diffusers 0.27.1__py3-none-any.whl → 0.32.2__py3-none-any.whl

diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting.py

@@ -9,13 +9,12 @@ import PIL.Image
 import torch
 from transformers import CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
-from ...loaders import LoraLoaderMixin
+from ...loaders import StableDiffusionLoraLoaderMixin
 from ...models import UNet2DConditionModel
 from ...schedulers import DDPMScheduler
 from ...utils import (
     BACKENDS_MAPPING,
     PIL_INTERPOLATION,
-    is_accelerate_available,
     is_bs4_available,
     is_ftfy_available,
     logging,
@@ -112,7 +111,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
+class IFInpaintingPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
     tokenizer: T5Tokenizer
     text_encoder: T5EncoderModel
 
@@ -142,6 +141,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
 
     _optional_components = ["tokenizer", "text_encoder", "safety_checker", "feature_extractor", "watermarker"]
     model_cpu_offload_seq = "text_encoder->unet"
+    _exclude_from_cpu_offload = ["watermarker"]
 
     def __init__(
         self,
@@ -183,21 +183,6 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.remove_all_hooks
-    def remove_all_hooks(self):
-        if is_accelerate_available():
-            from accelerate.hooks import remove_hook_from_module
-        else:
-            raise ImportError("Please install accelerate via `pip install accelerate`")
-
-        for model in [self.text_encoder, self.unet, self.safety_checker]:
-            if model is not None:
-                remove_hook_from_module(model, recurse=True)
-
-        self.unet_offload_hook = None
-        self.text_encoder_offload_hook = None
-        self.final_offload_hook = None
-
     @torch.no_grad()
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.encode_prompt
     def encode_prompt(
@@ -207,8 +192,8 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
         num_images_per_prompt: int = 1,
         device: Optional[torch.device] = None,
         negative_prompt: Optional[Union[str, List[str]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         clean_caption: bool = False,
     ):
         r"""
@@ -227,10 +212,10 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
                 The prompt or prompts not to guide the image generation. If not defined, one has to pass
                 `negative_prompt_embeds`. instead. If not defined, one has to pass `negative_prompt_embeds`. instead.
                 Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`).
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -365,9 +350,6 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
             nsfw_detected = None
             watermark_detected = None
 
-            if hasattr(self, "unet_offload_hook") and self.unet_offload_hook is not None:
-                self.unet_offload_hook.offload()
-
         return image, nsfw_detected, watermark_detected
 
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.prepare_extra_step_kwargs
@@ -446,7 +428,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
             and not isinstance(check_image_type, np.ndarray)
         ):
             raise ValueError(
-                "`image` has to be of type `torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
+                "`image` has to be of type `torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
                 f" {type(check_image_type)}"
             )
 
@@ -477,7 +459,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
             and not isinstance(check_image_type, np.ndarray)
         ):
             raise ValueError(
-                "`mask_image` has to be of type `torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
+                "`mask_image` has to be of type `torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
                 f" {type(check_image_type)}"
             )
 
@@ -654,7 +636,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
 
             for image_ in image:
                 image_ = image_.convert("RGB")
-                image_ = resize(image_, self.unet.sample_size)
+                image_ = resize(image_, self.unet.config.sample_size)
                 image_ = np.array(image_)
                 image_ = image_.astype(np.float32)
                 image_ = image_ / 127.5 - 1
@@ -701,7 +683,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
 
             for mask_image_ in mask_image:
                 mask_image_ = mask_image_.convert("L")
-                mask_image_ = resize(mask_image_, self.unet.sample_size)
+                mask_image_ = resize(mask_image_, self.unet.config.sample_size)
                 mask_image_ = np.array(mask_image_)
                 mask_image_ = mask_image_[None, None, :]
                 new_mask_image.append(mask_image_)
@@ -723,13 +705,15 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
 
