add model (#3230)

* add

* clean

* up

* clean up more

* fix more tests

* Improve docs further

* improve

* more fixes docs

* Improve docs more

* Update src/diffusers/models/unet_2d_condition.py

* fix

* up

* update doc links

* make fix-copies

* add safety checker and watermarker to stage 3 doc page code snippets

* speed optimizations docs

* memory optimization docs

* make style

* add watermarking snippets to doc string examples

* make style

* use pt_to_pil helper functions in doc strings

* skip mps tests

* Improve safety

* make style

* new logic

* fix

* fix bad onnx design

* make new stable diffusion upscale pipeline model arguments optional

* define has_nsfw_concept when non-pil output type

* lowercase linked to notebook name

---------
Co-authored-by: William Berman <WLBberman@gmail.com>

docs/source/en/_toctree.yml

@@ -152,6 +152,8 @@
       title: DDPM
     - local: api/pipelines/dit
       title: DiT
+    - local: api/pipelines/if
+      title: IF
     - local: api/pipelines/latent_diffusion
       title: Latent Diffusion
     - local: api/pipelines/paint_by_example

docs/source/en/api/pipelines/if.mdx

+<!--Copyright 2023 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.
+-->
+
+# IF
+
+## Overview
+
+DeepFloyd IF is a novel state-of-the-art open-source text-to-image model with a high degree of photorealism and language understanding. 
+The model is a modular composed of a frozen text encoder and three cascaded pixel diffusion modules: 
+- Stage 1: a base model that generates 64x64 px image based on text prompt,
+- Stage 2: a 64x64 px => 256x256 px super-resolution model, and a
+- Stage 3: a 256x256 px => 1024x1024 px super-resolution model
+Stage 1 and Stage 2 utilize a frozen text encoder based on the T5 transformer to extract text embeddings, 
+which are then fed into a UNet architecture enhanced with cross-attention and attention pooling. 
+Stage 3 is [Stability's x4 Upscaling model](https://huggingface.co/stabilityai/stable-diffusion-x4-upscaler).
+The result is a highly efficient model that outperforms current state-of-the-art models, achieving a zero-shot FID score of 6.66 on the COCO dataset. 
+Our work underscores the potential of larger UNet architectures in the first stage of cascaded diffusion models and depicts a promising future for text-to-image synthesis.
+
+## Usage
+
+Before you can use IF, you need to accept its usage conditions. To do so:
+1. Make sure to have a [Hugging Face account](https://huggingface.co/join) and be loggin in
+2. Accept the license on the model card of [DeepFloyd/IF-I-IF-v1.0](https://huggingface.co/DeepFloyd/IF-I-IF-v1.0) and [DeepFloyd/IF-II-L-v1.0](https://huggingface.co/DeepFloyd/IF-II-L-v1.0)
+3. Make sure to login locally. Install `huggingface_hub`
+```sh
+pip install huggingface_hub --upgrade
+```
+
+run the login function in a Python shell
+
+```py
+from huggingface_hub import login
+
+login()
+```
+
+and enter your [Hugging Face Hub access token](https://huggingface.co/docs/hub/security-tokens#what-are-user-access-tokens).
+
+Next we install `diffusers` and dependencies:
+
+```sh
+pip install diffusers accelerate transformers safetensors
+```
+
+The following sections give more in-detail examples of how to use IF. Specifically:
+
+- [Text-to-Image Generation](#text-to-image-generation)
+- [Image-to-Image Generation](#text-guided-image-to-image-generation)
+- [Inpainting](#text-guided-inpainting-generation)
+- [Reusing model weights](#converting-between-different-pipelines)
+- [Speed optimization](#optimizing-for-speed)
+- [Memory optimization](#optimizing-for-memory)
+
+**Available checkpoints**
+- *Stage-1*
+  - [DeepFloyd/IF-I-IF-v1.0](https://huggingface.co/DeepFloyd/IF-I-IF-v1.0)
+  - [DeepFloyd/IF-I-L-v1.0](https://huggingface.co/DeepFloyd/IF-I-L-v1.0)
+  - [DeepFloyd/IF-I-M-v1.0](https://huggingface.co/DeepFloyd/IF-I-M-v1.0)
+
+- *Stage-2*
+  - [DeepFloyd/IF-II-L-v1.0](https://huggingface.co/DeepFloyd/IF-II-L-v1.0)
+  - [DeepFloyd/IF-II-M-v1.0](https://huggingface.co/DeepFloyd/IF-II-M-v1.0)
+
+- *Stage-3*
+  - [stabilityai/stable-diffusion-x4-upscaler](https://huggingface.co/stabilityai/stable-diffusion-x4-upscaler)
+
+**Demo**
+[![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/DeepFloyd/IF)
+
+**Google Colab**
+[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/deepfloyd_if_free_tier_google_colab.ipynb)
+
+### Text-to-Image Generation
+
+By default diffusers makes use of [model cpu offloading](https://huggingface.co/docs/diffusers/optimization/fp16#model-offloading-for-fast-inference-and-memory-savings)
