diffusers 0.27.0__py3-none-any.whl → 0.32.2__py3-none-any.whl

diffusers/pipelines/kandinsky/pipeline_kandinsky_prior.py

@@ -115,14 +115,14 @@ class KandinskyPriorPipelineOutput(BaseOutput):
     Output class for KandinskyPriorPipeline.
 
     Args:
-        image_embeds (`torch.FloatTensor`)
+        image_embeds (`torch.Tensor`)
             clip image embeddings for text prompt
         negative_image_embeds (`List[PIL.Image.Image]` or `np.ndarray`)
             clip image embeddings for unconditional tokens
     """
 
-    image_embeds: Union[torch.FloatTensor, np.ndarray]
-    negative_image_embeds: Union[torch.FloatTensor, np.ndarray]
+    image_embeds: Union[torch.Tensor, np.ndarray]
+    negative_image_embeds: Union[torch.Tensor, np.ndarray]
 
 
 class KandinskyPriorPipeline(DiffusionPipeline):
@@ -134,7 +134,7 @@ class KandinskyPriorPipeline(DiffusionPipeline):
 
     Args:
         prior ([`PriorTransformer`]):
-            The canonincal unCLIP prior to approximate the image embedding from the text embedding.
+            The canonical unCLIP prior to approximate the image embedding from the text embedding.
         image_encoder ([`CLIPVisionModelWithProjection`]):
             Frozen image-encoder.
         text_encoder ([`CLIPTextModelWithProjection`]):
@@ -173,12 +173,12 @@ class KandinskyPriorPipeline(DiffusionPipeline):
     @replace_example_docstring(EXAMPLE_INTERPOLATE_DOC_STRING)
     def interpolate(
         self,
-        images_and_prompts: List[Union[str, PIL.Image.Image, torch.FloatTensor]],
+        images_and_prompts: List[Union[str, PIL.Image.Image, torch.Tensor]],
         weights: List[float],
         num_images_per_prompt: int = 1,
         num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         negative_prior_prompt: Optional[str] = None,
         negative_prompt: str = "",
         guidance_scale: float = 4.0,
@@ -188,7 +188,7 @@ class KandinskyPriorPipeline(DiffusionPipeline):
         Function invoked when using the prior pipeline for interpolation.
 
         Args:
-            images_and_prompts (`List[Union[str, PIL.Image.Image, torch.FloatTensor]]`):
+            images_and_prompts (`List[Union[str, PIL.Image.Image, torch.Tensor]]`):
                 list of prompts and images to guide the image generation.
             weights: (`List[float]`):
                 list of weights for each condition in `images_and_prompts`
@@ -200,7 +200,7 @@ class KandinskyPriorPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.
@@ -403,7 +403,7 @@ class KandinskyPriorPipeline(DiffusionPipeline):
         num_images_per_prompt: int = 1,
         num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         guidance_scale: float = 4.0,
         output_type: Optional[str] = "pt",
         return_dict: bool = True,
@@ -425,7 +425,7 @@ class KandinskyPriorPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2.py

@@ -123,15 +123,15 @@ class KandinskyV22Pipeline(DiffusionPipeline):
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
         self,
-        image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        negative_image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
+        image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        negative_image_embeds: Union[torch.Tensor, List[torch.Tensor]],
         height: int = 512,
         width: int = 512,
         num_inference_steps: int = 100,
         guidance_scale: float = 4.0,
         num_images_per_prompt: int = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
@@ -142,9 +142,9 @@ class KandinskyV22Pipeline(DiffusionPipeline):
         Function invoked when calling the pipeline for generation.
 