         return mask_image
 
-    # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if_img2img.IFImg2ImgPipeline.get_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.StableDiffusionImg2ImgPipeline.get_timesteps
     def get_timesteps(self, num_inference_steps, strength):
         # get the original timestep using init_timestep
         init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
 
         t_start = max(num_inference_steps - init_timestep, 0)
-        timesteps = self.scheduler.timesteps[t_start:]
+        timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
+        if hasattr(self.scheduler, "set_begin_index"):
+            self.scheduler.set_begin_index(t_start * self.scheduler.order)
 
         return timesteps, num_inference_steps - t_start
 
@@ -776,11 +760,11 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         clean_caption: bool = True,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
@@ -792,7 +776,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
             prompt (`str` or `List[str]`, *optional*):
                 The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
                 instead.
-            image (`torch.FloatTensor` or `PIL.Image.Image`):
+            image (`torch.Tensor` or `PIL.Image.Image`):
                 `Image`, or tensor representing an image batch, that will be used as the starting point for the
                 process.
             mask_image (`PIL.Image.Image`):
@@ -830,10 +814,10 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
                 to make generation deterministic.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -844,7 +828,7 @@ class IFInpaintingPipeline(DiffusionPipeline, LoraLoaderMixin):
                 Whether or not to return a [`~pipelines.stable_diffusion.IFPipelineOutput`] instead of a plain tuple.
             callback (`Callable`, *optional*):
                 A function that will be called every `callback_steps` steps during inference. The function will be
-                called with the following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+                called with the following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             callback_steps (`int`, *optional*, defaults to 1):
                 The frequency at which the `callback` function will be called. If not specified, the callback will be
                 called at every step.

diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting_superresolution.py

@@ -10,13 +10,12 @@ import torch
 import torch.nn.functional as F
 from transformers import CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
-from ...loaders import LoraLoaderMixin
+from ...loaders import StableDiffusionLoraLoaderMixin
 from ...models import UNet2DConditionModel
 from ...schedulers import DDPMScheduler
 from ...utils import (
     BACKENDS_MAPPING,
     PIL_INTERPOLATION,
-    is_accelerate_available,
     is_bs4_available,
     is_ftfy_available,
     logging,
@@ -114,7 +113,7 @@ EXAMPLE_DOC_STRING = """
     """
 
 
-class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
+class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
     tokenizer: T5Tokenizer
     text_encoder: T5EncoderModel
 
@@ -145,6 +144,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
 
     model_cpu_offload_seq = "text_encoder->unet"
     _optional_components = ["tokenizer", "text_encoder", "safety_checker", "feature_extractor", "watermarker"]
+    _exclude_from_cpu_offload = ["watermarker"]
 
     def __init__(
         self,
@@ -193,21 +193,6 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.remove_all_hooks
-    def remove_all_hooks(self):
-        if is_accelerate_available():
-            from accelerate.hooks import remove_hook_from_module
-        else:
-            raise ImportError("Please install accelerate via `pip install accelerate`")
-
-        for model in [self.text_encoder, self.unet, self.safety_checker]:
-            if model is not None:
-                remove_hook_from_module(model, recurse=True)
-
-        self.unet_offload_hook = None
-        self.text_encoder_offload_hook = None
-        self.final_offload_hook = None
-
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline._text_preprocessing
     def _text_preprocessing(self, text, clean_caption=False):
         if clean_caption and not is_bs4_available():
@@ -357,8 +342,8 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         num_images_per_prompt: int = 1,
         device: Optional[torch.device] = None,
         negative_prompt: Optional[Union[str, List[str]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         clean_caption: bool = False,
     ):
         r"""
@@ -377,10 +362,10 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
                 The prompt or prompts not to guide the image generation. If not defined, one has to pass
                 `negative_prompt_embeds`. instead. If not defined, one has to pass `negative_prompt_embeds`. instead.
                 Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`).
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -515,9 +500,6 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             nsfw_detected = None
             watermark_detected = None
 
