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

diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion.py

@@ -81,7 +81,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
     @torch.no_grad()
     def image_variation(
         self,
-        image: Union[torch.FloatTensor, PIL.Image.Image],
+        image: Union[torch.Tensor, PIL.Image.Image],
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 50,
@@ -90,10 +90,10 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
     ):
         r"""
@@ -123,7 +123,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
             generator (`torch.Generator`, *optional*):
                 A [`torch.Generator`](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 is generated by sampling using the supplied random `generator`.
@@ -134,7 +134,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
                 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)`.
+                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.
@@ -202,10 +202,10 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
     ):
         r"""
@@ -235,7 +235,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
             generator (`torch.Generator`, *optional*):
                 A [`torch.Generator`](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 is generated by sampling using the supplied random `generator`.
@@ -246,7 +246,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
                 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)`.
+                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.
@@ -311,10 +311,10 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
     ):
         r"""
@@ -344,7 +344,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 A [`torch.Generator`](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 is generated by sampling using the supplied random `generator`.
@@ -355,7 +355,7 @@ class VersatileDiffusionPipeline(DiffusionPipeline):
                 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)`.
+                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/deprecated/versatile_diffusion/pipeline_versatile_diffusion_dual_guided.py

@@ -348,7 +348,12 @@ class VersatileDiffusionDualGuidedPipeline(DiffusionPipeline):
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
-        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
+        shape = (
+            batch_size,
+            num_channels_latents,
+            int(height) // self.vae_scale_factor,
+            int(width) // self.vae_scale_factor,
+        )
         if isinstance(generator, list) and len(generator) != batch_size:
             raise ValueError(
                 f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
@@ -390,10 +395,10 @@ class VersatileDiffusionDualGuidedPipeline(DiffusionPipeline):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         **kwargs,
     ):
@@ -424,7 +429,7 @@ class VersatileDiffusionDualGuidedPipeline(DiffusionPipeline):
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 A [`torch.Generator`](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 is generated by sampling using the supplied random `generator`.
@@ -434,7 +439,7 @@ class VersatileDiffusionDualGuidedPipeline(DiffusionPipeline):
                 Whether or not to return a [`~pipelines.ImagePipelineOutput`] 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)`.
+                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/deprecated/versatile_diffusion/pipeline_versatile_diffusion_image_variation.py

@@ -197,7 +197,7 @@ class VersatileDiffusionImageVariationPipeline(DiffusionPipeline):
             and not isinstance(image, list)
         ):
             raise ValueError(
-                "`image` has to be of type `torch.FloatTensor` or `PIL.Image.Image` or `List[PIL.Image.Image]` but is"
+                "`image` has to be of type `torch.Tensor` or `PIL.Image.Image` or `List[PIL.Image.Image]` but is"
                 f" {type(image)}"
             )
 
@@ -214,7 +214,12 @@ class VersatileDiffusionImageVariationPipeline(DiffusionPipeline):
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
-        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
+        shape = (
+            batch_size,
+            num_channels_latents,
+            int(height) // self.vae_scale_factor,
+            int(width) // self.vae_scale_factor,
+        )
         if isinstance(generator, list) and len(generator) != batch_size:
             raise ValueError(
                 f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
@@ -242,10 +247,10 @@ class VersatileDiffusionImageVariationPipeline(DiffusionPipeline):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         **kwargs,
     ):
@@ -276,7 +281,7 @@ class VersatileDiffusionImageVariationPipeline(DiffusionPipeline):
             generator (`torch.Generator`, *optional*):
                 A [`torch.Generator`](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 is generated by sampling using the supplied random `generator`.
@@ -287,7 +292,7 @@ class VersatileDiffusionImageVariationPipeline(DiffusionPipeline):
                 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)`.
+                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/deprecated/versatile_diffusion/pipeline_versatile_diffusion_text_to_image.py

@@ -300,7 +300,12 @@ class VersatileDiffusionTextToImagePipeline(DiffusionPipeline):
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
-        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
+        shape = (
+            batch_size,
+            num_channels_latents,
+            int(height) // self.vae_scale_factor,
+            int(width) // self.vae_scale_factor,
+        )
         if isinstance(generator, list) and len(generator) != batch_size:
             raise ValueError(
                 f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
@@ -328,10 +333,10 @@ class VersatileDiffusionTextToImagePipeline(DiffusionPipeline):
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         **kwargs,
     ):
@@ -362,7 +367,7 @@ class VersatileDiffusionTextToImagePipeline(DiffusionPipeline):
             generator (`torch.Generator`, *optional*):
                 A [`torch.Generator`](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 is generated by sampling using the supplied random `generator`.
@@ -373,7 +378,7 @@ class VersatileDiffusionTextToImagePipeline(DiffusionPipeline):
                 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)`.
+                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/deprecated/vq_diffusion/pipeline_vq_diffusion.py

@@ -169,10 +169,10 @@ class VQDiffusionPipeline(DiffusionPipeline):
         truncation_rate: float = 1.0,
         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: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
     ) -> Union[ImagePipelineOutput, Tuple]:
         """
@@ -196,7 +196,7 @@ class VQDiffusionPipeline(DiffusionPipeline):
             generator (`torch.Generator`, *optional*):
                 A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
                 generation deterministic.
-            latents (`torch.FloatTensor` of shape (batch), *optional*):
+            latents (`torch.Tensor` of shape (batch), *optional*):
                 Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Must be valid embedding indices.If not provided, a latents tensor will be generated of
                 completely masked latent pixels.
@@ -206,7 +206,7 @@ class VQDiffusionPipeline(DiffusionPipeline):
                 Whether or not to return a [`~pipelines.ImagePipelineOutput`] 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)`.
+                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.
@@ -301,7 +301,7 @@ class VQDiffusionPipeline(DiffusionPipeline):
 
         return ImagePipelineOutput(images=image)
 
