) if self.cross_attention_kwargs is not None else None
         )
         (
             prompt_embeds,
@@ -1024,7 +1156,7 @@ class StableDiffusionXLControlNetPipeline(
             prompt_2,
             device,
             num_images_per_prompt,
-            do_classifier_free_guidance,
+            self.do_classifier_free_guidance,
             negative_prompt,
             negative_prompt_2,
             prompt_embeds=prompt_embeds,
@@ -1032,9 +1164,15 @@ class StableDiffusionXLControlNetPipeline(
             pooled_prompt_embeds=pooled_prompt_embeds,
             negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
             lora_scale=text_encoder_lora_scale,
-            clip_skip=clip_skip,
+            clip_skip=self.clip_skip,
         )
 
+        # 3.2 Encode ip_adapter_image
+        if ip_adapter_image is not None:
+            image_embeds, negative_image_embeds = self.encode_image(ip_adapter_image, device, num_images_per_prompt)
+            if self.do_classifier_free_guidance:
+                image_embeds = torch.cat([negative_image_embeds, image_embeds])
+
         # 4. Prepare image
         if isinstance(controlnet, ControlNetModel):
             image = self.prepare_image(
@@ -1045,7 +1183,7 @@ class StableDiffusionXLControlNetPipeline(
                 num_images_per_prompt=num_images_per_prompt,
                 device=device,
                 dtype=controlnet.dtype,
-                do_classifier_free_guidance=do_classifier_free_guidance,
+                do_classifier_free_guidance=self.do_classifier_free_guidance,
                 guess_mode=guess_mode,
             )
             height, width = image.shape[-2:]
@@ -1061,7 +1199,7 @@ class StableDiffusionXLControlNetPipeline(
                     num_images_per_prompt=num_images_per_prompt,
                     device=device,
                     dtype=controlnet.dtype,
-                    do_classifier_free_guidance=do_classifier_free_guidance,
+                    do_classifier_free_guidance=self.do_classifier_free_guidance,
                     guess_mode=guess_mode,
                 )
 
@@ -1075,6 +1213,7 @@ class StableDiffusionXLControlNetPipeline(
         # 5. Prepare timesteps
         self.scheduler.set_timesteps(num_inference_steps, device=device)
         timesteps = self.scheduler.timesteps
+        self._num_timesteps = len(timesteps)
 
         # 6. Prepare latent variables
         num_channels_latents = self.unet.config.in_channels
@@ -1089,6 +1228,14 @@ class StableDiffusionXLControlNetPipeline(
             latents,
         )
 
+        # 6.5 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)
+            timestep_cond = self.get_guidance_scale_embedding(
+                guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim
+            ).to(device=device, dtype=latents.dtype)
+
         # 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)
 
@@ -1133,7 +1280,7 @@ class StableDiffusionXLControlNetPipeline(
         else:
             negative_add_time_ids = add_time_ids
 
-        if do_classifier_free_guidance:
+        if self.do_classifier_free_guidance:
             prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)
             add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)
             add_time_ids = torch.cat([negative_add_time_ids, add_time_ids], dim=0)
@@ -1154,13 +1301,13 @@ class StableDiffusionXLControlNetPipeline(
                 if (is_unet_compiled and is_controlnet_compiled) and is_torch_higher_equal_2_1:
                     torch._inductor.cudagraph_mark_step_begin()
                 # 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 self.do_classifier_free_guidance else latents
                 latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
 
                 added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
 
                 # controlnet(s) inference
-                if guess_mode and do_classifier_free_guidance:
+                if guess_mode and self.do_classifier_free_guidance:
                     # Infer ControlNet only for the conditional batch.
                     control_model_input = latents
                     control_model_input = self.scheduler.scale_model_input(control_model_input, t)
@@ -1193,19 +1340,23 @@ class StableDiffusionXLControlNetPipeline(
                     return_dict=False,
                 )
 
-                if guess_mode and do_classifier_free_guidance:
+                if guess_mode and self.do_classifier_free_guidance:
                     # Infered ControlNet only for the conditional batch.
                     # To apply the output of ControlNet to both the unconditional and conditional batches,
                     # add 0 to the unconditional batch to keep it unchanged.
                     down_block_res_samples = [torch.cat([torch.zeros_like(d), d]) for d in down_block_res_samples]
                     mid_block_res_sample = torch.cat([torch.zeros_like(mid_block_res_sample), mid_block_res_sample])
 
+                if ip_adapter_image is not None:
+                    added_cond_kwargs["image_embeds"] = image_embeds
+
                 # predict the noise residual
                 noise_pred = self.unet(
                     latent_model_input,
                     t,
                     encoder_hidden_states=prompt_embeds,
-                    cross_attention_kwargs=cross_attention_kwargs,
+                    timestep_cond=timestep_cond,
+                    cross_attention_kwargs=self.cross_attention_kwargs,
                     down_block_additional_residuals=down_block_res_samples,
                     mid_block_additional_residual=mid_block_res_sample,
                     added_cond_kwargs=added_cond_kwargs,
@@ -1213,13 +1364,23 @@ class StableDiffusionXLControlNetPipeline(
                 )[0]
 
                 # perform guidance
-                if do_classifier_free_guidance:
+                if self.do_classifier_free_guidance:
                     noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                     noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
                 # compute the previous noisy sample x_t -> x_t-1
                 latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
+                if callback_on_step_end is not None:
+                    callback_kwargs = {}
+                    for k in callback_on_step_end_tensor_inputs:
+                        callback_kwargs[k] = locals()[k]
+                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
+
+                    latents = callback_outputs.pop("latents", latents)
+                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
+                    negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)
+
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
                     progress_bar.update()