-            if hasattr(self, "unet_offload_hook") and self.unet_offload_hook is not None:
-                self.unet_offload_hook.offload()
-
         return image, nsfw_detected, watermark_detected
 
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.prepare_extra_step_kwargs
@@ -597,7 +579,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             and not isinstance(check_image_type, np.ndarray)
         ):
             raise ValueError(
-                "`image` has to be of type `torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
+                "`image` has to be of type `torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
                 f" {type(check_image_type)}"
             )
 
@@ -628,7 +610,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             and not isinstance(check_image_type, np.ndarray)
         ):
             raise ValueError(
-                "`original_image` has to be of type `torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
+                "`original_image` has to be of type `torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
                 f" {type(check_image_type)}"
             )
 
@@ -661,7 +643,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             and not isinstance(check_image_type, np.ndarray)
         ):
             raise ValueError(
-                "`mask_image` has to be of type `torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
+                "`mask_image` has to be of type `torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
                 f" {type(check_image_type)}"
             )
 
@@ -698,7 +680,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
 
             for image_ in image:
                 image_ = image_.convert("RGB")
-                image_ = resize(image_, self.unet.sample_size)
+                image_ = resize(image_, self.unet.config.sample_size)
                 image_ = np.array(image_)
                 image_ = image_.astype(np.float32)
                 image_ = image_ / 127.5 - 1
@@ -778,7 +760,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
 
             for mask_image_ in mask_image:
                 mask_image_ = mask_image_.convert("L")
-                mask_image_ = resize(mask_image_, self.unet.sample_size)
+                mask_image_ = resize(mask_image_, self.unet.config.sample_size)
                 mask_image_ = np.array(mask_image_)
                 mask_image_ = mask_image_[None, None, :]
                 new_mask_image.append(mask_image_)
@@ -800,13 +782,15 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
 
         return mask_image
 
-    # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if_img2img.IFImg2ImgPipeline.get_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.StableDiffusionImg2ImgPipeline.get_timesteps
     def get_timesteps(self, num_inference_steps, strength):
         # get the original timestep using init_timestep
         init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
 
         t_start = max(num_inference_steps - init_timestep, 0)
-        timesteps = self.scheduler.timesteps[t_start:]
+        timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
+        if hasattr(self.scheduler, "set_begin_index"):
+            self.scheduler.set_begin_index(t_start * self.scheduler.order)
 
         return timesteps, num_inference_steps - t_start
 
@@ -839,7 +823,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
         self,
-        image: Union[PIL.Image.Image, np.ndarray, torch.FloatTensor],
+        image: Union[PIL.Image.Image, np.ndarray, torch.Tensor],
         original_image: Union[
             PIL.Image.Image, torch.Tensor, np.ndarray, List[PIL.Image.Image], List[torch.Tensor], List[np.ndarray]
         ] = None,
@@ -855,11 +839,11 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         noise_level: int = 0,
@@ -869,10 +853,10 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         Function invoked when calling the pipeline for generation.
 
         Args:
-            image (`torch.FloatTensor` or `PIL.Image.Image`):
+            image (`torch.Tensor` or `PIL.Image.Image`):
                 `Image`, or tensor representing an image batch, that will be used as the starting point for the
                 process.
-            original_image (`torch.FloatTensor` or `PIL.Image.Image`):
+            original_image (`torch.Tensor` or `PIL.Image.Image`):
                 The original image that `image` was varied from.
             mask_image (`PIL.Image.Image`):
                 `Image`, or tensor representing an image batch, to mask `image`. White pixels in the mask will be
@@ -912,10 +896,10 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
                 to make generation deterministic.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -926,7 +910,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
                 Whether or not to return a [`~pipelines.stable_diffusion.IFPipelineOutput`] instead of a plain tuple.
             callback (`Callable`, *optional*):
                 A function that will be called every `callback_steps` steps during inference. The function will be
-                called with the following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+                called with the following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             callback_steps (`int`, *optional*, defaults to 1):
                 The frequency at which the `callback` function will be called. If not specified, the callback will be
                 called at every step.