+to run the whole IF pipeline with as little as 14 GB of VRAM.
+
+```python
+from diffusers import DiffusionPipeline
+from diffusers.utils import pt_to_pil
+import torch
+
+# stage 1
+stage_1 = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+stage_1.enable_model_cpu_offload()
+
+# stage 2
+stage_2 = DiffusionPipeline.from_pretrained(
+    "DeepFloyd/IF-II-L-v1.0", text_encoder=None, variant="fp16", torch_dtype=torch.float16
+)
+stage_2.enable_model_cpu_offload()
+
+# stage 3
+safety_modules = {
+    "feature_extractor": stage_1.feature_extractor,
+    "safety_checker": stage_1.safety_checker,
+    "watermarker": stage_1.watermarker,
+}
+stage_3 = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-x4-upscaler", **safety_modules, torch_dtype=torch.float16
+)
+stage_3.enable_model_cpu_offload()
+
+prompt = 'a photo of a kangaroo wearing an orange hoodie and blue sunglasses standing in front of the eiffel tower holding a sign that says "very deep learning"'
+generator = torch.manual_seed(1)
+
+# text embeds
+prompt_embeds, negative_embeds = stage_1.encode_prompt(prompt)
+
+# stage 1
+image = stage_1(
+    prompt_embeds=prompt_embeds, negative_prompt_embeds=negative_embeds, generator=generator, output_type="pt"
+).images
+pt_to_pil(image)[0].save("./if_stage_I.png")
+
+# stage 2
+image = stage_2(
+    image=image,
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    generator=generator,
+    output_type="pt",
+).images
+pt_to_pil(image)[0].save("./if_stage_II.png")
+
+# stage 3
+image = stage_3(prompt=prompt, image=image, noise_level=100, generator=generator).images
+image[0].save("./if_stage_III.png")
+```
+
+### Text Guided Image-to-Image Generation
+
+The same IF model weights can be used for text-guided image-to-image translation or image variation.
+In this case just make sure to load the weights using the [`IFInpaintingPipeline`] and [`IFInpaintingSuperResolutionPipeline`] pipelines.
+
+**Note**: You can also directly move the weights of the text-to-image pipelines to the image-to-image pipelines
+without loading them twice by making use of the [`~DiffusionPipeline.components()`] function as explained [here](#converting-between-different-pipelines).
+
+```python
+from diffusers import IFImg2ImgPipeline, IFImg2ImgSuperResolutionPipeline, DiffusionPipeline
+from diffusers.utils import pt_to_pil
+
+import torch
+
+from PIL import Image
+import requests
+from io import BytesIO
+
+# download image
+url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/assets/stable-samples/img2img/sketch-mountains-input.jpg"
+response = requests.get(url)
+original_image = Image.open(BytesIO(response.content)).convert("RGB")
+original_image = original_image.resize((768, 512))
+
+# stage 1
+stage_1 = IFImg2ImgPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+stage_1.enable_model_cpu_offload()
+
+# stage 2
+stage_2 = IFImg2ImgSuperResolutionPipeline.from_pretrained(
+    "DeepFloyd/IF-II-L-v1.0", text_encoder=None, variant="fp16", torch_dtype=torch.float16
+)
+stage_2.enable_model_cpu_offload()
+
+# stage 3
+safety_modules = {
+    "feature_extractor": stage_1.feature_extractor,
+    "safety_checker": stage_1.safety_checker,
+    "watermarker": stage_1.watermarker,
+}
+stage_3 = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-x4-upscaler", **safety_modules, torch_dtype=torch.float16
+)
+stage_3.enable_model_cpu_offload()
+
+prompt = "A fantasy landscape in style minecraft"
+generator = torch.manual_seed(1)
+
+# text embeds
+prompt_embeds, negative_embeds = stage_1.encode_prompt(prompt)
+
+# stage 1
+image = stage_1(
+    image=original_image,
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    generator=generator,
+    output_type="pt",
+).images
+pt_to_pil(image)[0].save("./if_stage_I.png")
+
+# stage 2
+image = stage_2(
+    image=image,
+    original_image=original_image,
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    generator=generator,
+    output_type="pt",
+).images
+pt_to_pil(image)[0].save("./if_stage_II.png")
+
+# stage 3
+image = stage_3(prompt=prompt, image=image, generator=generator, noise_level=100).images
+image[0].save("./if_stage_III.png")
+```
+
+### Text Guided Inpainting Generation
+
+The same IF model weights can be used for text-guided image-to-image translation or image variation.
+In this case just make sure to load the weights using the [`IFInpaintingPipeline`] and [`IFInpaintingSuperResolutionPipeline`] pipelines.
+
+**Note**: You can also directly move the weights of the text-to-image pipelines to the image-to-image pipelines
