Docs: Stable Diffusion pipeline (#386)

* Initial description of Stable Diffusion pipeline.

* Placeholder docstrings to test preview.

* Add docstrings to Stable Diffusion pipeline.

* Style

* Docs for all the SD pipelines + attention slicing.

* Style: wrap long lines.

docs/source/api/pipelines/stable_diffusion.mdx

@@ -1 +1,40 @@
-# GLIDE MODEL
+# Stable diffusion pipelines
+
+## Overview
+
+Stable Diffusion is a text-to-image _latent diffusion_ model created by the researchers and engineers from [CompVis](https://github.com/CompVis), [Stability AI](https://stability.ai/) and [LAION](https://laion.ai/). It's trained on 512x512 images from a subset of the [LAION-5B](https://laion.ai/blog/laion-5b/) dataset. This model uses a frozen CLIP ViT-L/14 text encoder to condition the model on text prompts. With its 860M UNet and 123M text encoder, the model is relatively lightweight and can run on consumer GPUs.
+
+Latent diffusion is the research on top of which Stable Diffusion was built. It was proposed in [High-Resolution Image Synthesis with Latent Diffusion Models](https://arxiv.org/abs/2112.10752) by Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, Björn Ommer. You can learn more details about it in the [specific pipeline for latent diffusion](pipelines/latent_diffusion) that is part of 🤗 Diffusers.
+
+For more details about how Stable Diffusion works and how it differs from the base latent diffusion model, please refer to the official [launch announcement post](https://stability.ai/blog/stable-diffusion-announcement) and [this section of our own blog post](https://huggingface.co/blog/stable_diffusion#how-does-stable-diffusion-work).
+
+## Tips
+
+- To tweak your prompts on a specific result you liked, you can generate your own latents, as demonstrated in the following notebook: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pcuenca/diffusers-examples/blob/main/notebooks/stable-diffusion-seeds.ipynb)
+- TODO: some interesting Tips
+
+## Available pipelines
+
+| Pipeline | Tasks | Colab | Demo
+ |---|---|:---:|:---:|
+ | [pipeline_stable_diffusion.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py) | *Text-to-Image Generation* | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/stable_diffusion.ipynb) | [🤗 Stable Diffusion](https://huggingface.co/spaces/stabilityai/stable-diffusion)
+ | [pipeline_stable_diffusion_img2img.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py) | *Image-to-Image Text-Guided Generation* | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/patil-suraj/Notebooks/blob/master/image_2_image_using_diffusers.ipynb) | [🤗 Diffuse the Rest](https://huggingface.co/spaces/huggingface/diffuse-the-rest)
+ | [pipeline_stable_diffusion_inpaint.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py) | **Experimental** – *Text-Guided Image Inpainting* | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/patil-suraj/Notebooks/blob/master/in_painting_with_stable_diffusion_using_diffusers.ipynb) | Coming soon
+
+## API
+
+[[autodoc]] StableDiffusionPipeline
+     - __call__
+     - enable_attention_slicing
+     - disable_attention_slicing
+
+[[autodoc]] StableDiffusionImg2ImgPipeline
+     - __call__
+     - enable_attention_slicing
+     - disable_attention_slicing
+
+[[autodoc]] StableDiffusionInpaintPipeline
+     - __call__
+     - enable_attention_slicing
+     - disable_attention_slicing
+

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -14,6 +14,33 @@ from .safety_checker import StableDiffusionSafetyChecker
 
 
 class StableDiffusionPipeline(DiffusionPipeline):
+    r"""
+    Pipeline for text-to-image generation using Stable Diffusion.
+
+    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
+    library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
+
+    Args:
+        vae ([`AutoencoderKL`]):
+            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
+        text_encoder ([`CLIPTextModel`]):
+            Frozen text-encoder. Stable Diffusion uses the text portion of
+            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
+            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
+        tokenizer (`CLIPTokenizer`):
+            Tokenizer of class
+            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
+        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
+        scheduler ([`SchedulerMixin`]):
+            A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
+            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
+        safety_checker ([`StableDiffusionSafetyChecker`]):
+            Classification module that estimates whether generated images could be considered offsensive or harmful.
+            Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
+        feature_extractor ([`CLIPFeatureExtractor`]):
+            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
+    """
+
     def __init__(
         self,
         vae: AutoencoderKL,
@@ -37,6 +64,18 @@ class StableDiffusionPipeline(DiffusionPipeline):
         )
 
     def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
+        r"""
+        Enable sliced attention computation.
+
+        When this option is enabled, the attention module will split the input batch in slices, to compute attention in
+        several steps. This is useful to save some memory in exchange for a small speed decrease.
+
+        Args:
+            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
+                When `"auto"`, halves the input batch to the attention heads, so attention will be computed in two
+                steps. If a number is provided, use as many slices as `attention_head_dim // slice_size`. In this case,
+                `attention_head_dim` must be a multiple of `slice_size`.
+        """
         if slice_size == "auto":
             # half the attention head size is usually a good trade-off between
             # speed and memory
@@ -44,6 +83,10 @@ class StableDiffusionPipeline(DiffusionPipeline):
         self.unet.set_attention_slice(slice_size)
 
     def disable_attention_slicing(self):
+        r"""
+        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
+        back to computing attention in one step.
+        """
         # set slice_size = `None` to disable `set_attention_slice`
         self.enable_attention_slice(None)
 