diffusers/pipelines/deepfloyd_if/pipeline_if_superresolution.py

@@ -10,12 +10,11 @@ import torch
 import torch.nn.functional as F
 from transformers import CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
-from ...loaders import LoraLoaderMixin
+from ...loaders import StableDiffusionLoraLoaderMixin
 from ...models import UNet2DConditionModel
 from ...schedulers import DDPMScheduler
 from ...utils import (
     BACKENDS_MAPPING,
-    is_accelerate_available,
     is_bs4_available,
     is_ftfy_available,
     logging,
@@ -70,7 +69,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
+class IFSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
     tokenizer: T5Tokenizer
     text_encoder: T5EncoderModel
 
@@ -101,6 +100,7 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
 
     _optional_components = ["tokenizer", "text_encoder", "safety_checker", "feature_extractor", "watermarker"]
     model_cpu_offload_seq = "text_encoder->unet"
+    _exclude_from_cpu_offload = ["watermarker"]
 
     def __init__(
         self,
@@ -149,21 +149,6 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.remove_all_hooks
-    def remove_all_hooks(self):
-        if is_accelerate_available():
-            from accelerate.hooks import remove_hook_from_module
-        else:
-            raise ImportError("Please install accelerate via `pip install accelerate`")
-
-        for model in [self.text_encoder, self.unet, self.safety_checker]:
-            if model is not None:
-                remove_hook_from_module(model, recurse=True)
-
-        self.unet_offload_hook = None
-        self.text_encoder_offload_hook = None
-        self.final_offload_hook = None
-
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline._text_preprocessing
     def _text_preprocessing(self, text, clean_caption=False):
         if clean_caption and not is_bs4_available():
@@ -313,8 +298,8 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         num_images_per_prompt: int = 1,
         device: Optional[torch.device] = None,
         negative_prompt: Optional[Union[str, List[str]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         clean_caption: bool = False,
     ):
         r"""
@@ -333,10 +318,10 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
                 The prompt or prompts not to guide the image generation. If not defined, one has to pass
                 `negative_prompt_embeds`. instead. If not defined, one has to pass `negative_prompt_embeds`. instead.
                 Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`).
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -471,9 +456,6 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             nsfw_detected = None
             watermark_detected = None
 
-            if hasattr(self, "unet_offload_hook") and self.unet_offload_hook is not None:
-                self.unet_offload_hook.offload()
-
         return image, nsfw_detected, watermark_detected
 
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline.prepare_extra_step_kwargs
@@ -555,7 +537,7 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             and not isinstance(check_image_type, np.ndarray)
         ):
             raise ValueError(
-                "`image` has to be of type `torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
+                "`image` has to be of type `torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, or List[...] but is"
                 f" {type(check_image_type)}"
             )
 
@@ -626,7 +608,7 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         prompt: Union[str, List[str]] = None,
         height: int = None,
         width: int = None,
-        image: Union[PIL.Image.Image, np.ndarray, torch.FloatTensor] = None,
+        image: Union[PIL.Image.Image, np.ndarray, torch.Tensor] = None,
         num_inference_steps: int = 50,
         timesteps: List[int] = None,
         guidance_scale: float = 4.0,
@@ -634,11 +616,11 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         noise_level: int = 250,
@@ -655,7 +637,7 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
                 The height in pixels of the generated image.
             width (`int`, *optional*, defaults to None):
                 The width in pixels of the generated image.
-            image (`PIL.Image.Image`, `np.ndarray`, `torch.FloatTensor`):
+            image (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`):
                 The image to be upscaled.
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
@@ -681,10 +663,10 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
                 to make generation deterministic.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -695,7 +677,7 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
                 Whether or not to return a [`~pipelines.stable_diffusion.IFPipelineOutput`] instead of a plain tuple.
             callback (`Callable`, *optional*):
                 A function that will be called every `callback_steps` steps during inference. The function will be
-                called with the following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+                called with the following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             callback_steps (`int`, *optional*, defaults to 1):
                 The frequency at which the `callback` function will be called. If not specified, the callback will be
                 called at every step.
@@ -775,6 +757,9 @@ class IFSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
             self.scheduler.set_timesteps(num_inference_steps, device=device)
             timesteps = self.scheduler.timesteps
 