+without loading them twice by making use of the [`~DiffusionPipeline.components()`] function as explained [here](#converting-between-different-pipelines).
+
+```python
+from diffusers import IFInpaintingPipeline, IFInpaintingSuperResolutionPipeline, DiffusionPipeline
+from diffusers.utils import pt_to_pil
+import torch
+
+from PIL import Image
+import requests
+from io import BytesIO
+
+# download image
+url = "https://huggingface.co/datasets/diffusers/docs-images/resolve/main/if/person.png"
+response = requests.get(url)
+original_image = Image.open(BytesIO(response.content)).convert("RGB")
+original_image = original_image
+
+# download mask
+url = "https://huggingface.co/datasets/diffusers/docs-images/resolve/main/if/glasses_mask.png"
+response = requests.get(url)
+mask_image = Image.open(BytesIO(response.content))
+mask_image = mask_image
+
+# stage 1
+stage_1 = IFInpaintingPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+stage_1.enable_model_cpu_offload()
+
+# stage 2
+stage_2 = IFInpaintingSuperResolutionPipeline.from_pretrained(
+    "DeepFloyd/IF-II-L-v1.0", text_encoder=None, variant="fp16", torch_dtype=torch.float16
+)
+stage_2.enable_model_cpu_offload()
+
+# stage 3
+safety_modules = {
+    "feature_extractor": stage_1.feature_extractor,
+    "safety_checker": stage_1.safety_checker,
+    "watermarker": stage_1.watermarker,
+}
+stage_3 = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-x4-upscaler", **safety_modules, torch_dtype=torch.float16
+)
+stage_3.enable_model_cpu_offload()
+
+prompt = "blue sunglasses"
+generator = torch.manual_seed(1)
+
+# text embeds
+prompt_embeds, negative_embeds = stage_1.encode_prompt(prompt)
+
+# stage 1
+image = stage_1(
+    image=original_image,
+    mask_image=mask_image,
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    generator=generator,
+    output_type="pt",
+).images
+pt_to_pil(image)[0].save("./if_stage_I.png")
+
+# stage 2
+image = stage_2(
+    image=image,
+    original_image=original_image,
+    mask_image=mask_image,
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    generator=generator,
+    output_type="pt",
+).images
+pt_to_pil(image)[0].save("./if_stage_II.png")
+
+# stage 3
+image = stage_3(prompt=prompt, image=image, generator=generator, noise_level=100).images
+image[0].save("./if_stage_III.png")
+```
+
+### Converting between different pipelines
+
+In addition to being loaded with `from_pretrained`, Pipelines can also be loaded directly from each other.
+
+```python
+from diffusers import IFPipeline, IFSuperResolutionPipeline
+
+pipe_1 = IFPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0")
+pipe_2 = IFSuperResolutionPipeline.from_pretrained("DeepFloyd/IF-II-L-v1.0")
+
+
+from diffusers import IFImg2ImgPipeline, IFImg2ImgSuperResolutionPipeline
+
+pipe_1 = IFImg2ImgPipeline(**pipe_1.components)
+pipe_2 = IFImg2ImgSuperResolutionPipeline(**pipe_2.components)
+
+
+from diffusers import IFInpaintingPipeline, IFInpaintingSuperResolutionPipeline
+
+pipe_1 = IFInpaintingPipeline(**pipe_1.components)
+pipe_2 = IFInpaintingSuperResolutionPipeline(**pipe_2.components)
+```
+
+### Optimizing for speed
+
+The simplest optimization to run IF faster is to move all model components to the GPU.
+
+```py
+pipe = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+pipe.to("cuda")
+```
+
+You can also run the diffusion process for a shorter number of timesteps.
+
+This can either be done with the `num_inference_steps` argument
+
+```py
+pipe("<prompt>", num_inference_steps=30)
+```
+
+Or with the `timesteps` argument
+
+```py
+from diffusers.pipelines.deepfloyd_if import fast27_timesteps
+
+pipe("<prompt>", timesteps=fast27_timesteps)
+```
+
+When doing image variation or inpainting, you can also decrease the number of timesteps
+with the strength argument. The strength argument is the amount of noise to add to 
+the input image which also determines how many steps to run in the denoising process.
+A smaller number will vary the image less but run faster.
+
+```py
+pipe = IFImg2ImgPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+pipe.to("cuda")
+
+image = pipe(image=image, prompt="<prompt>", strength=0.3).images
+```
+
+You can also use [`torch.compile`](../../optimization/torch2.0). Note that we have not exhaustively tested `torch.compile`
+with IF and it might not give expected results.
+
+```py
+import torch
+
+pipe = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+pipe.to("cuda")
+
+pipe.text_encoder = torch.compile(pipe.text_encoder)
+pipe.unet = torch.compile(pipe.unet)
+```
+
+### Optimizing for memory
+
+When optimizing for GPU memory, we can use the standard diffusers cpu offloading APIs.