@@ -62,6 +105,49 @@ class StableDiffusionPipeline(DiffusionPipeline):
         return_dict: bool = True,
         **kwargs,
     ):
+        r"""
+        Function invoked when calling the pipeline for generation.
+
+        Args:
+            prompt (`str` or `List[str]`):
+                The prompt or prompts to guide the image generation.
+            height (`int`, *optional*, defaults to 512):
+                The height in pixels of the generated image.
+            width (`int`, *optional*, defaults to 512):
+                The width in pixels of the generated image.
+            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
+                expense of slower inference.
+            guidance_scale (`float`, *optional*, defaults to 7.5):
+                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
+                `guidance_scale` is defined as `w` of equation 2. of [Imagen
+                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
+                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
+                usually at the expense of lower image quality.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
+            generator (`torch.Generator`, *optional*):
+                A [torch generator](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make generation
+                deterministic.
+            latents (`torch.FloatTensor`, *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`.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generate image. Choose between
+                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `nd.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
+                plain tuple.
+
+        Returns:
+            `~pipelines.stable_diffusion.StableDiffusionPipelineOutput` if `return_dict` is True, otherwise a tuple.
+            When returning a tuple, the first element is a list with the generated images, and the second element is a
+            list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content, according to the `safety_checker`.
+        """
+
         if "torch_device" in kwargs:
             device = kwargs.pop("torch_device")
             warnings.warn(

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -25,6 +25,33 @@ def preprocess(image):
 
 
 class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
+    r"""
+    Pipeline for text-guided image to image generation using Stable Diffusion.
+
+    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
+    library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
+
+    Args:
+        vae ([`AutoencoderKL`]):
+            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
+        text_encoder ([`CLIPTextModel`]):
+            Frozen text-encoder. Stable Diffusion uses the text portion of
+            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
+            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
+        tokenizer (`CLIPTokenizer`):
+            Tokenizer of class
+            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
+        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
+        scheduler ([`SchedulerMixin`]):
+            A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
+            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
+        safety_checker ([`StableDiffusionSafetyChecker`]):
+            Classification module that estimates whether generated images could be considered offsensive or harmful.
+            Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
+        feature_extractor ([`CLIPFeatureExtractor`]):
+            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
+    """
+
     def __init__(
         self,
         vae: AutoencoderKL,
@@ -48,6 +75,18 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         )
 
     def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
+        r"""
+        Enable sliced attention computation.
+
+        When this option is enabled, the attention module will split the input batch in slices, to compute attention in
+        several steps. This is useful to save some memory in exchange for a small speed decrease.
+
+        Args:
+            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
+                When `"auto"`, halves the input batch to the attention heads, so attention will be computed in two
+                steps. If a number is provided, use as many slices as `attention_head_dim // slice_size`. In this case,
+                `attention_head_dim` must be a multiple of `slice_size`.
+        """
         if slice_size == "auto":
             # half the attention head size is usually a good trade-off between
             # speed and memory
@@ -55,6 +94,10 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         self.unet.set_attention_slice(slice_size)
 
     def disable_attention_slicing(self):
+        r"""
+        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
+        back to computing attention in one step.
+        """
         # set slice_size = `None` to disable `set_attention_slice`
         self.enable_attention_slice(None)
 
@@ -71,7 +114,49 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
     ):
+        r"""
+        Function invoked when calling the pipeline for generation.
 
+        Args:
+            prompt (`str` or `List[str]`):
+                The prompt or prompts to guide the image generation.
+            init_image (`torch.FloatTensor` or `PIL.Image.Image`):
+                `Image`, or tensor representing an image batch, that will be used as the starting point for the
+                process.
+            strength (`float`, *optional*, defaults to 0.8):
+                Conceptually, indicates how much to transform the reference `init_image`. Must be between 0 and 1.
+                `init_image` will be used as a starting point, adding more noise to it the larger the `strength`. The
+                number of 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 `init_image`.
+            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
+                expense of slower inference. This parameter will be modulated by `strength`.
+            guidance_scale (`float`, *optional*, defaults to 7.5):
+                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
+                `guidance_scale` is defined as `w` of equation 2. of [Imagen
+                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
+                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
+                usually at the expense of lower image quality.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
+            generator (`torch.Generator`, *optional*):
+                A [torch generator](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make generation
+                deterministic.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generate image. Choose between
+                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `nd.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
+                plain tuple.
+
+        Returns:
+            `~pipelines.stable_diffusion.StableDiffusionPipelineOutput` if `return_dict` is True, otherwise a tuple.
+            When returning a tuple, the first element is a list with the generated images, and the second element is a
+            list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content, according to the `safety_checker`.
+        """
         if isinstance(prompt, str):
             batch_size = 1
         elif isinstance(prompt, list):