+        if hasattr(self.scheduler, "set_begin_index"):
+            self.scheduler.set_begin_index(0)
+
         # 5. Prepare intermediate images
         num_channels = self.unet.config.in_channels // 2
         intermediate_images = self.prepare_intermediate_images(

diffusers/pipelines/deepfloyd_if/pipeline_output.py

@@ -9,16 +9,17 @@ from ...utils import BaseOutput
 
 @dataclass
 class IFPipelineOutput(BaseOutput):
-    """
-    Args:
+    r"""
     Output class for Stable Diffusion pipelines.
-        images (`List[PIL.Image.Image]` or `np.ndarray`)
+
+    Args:
+        images (`List[PIL.Image.Image]` or `np.ndarray`):
             List of denoised PIL images of length `batch_size` or numpy array of shape `(batch_size, height, width,
             num_channels)`. PIL images or numpy array present the denoised images of the diffusion pipeline.
-        nsfw_detected (`List[bool]`)
+        nsfw_detected (`List[bool]`):
             List of flags denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content or a watermark. `None` if safety checking could not be performed.
-        watermark_detected (`List[bool]`)
+        watermark_detected (`List[bool]`):
             List of flags denoting whether the corresponding generated image likely has a watermark. `None` if safety
             checking could not be performed.
     """

diffusers/pipelines/deepfloyd_if/watermark.py

@@ -17,7 +17,7 @@ class IFWatermarker(ModelMixin, ConfigMixin):
         self.watermark_image_as_pil = None
 
     def apply_watermark(self, images: List[PIL.Image.Image], sample_size=None):
-        # copied from https://github.com/deep-floyd/IF/blob/b77482e36ca2031cb94dbca1001fc1e6400bf4ab/deepfloyd_if/modules/base.py#L287
+        # Copied from https://github.com/deep-floyd/IF/blob/b77482e36ca2031cb94dbca1001fc1e6400bf4ab/deepfloyd_if/modules/base.py#L287
 
         h = images[0].height
         w = images[0].width

diffusers/pipelines/deprecated/alt_diffusion/modeling_roberta_series.py

@@ -13,27 +13,27 @@ class TransformationModelOutput(ModelOutput):
     Base class for text model's outputs that also contains a pooling of the last hidden states.
 
     Args:
-        text_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`):
+        text_embeds (`torch.Tensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`):
             The text embeddings obtained by applying the projection layer to the pooler_output.
-        last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+        last_hidden_state (`torch.Tensor` of shape `(batch_size, sequence_length, hidden_size)`):
             Sequence of hidden-states at the output of the last layer of the model.
-        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
-            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
-            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+        hidden_states (`tuple(torch.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.Tensor` (one for the output of the embeddings, if the model has an embedding layer, + one
+            for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
 
             Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
-        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
-            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+        attentions (`tuple(torch.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.Tensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
             sequence_length)`.
 
             Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
             heads.
     """
 
-    projection_state: Optional[torch.FloatTensor] = None
-    last_hidden_state: torch.FloatTensor = None
-    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
-    attentions: Optional[Tuple[torch.FloatTensor]] = None
+    projection_state: Optional[torch.Tensor] = None
+    last_hidden_state: torch.Tensor = None
+    hidden_states: Optional[Tuple[torch.Tensor]] = None
+    attentions: Optional[Tuple[torch.Tensor]] = None
 
 
 class RobertaSeriesConfig(XLMRobertaConfig):