         Args:
-            image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for text prompt, that will be used to condition the image generation.
-            negative_image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            negative_image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for negative text prompt, will be used to condition the image generation.
             height (`int`, *optional*, defaults to 512):
                 The height in pixels of the generated image.
@@ -164,7 +164,7 @@ class KandinskyV22Pipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_combined.py

@@ -119,7 +119,7 @@ class KandinskyV22CombinedPipeline(DiffusionPipeline):
         movq ([`VQModel`]):
             MoVQ Decoder to generate the image from the latents.
         prior_prior ([`PriorTransformer`]):
-            The canonincal unCLIP prior to approximate the image embedding from the text embedding.
+            The canonical unCLIP prior to approximate the image embedding from the text embedding.
         prior_image_encoder ([`CLIPVisionModelWithProjection`]):
             Frozen image-encoder.
         prior_text_encoder ([`CLIPTextModelWithProjection`]):
@@ -135,6 +135,7 @@ class KandinskyV22CombinedPipeline(DiffusionPipeline):
 
     model_cpu_offload_seq = "prior_text_encoder->prior_image_encoder->unet->movq"
     _load_connected_pipes = True
+    _exclude_from_cpu_offload = ["prior_prior"]
 
     def __init__(
         self,
@@ -178,7 +179,7 @@ class KandinskyV22CombinedPipeline(DiffusionPipeline):
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
@@ -186,8 +187,8 @@ class KandinskyV22CombinedPipeline(DiffusionPipeline):
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -212,9 +213,9 @@ class KandinskyV22CombinedPipeline(DiffusionPipeline):
         prior_guidance_scale: float = 4.0,
         prior_num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         return_dict: bool = True,
         prior_callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
@@ -258,7 +259,7 @@ class KandinskyV22CombinedPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.
@@ -346,7 +347,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
         movq ([`VQModel`]):
             MoVQ Decoder to generate the image from the latents.
         prior_prior ([`PriorTransformer`]):
-            The canonincal unCLIP prior to approximate the image embedding from the text embedding.
+            The canonical unCLIP prior to approximate the image embedding from the text embedding.
         prior_image_encoder ([`CLIPVisionModelWithProjection`]):
             Frozen image-encoder.
         prior_text_encoder ([`CLIPTextModelWithProjection`]):
@@ -362,6 +363,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
 
     model_cpu_offload_seq = "prior_text_encoder->prior_image_encoder->unet->movq"
     _load_connected_pipes = True
+    _exclude_from_cpu_offload = ["prior_prior"]
 
     def __init__(
         self,
@@ -405,17 +407,17 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_model_cpu_offload(self, gpu_id=0):
+    def enable_model_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
         to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
         method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
         `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
         """
-        self.prior_pipe.enable_model_cpu_offload()
-        self.decoder_pipe.enable_model_cpu_offload()
+        self.prior_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
@@ -423,8 +425,8 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -440,7 +442,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
     def __call__(
         self,
         prompt: Union[str, List[str]],
-        image: Union[torch.FloatTensor, PIL.Image.Image, List[torch.FloatTensor], List[PIL.Image.Image]],
+        image: Union[torch.Tensor, PIL.Image.Image, List[torch.Tensor], List[PIL.Image.Image]],
         negative_prompt: Optional[Union[str, List[str]]] = None,
         num_inference_steps: int = 100,
         guidance_scale: float = 4.0,
@@ -451,9 +453,9 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
         prior_guidance_scale: float = 4.0,
         prior_num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         return_dict: bool = True,
         prior_callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
@@ -467,7 +469,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
         Args:
             prompt (`str` or `List[str]`):
                 The prompt or prompts to guide the image generation.
-            image (`torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.FloatTensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
+            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
                 `Image`, or tensor representing an image batch, that will be used as the starting point for the
                 process. Can also accept image latents as `image`, if passing latents directly, it will not be encoded
                 again.
@@ -507,7 +509,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.
@@ -516,7 +518,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
                 (`np.array`) or `"pt"` (`torch.Tensor`).
             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)`.
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             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.
@@ -545,7 +547,7 @@ class KandinskyV22Img2ImgCombinedPipeline(DiffusionPipeline):
         negative_image_embeds = prior_outputs[1]
 
         prompt = [prompt] if not isinstance(prompt, (list, tuple)) else prompt
-        image = [image] if isinstance(prompt, PIL.Image.Image) else image
+        image = [image] if isinstance(image, PIL.Image.Image) else image
 