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -39,6 +39,33 @@ def preprocess_mask(mask):
 
 
 class StableDiffusionInpaintPipeline(DiffusionPipeline):
+    r"""
+    Pipeline for text-guided image inpainting using Stable Diffusion. *This is an experimental feature*.
+
+    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
+    library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
+
+    Args:
+        vae ([`AutoencoderKL`]):
+            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
+        text_encoder ([`CLIPTextModel`]):
+            Frozen text-encoder. Stable Diffusion uses the text portion of
+            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
+            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
+        tokenizer (`CLIPTokenizer`):
+            Tokenizer of class
+            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
+        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
+        scheduler ([`SchedulerMixin`]):
+            A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
+            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
+        safety_checker ([`StableDiffusionSafetyChecker`]):
+            Classification module that estimates whether generated images could be considered offsensive or harmful.
+            Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
+        feature_extractor ([`CLIPFeatureExtractor`]):
+            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
+    """
+
     def __init__(
         self,
         vae: AutoencoderKL,
@@ -62,6 +89,18 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
         )
 
     def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
+        r"""
+        Enable sliced attention computation.
+
+        When this option is enabled, the attention module will split the input batch in slices, to compute attention in
+        several steps. This is useful to save some memory in exchange for a small speed decrease.
+
+        Args:
+            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
+                When `"auto"`, halves the input batch to the attention heads, so attention will be computed in two
+                steps. If a number is provided, use as many slices as `attention_head_dim // slice_size`. In this case,
+                `attention_head_dim` must be a multiple of `slice_size`.
+        """
         if slice_size == "auto":
             # half the attention head size is usually a good trade-off between
             # speed and memory
@@ -69,6 +108,10 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
         self.unet.set_attention_slice(slice_size)
 
     def disable_attention_slicing(self):
+        r"""
+        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
+        back to computing attention in one step.
+        """
         # set slice_size = `None` to disable `set_attention_slice`
         self.enable_attention_slice(None)
 
@@ -86,7 +129,52 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
     ):
+        r"""
+        Function invoked when calling the pipeline for generation.
 
+        Args:
+            prompt (`str` or `List[str]`):
+                The prompt or prompts to guide the image generation.
+            init_image (`torch.FloatTensor` or `PIL.Image.Image`):
+                `Image`, or tensor representing an image batch, that will be used as the starting point for the
+                process. This is the image whose masked region will be inpainted.
+            mask_image (`torch.FloatTensor` or `PIL.Image.Image`):
+                `Image`, or tensor representing an image batch, to mask `init_image`. White pixels in the mask will be
+                replaced by noise and therefore repainted, while black pixels will be preserved. The mask image will be
+                converted to a single channel (luminance) before use.
+            strength (`float`, *optional*, defaults to 0.8):
+                Conceptually, indicates how much to inpaint the masked area. Must be between 0 and 1. When `strength`
+                is 1, the denoising process will be run on the masked area for the full number of iterations specified
+                in `num_inference_steps`. `init_image` will be used as a reference for the masked area, adding more
+                noise to that region the larger the `strength`. If `strength` is 0, no inpainting will occur.
+            num_inference_steps (`int`, *optional*, defaults to 50):
+                The reference number of denoising steps. More denoising steps usually lead to a higher quality image at
+                the expense of slower inference. This parameter will be modulated by `strength`, as explained above.
+            guidance_scale (`float`, *optional*, defaults to 7.5):
+                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
+                `guidance_scale` is defined as `w` of equation 2. of [Imagen
+                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
+                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
+                usually at the expense of lower image quality.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
+            generator (`torch.Generator`, *optional*):
+                A [torch generator](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make generation
+                deterministic.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generate image. Choose between
+                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `nd.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
+                plain tuple.
+
+        Returns:
+            `~pipelines.stable_diffusion.StableDiffusionPipelineOutput` if `return_dict` is True, otherwise a tuple.
+            When returning a tuple, the first element is a list with the generated images, and the second element is a
+            list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content, according to the `safety_checker`.
+        """
         if isinstance(prompt, str):
             batch_size = 1
         elif isinstance(prompt, list):