+
+Either the model based CPU offloading,
+
+```py
+pipe = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+pipe.enable_model_cpu_offload()
+```
+
+or the more aggressive layer based CPU offloading.
+
+```py
+pipe = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-IF-v1.0", variant="fp16", torch_dtype=torch.float16)
+pipe.enable_sequential_cpu_offload()
+```
+
+Additionally, T5 can be loaded in 8bit precision
+
+```py
+from transformers import T5EncoderModel
+
+text_encoder = T5EncoderModel.from_pretrained(
+    "DeepFloyd/IF-I-IF-v1.0", subfolder="text_encoder", device_map="auto", load_in_8bit=True, variant="8bit"
+)
+
+from diffusers import DiffusionPipeline
+
+pipe = DiffusionPipeline.from_pretrained(
+    "DeepFloyd/IF-I-IF-v1.0",
+    text_encoder=text_encoder,  # pass the previously instantiated 8bit text encoder
+    unet=None,
+    device_map="auto",
+)
+
+prompt_embeds, negative_embeds = pipe.encode_prompt("<prompt>")
+```
+
+For CPU RAM constrained machines like google colab free tier where we can't load all 
+model components to the CPU at once, we can manually only load the pipeline with
+the text encoder or unet when the respective model components are needed.
+
+```py
+from diffusers import IFPipeline, IFSuperResolutionPipeline
+import torch
+import gc
+from transformers import T5EncoderModel
+from diffusers.utils import pt_to_pil
+
+text_encoder = T5EncoderModel.from_pretrained(
+    "DeepFloyd/IF-I-IF-v1.0", subfolder="text_encoder", device_map="auto", load_in_8bit=True, variant="8bit"
+)
+
+# text to image
+
+pipe = DiffusionPipeline.from_pretrained(
+    "DeepFloyd/IF-I-IF-v1.0",
+    text_encoder=text_encoder,  # pass the previously instantiated 8bit text encoder
+    unet=None,
+    device_map="auto",
+)
+
+prompt = 'a photo of a kangaroo wearing an orange hoodie and blue sunglasses standing in front of the eiffel tower holding a sign that says "very deep learning"'
+prompt_embeds, negative_embeds = pipe.encode_prompt(prompt)
+
+# Remove the pipeline so we can re-load the pipeline with the unet
+del text_encoder
+del pipe
+gc.collect()
+torch.cuda.empty_cache()
+
+pipe = IFPipeline.from_pretrained(
+    "DeepFloyd/IF-I-IF-v1.0", text_encoder=None, variant="fp16", torch_dtype=torch.float16, device_map="auto"
+)
+
+generator = torch.Generator().manual_seed(0)
+image = pipe(
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    output_type="pt",
+    generator=generator,
+).images
+
+pt_to_pil(image)[0].save("./if_stage_I.png")
+
+# Remove the pipeline so we can load the super-resolution pipeline
+del pipe
+gc.collect()
+torch.cuda.empty_cache()
+
+# First super resolution
+
+pipe = IFSuperResolutionPipeline.from_pretrained(
+    "DeepFloyd/IF-II-L-v1.0", text_encoder=None, variant="fp16", torch_dtype=torch.float16, device_map="auto"
+)
+
+generator = torch.Generator().manual_seed(0)
+image = pipe(
+    image=image,
+    prompt_embeds=prompt_embeds,
+    negative_prompt_embeds=negative_embeds,
+    output_type="pt",
+    generator=generator,
+).images
+
+pt_to_pil(image)[0].save("./if_stage_II.png")
+```
+
+
+## Available Pipelines:
+
+| Pipeline | Tasks | Colab
+|---|---|:---:|
+| [pipeline_if.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/deepfloyd_if/pipeline_if.py) | *Text-to-Image Generation* | - |
+| [pipeline_if_superresolution.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/deepfloyd_if/pipeline_if.py) | *Text-to-Image Generation* | - |
+| [pipeline_if_img2img.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/deepfloyd_if/pipeline_if_img2img.py) | *Image-to-Image Generation* | - |
+| [pipeline_if_img2img_superresolution.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/deepfloyd_if/pipeline_if_img2img_superresolution.py) | *Image-to-Image Generation* | - |
+| [pipeline_if_inpainting.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting.py) | *Image-to-Image Generation* | - |
+| [pipeline_if_inpainting_superresolution.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting_superresolution.py) | *Image-to-Image Generation* | - |
+
+## IFPipeline
+[[autodoc]] IFPipeline
+	- all
+	- __call__
+
+## IFSuperResolutionPipeline
+[[autodoc]] IFSuperResolutionPipeline
+	- all
+	- __call__
+
+## IFImg2ImgPipeline
+[[autodoc]] IFImg2ImgPipeline
+	- all
+	- __call__
+
+## IFImg2ImgSuperResolutionPipeline
+[[autodoc]] IFImg2ImgSuperResolutionPipeline
+	- all
+	- __call__
+
+## IFInpaintingPipeline
+[[autodoc]] IFInpaintingPipeline
+	- all
+	- __call__
+
+## IFInpaintingSuperResolutionPipeline
+[[autodoc]] IFInpaintingSuperResolutionPipeline
+	- all
+	- __call__