         if len(prompt) < image_embeds.shape[0] and image_embeds.shape[0] % len(prompt) == 0:
             prompt = (image_embeds.shape[0] // len(prompt)) * prompt
@@ -594,7 +596,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
         movq ([`VQModel`]):
             MoVQ Decoder to generate the image from the latents.
         prior_prior ([`PriorTransformer`]):
-            The canonincal unCLIP prior to approximate the image embedding from the text embedding.
+            The canonical unCLIP prior to approximate the image embedding from the text embedding.
         prior_image_encoder ([`CLIPVisionModelWithProjection`]):
             Frozen image-encoder.
         prior_text_encoder ([`CLIPTextModelWithProjection`]):
@@ -610,6 +612,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
 
     model_cpu_offload_seq = "prior_text_encoder->prior_image_encoder->unet->movq"
     _load_connected_pipes = True
+    _exclude_from_cpu_offload = ["prior_prior"]
 
     def __init__(
         self,
@@ -653,7 +656,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
@@ -661,8 +664,8 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -678,8 +681,8 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
     def __call__(
         self,
         prompt: Union[str, List[str]],
-        image: Union[torch.FloatTensor, PIL.Image.Image, List[torch.FloatTensor], List[PIL.Image.Image]],
-        mask_image: Union[torch.FloatTensor, PIL.Image.Image, List[torch.FloatTensor], List[PIL.Image.Image]],
+        image: Union[torch.Tensor, PIL.Image.Image, List[torch.Tensor], List[PIL.Image.Image]],
+        mask_image: Union[torch.Tensor, PIL.Image.Image, List[torch.Tensor], List[PIL.Image.Image]],
         negative_prompt: Optional[Union[str, List[str]]] = None,
         num_inference_steps: int = 100,
         guidance_scale: float = 4.0,
@@ -689,7 +692,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
         prior_guidance_scale: float = 4.0,
         prior_num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         prior_callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
@@ -704,7 +707,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
         Args:
             prompt (`str` or `List[str]`):
                 The prompt or prompts to guide the image generation.
-            image (`torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.FloatTensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
+            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
                 `Image`, or tensor representing an image batch, that will be used as the starting point for the
                 process. Can also accept image latents as `image`, if passing latents directly, it will not be encoded
                 again.
@@ -743,7 +746,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.
@@ -810,7 +813,7 @@ class KandinskyV22InpaintCombinedPipeline(DiffusionPipeline):
         negative_image_embeds = prior_outputs[1]
 
         prompt = [prompt] if not isinstance(prompt, (list, tuple)) else prompt
-        image = [image] if isinstance(prompt, PIL.Image.Image) else image
+        image = [image] if isinstance(image, PIL.Image.Image) else image
         mask_image = [mask_image] if isinstance(mask_image, PIL.Image.Image) else mask_image
 
         if len(prompt) < image_embeds.shape[0] and image_embeds.shape[0] % len(prompt) == 0:

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_controlnet.py

@@ -151,18 +151,18 @@ class KandinskyV22ControlnetPipeline(DiffusionPipeline):
     @torch.no_grad()
     def __call__(
         self,
-        image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        negative_image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        hint: torch.FloatTensor,
+        image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        negative_image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        hint: torch.Tensor,
         height: int = 512,
         width: int = 512,
         num_inference_steps: int = 100,
         guidance_scale: float = 4.0,
         num_images_per_prompt: int = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         return_dict: bool = True,
     ):
@@ -172,11 +172,11 @@ class KandinskyV22ControlnetPipeline(DiffusionPipeline):
         Args:
             prompt (`str` or `List[str]`):
                 The prompt or prompts to guide the image generation.
-            hint (`torch.FloatTensor`):
+            hint (`torch.Tensor`):
                 The controlnet condition.
-            image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for text prompt, that will be used to condition the image generation.
-            negative_image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            negative_image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for negative text prompt, will be used to condition the image generation.
             negative_prompt (`str` or `List[str]`, *optional*):
                 The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
@@ -199,7 +199,7 @@ class KandinskyV22ControlnetPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.
@@ -208,7 +208,7 @@ class KandinskyV22ControlnetPipeline(DiffusionPipeline):
                 (`np.array`) or `"pt"` (`torch.Tensor`).
             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)`.
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             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.