-    def truncate(self, log_p_x_0: torch.FloatTensor, truncation_rate: float) -> torch.FloatTensor:
+    def truncate(self, log_p_x_0: torch.Tensor, truncation_rate: float) -> torch.Tensor:
         """
         Truncates `log_p_x_0` such that for each column vector, the total cumulative probability is `truncation_rate`
         The lowest probabilities that would increase the cumulative probability above `truncation_rate` are set to

diffusers/pipelines/dit/pipeline_dit.py

@@ -22,7 +22,7 @@ from typing import Dict, List, Optional, Tuple, Union
 
 import torch
 
-from ...models import AutoencoderKL, Transformer2DModel
+from ...models import AutoencoderKL, DiTTransformer2DModel
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
@@ -36,8 +36,8 @@ class DiTPipeline(DiffusionPipeline):
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
 
     Parameters:
-        transformer ([`Transformer2DModel`]):
-            A class conditioned `Transformer2DModel` to denoise the encoded image latents.
+        transformer ([`DiTTransformer2DModel`]):
+            A class conditioned `DiTTransformer2DModel` to denoise the encoded image latents.
         vae ([`AutoencoderKL`]):
             Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
         scheduler ([`DDIMScheduler`]):
@@ -48,7 +48,7 @@ class DiTPipeline(DiffusionPipeline):
 
     def __init__(
         self,
-        transformer: Transformer2DModel,
+        transformer: DiTTransformer2DModel,
         vae: AutoencoderKL,
         scheduler: KarrasDiffusionSchedulers,
         id2label: Optional[Dict[int, str]] = None,
@@ -227,6 +227,9 @@ class DiTPipeline(DiffusionPipeline):
         if output_type == "pil":
             samples = self.numpy_to_pil(samples)
 
+        # Offload all models
+        self.maybe_free_model_hooks()
+
         if not return_dict:
             return (samples,)
 

diffusers/pipelines/flux/__init__.py

@@ -0,0 +1,69 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_additional_imports = {}
+_import_structure = {"pipeline_output": ["FluxPipelineOutput", "FluxPriorReduxPipelineOutput"]}
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["modeling_flux"] = ["ReduxImageEncoder"]
+    _import_structure["pipeline_flux"] = ["FluxPipeline"]
+    _import_structure["pipeline_flux_control"] = ["FluxControlPipeline"]
+    _import_structure["pipeline_flux_control_img2img"] = ["FluxControlImg2ImgPipeline"]
+    _import_structure["pipeline_flux_control_inpaint"] = ["FluxControlInpaintPipeline"]
+    _import_structure["pipeline_flux_controlnet"] = ["FluxControlNetPipeline"]
+    _import_structure["pipeline_flux_controlnet_image_to_image"] = ["FluxControlNetImg2ImgPipeline"]
+    _import_structure["pipeline_flux_controlnet_inpainting"] = ["FluxControlNetInpaintPipeline"]
+    _import_structure["pipeline_flux_fill"] = ["FluxFillPipeline"]
+    _import_structure["pipeline_flux_img2img"] = ["FluxImg2ImgPipeline"]
+    _import_structure["pipeline_flux_inpaint"] = ["FluxInpaintPipeline"]
+    _import_structure["pipeline_flux_prior_redux"] = ["FluxPriorReduxPipeline"]
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *  # noqa F403
+    else:
+        from .modeling_flux import ReduxImageEncoder
+        from .pipeline_flux import FluxPipeline
+        from .pipeline_flux_control import FluxControlPipeline
+        from .pipeline_flux_control_img2img import FluxControlImg2ImgPipeline
+        from .pipeline_flux_control_inpaint import FluxControlInpaintPipeline
+        from .pipeline_flux_controlnet import FluxControlNetPipeline
+        from .pipeline_flux_controlnet_image_to_image import FluxControlNetImg2ImgPipeline
+        from .pipeline_flux_controlnet_inpainting import FluxControlNetInpaintPipeline
+        from .pipeline_flux_fill import FluxFillPipeline
+        from .pipeline_flux_img2img import FluxImg2ImgPipeline
+        from .pipeline_flux_inpaint import FluxInpaintPipeline
+        from .pipeline_flux_prior_redux import FluxPriorReduxPipeline
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)
+    for name, value in _additional_imports.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/flux/modeling_flux.py

@@ -0,0 +1,47 @@
+# Copyright 2024 Black Forest Labs and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+from dataclasses import dataclass
+from typing import Optional
+
+import torch
+import torch.nn as nn
+
+from ...configuration_utils import ConfigMixin, register_to_config
+from ...models.modeling_utils import ModelMixin
+from ...utils import BaseOutput
+
+
+@dataclass
+class ReduxImageEncoderOutput(BaseOutput):
+    image_embeds: Optional[torch.Tensor] = None
+
+
+class ReduxImageEncoder(ModelMixin, ConfigMixin):
+    @register_to_config
+    def __init__(
+        self,
+        redux_dim: int = 1152,
+        txt_in_features: int = 4096,
+    ) -> None:
+        super().__init__()
+
+        self.redux_up = nn.Linear(redux_dim, txt_in_features * 3)
+        self.redux_down = nn.Linear(txt_in_features * 3, txt_in_features)
+
+    def forward(self, x: torch.Tensor) -> ReduxImageEncoderOutput:
+        projected_x = self.redux_down(nn.functional.silu(self.redux_up(x)))
+
+        return ReduxImageEncoderOutput(image_embeds=projected_x)