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_controlnet_img2img.py

@@ -206,10 +206,10 @@ class KandinskyV22ControlnetImg2ImgPipeline(DiffusionPipeline):
     @torch.no_grad()
     def __call__(
         self,
-        image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        image: Union[torch.FloatTensor, PIL.Image.Image, List[torch.FloatTensor], List[PIL.Image.Image]],
-        negative_image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        hint: torch.FloatTensor,
+        image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        image: Union[torch.Tensor, PIL.Image.Image, List[torch.Tensor], List[PIL.Image.Image]],
+        negative_image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        hint: torch.Tensor,
         height: int = 512,
         width: int = 512,
         num_inference_steps: int = 100,
@@ -218,7 +218,7 @@ class KandinskyV22ControlnetImg2ImgPipeline(DiffusionPipeline):
         num_images_per_prompt: int = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
         output_type: Optional[str] = "pil",
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         return_dict: bool = True,
     ):
@@ -226,9 +226,9 @@ class KandinskyV22ControlnetImg2ImgPipeline(DiffusionPipeline):
         Function invoked when calling the pipeline for generation.
 
         Args:
-            image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for text prompt, that will be used to condition the image generation.
-            image (`torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.FloatTensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
+            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
                 `Image`, or tensor representing an image batch, that will be used as the starting point for the
                 process. Can also accept image latents as `image`, if passing latents directly, it will not be encoded
                 again.
@@ -238,9 +238,9 @@ class KandinskyV22ControlnetImg2ImgPipeline(DiffusionPipeline):
                 denoising steps depends on the amount of noise initially added. When `strength` is 1, added noise will
                 be maximum and the denoising process will run for the full number of iterations specified in
                 `num_inference_steps`. A value of 1, therefore, essentially ignores `image`.
-            hint (`torch.FloatTensor`):
+            hint (`torch.Tensor`):
                 The controlnet condition.
-            negative_image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            negative_image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for negative text prompt, will be used to condition the image generation.
             height (`int`, *optional*, defaults to 512):
                 The height in pixels of the generated image.
@@ -265,7 +265,7 @@ class KandinskyV22ControlnetImg2ImgPipeline(DiffusionPipeline):
                 (`np.array`) or `"pt"` (`torch.Tensor`).
             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)`.
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             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.

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_img2img.py

@@ -190,9 +190,9 @@ class KandinskyV22Img2ImgPipeline(DiffusionPipeline):
     @torch.no_grad()
     def __call__(
         self,
-        image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        image: Union[torch.FloatTensor, PIL.Image.Image, List[torch.FloatTensor], List[PIL.Image.Image]],
-        negative_image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
+        image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        image: Union[torch.Tensor, PIL.Image.Image, List[torch.Tensor], List[PIL.Image.Image]],
+        negative_image_embeds: Union[torch.Tensor, List[torch.Tensor]],
         height: int = 512,
         width: int = 512,
         num_inference_steps: int = 100,
@@ -210,9 +210,9 @@ class KandinskyV22Img2ImgPipeline(DiffusionPipeline):
         Function invoked when calling the pipeline for generation.
 
         Args:
-            image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for text prompt, that will be used to condition the image generation.
-            image (`torch.FloatTensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.FloatTensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
+            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
                 `Image`, or tensor representing an image batch, that will be used as the starting point for the
                 process. Can also accept image latents as `image`, if passing latents directly, it will not be encoded
                 again.
@@ -222,7 +222,7 @@ class KandinskyV22Img2ImgPipeline(DiffusionPipeline):
                 denoising steps depends on the amount of noise initially added. When `strength` is 1, added noise will
                 be maximum and the denoising process will run for the full number of iterations specified in
                 `num_inference_steps`. A value of 1, therefore, essentially ignores `image`.
-            negative_image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            negative_image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for negative text prompt, will be used to condition the image generation.
             height (`int`, *optional*, defaults to 512):
                 The height in pixels of the generated image.

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_inpainting.py

@@ -294,17 +294,17 @@ class KandinskyV22InpaintPipeline(DiffusionPipeline):
     @torch.no_grad()
     def __call__(
         self,
-        image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
-        image: Union[torch.FloatTensor, PIL.Image.Image],
-        mask_image: Union[torch.FloatTensor, PIL.Image.Image, np.ndarray],
-        negative_image_embeds: Union[torch.FloatTensor, List[torch.FloatTensor]],
+        image_embeds: Union[torch.Tensor, List[torch.Tensor]],
+        image: Union[torch.Tensor, PIL.Image.Image],
+        mask_image: Union[torch.Tensor, PIL.Image.Image, np.ndarray],
+        negative_image_embeds: Union[torch.Tensor, List[torch.Tensor]],
         height: int = 512,
         width: int = 512,
         num_inference_steps: int = 100,
         guidance_scale: float = 4.0,
         num_images_per_prompt: int = 1,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
@@ -315,7 +315,7 @@ class KandinskyV22InpaintPipeline(DiffusionPipeline):
         Function invoked when calling the pipeline for generation.
 
         Args:
-            image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for text prompt, that will be used to condition the image generation.
             image (`PIL.Image.Image`):
                 `Image`, or tensor representing an image batch which will be inpainted, *i.e.* parts of the image will
@@ -325,7 +325,7 @@ class KandinskyV22InpaintPipeline(DiffusionPipeline):
                 black pixels will be preserved. If `mask_image` is a PIL image, it will be converted to a single
                 channel (luminance) before use. If it's a tensor, it should contain one color channel (L) instead of 3,
                 so the expected shape would be `(B, H, W, 1)`.
-            negative_image_embeds (`torch.FloatTensor` or `List[torch.FloatTensor]`):
+            negative_image_embeds (`torch.Tensor` or `List[torch.Tensor]`):
                 The clip image embeddings for negative text prompt, will be used to condition the image generation.
             height (`int`, *optional*, defaults to 512):
                 The height in pixels of the generated image.
@@ -345,7 +345,7 @@ class KandinskyV22InpaintPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.

diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_prior.py

@@ -90,7 +90,7 @@ class KandinskyV22PriorPipeline(DiffusionPipeline):
 
     Args:
         prior ([`PriorTransformer`]):
-            The canonincal unCLIP prior to approximate the image embedding from the text embedding.
+            The canonical unCLIP prior to approximate the image embedding from the text embedding.
         image_encoder ([`CLIPVisionModelWithProjection`]):
             Frozen image-encoder.
         text_encoder ([`CLIPTextModelWithProjection`]):
@@ -132,12 +132,12 @@ class KandinskyV22PriorPipeline(DiffusionPipeline):
     @replace_example_docstring(EXAMPLE_INTERPOLATE_DOC_STRING)
     def interpolate(
         self,
-        images_and_prompts: List[Union[str, PIL.Image.Image, torch.FloatTensor]],
+        images_and_prompts: List[Union[str, PIL.Image.Image, torch.Tensor]],
         weights: List[float],
         num_images_per_prompt: int = 1,
         num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         negative_prior_prompt: Optional[str] = None,
         negative_prompt: str = "",
         guidance_scale: float = 4.0,
@@ -147,7 +147,7 @@ class KandinskyV22PriorPipeline(DiffusionPipeline):
         Function invoked when using the prior pipeline for interpolation.
 
         Args:
-            images_and_prompts (`List[Union[str, PIL.Image.Image, torch.FloatTensor]]`):
+            images_and_prompts (`List[Union[str, PIL.Image.Image, torch.Tensor]]`):
                 list of prompts and images to guide the image generation.
             weights: (`List[float]`):
                 list of weights for each condition in `images_and_prompts`
@@ -159,7 +159,7 @@ class KandinskyV22PriorPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.
@@ -376,7 +376,7 @@ class KandinskyV22PriorPipeline(DiffusionPipeline):
         num_images_per_prompt: int = 1,
         num_inference_steps: int = 25,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
         guidance_scale: float = 4.0,
         output_type: Optional[str] = "pt",  # pt only
         return_dict: bool = True,
@@ -400,7 +400,7 @@ class KandinskyV22PriorPipeline(DiffusionPipeline):
             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.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will ge generated by sampling using the supplied random `generator`.