docs/source/en/api/pipelines/overview.mdx

@@ -51,6 +51,9 @@ available a colab notebook to directly try them out.
 | [dance_diffusion](./dance_diffusion) | [**Dance Diffusion**](https://github.com/williamberman/diffusers.git) | Unconditional Audio Generation |
 | [ddpm](./ddpm) | [**Denoising Diffusion Probabilistic Models**](https://arxiv.org/abs/2006.11239) | Unconditional Image Generation |
 | [ddim](./ddim) | [**Denoising Diffusion Implicit Models**](https://arxiv.org/abs/2010.02502) | Unconditional Image Generation |
+| [if](./if) | [**IF**](https://github.com/deep-floyd/IF) | 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/deepfloyd_if_free_tier_google_colab.ipynb)
+| [if_img2img](./if) | [**IF**](https://github.com/deep-floyd/IF) | Image-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/deepfloyd_if_free_tier_google_colab.ipynb)
+| [if_inpainting](./if) | [**IF**](https://github.com/deep-floyd/IF) | Image-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/deepfloyd_if_free_tier_google_colab.ipynb)
 | [latent_diffusion](./latent_diffusion) | [**High-Resolution Image Synthesis with Latent Diffusion Models**](https://arxiv.org/abs/2112.10752)| Text-to-Image Generation | 
 | [latent_diffusion](./latent_diffusion) | [**High-Resolution Image Synthesis with Latent Diffusion Models**](https://arxiv.org/abs/2112.10752)| Super Resolution Image-to-Image | 
 | [latent_diffusion_uncond](./latent_diffusion_uncond) | [**High-Resolution Image Synthesis with Latent Diffusion Models**](https://arxiv.org/abs/2112.10752) | Unconditional Image Generation | 

docs/source/en/index.mdx

@@ -58,6 +58,9 @@ The library has three main components:
 | [dance_diffusion](./api/pipelines/dance_diffusion) | [Dance Diffusion](https://github.com/williamberman/diffusers.git) | Unconditional Audio Generation |
 | [ddpm](./api/pipelines/ddpm) | [Denoising Diffusion Probabilistic Models](https://arxiv.org/abs/2006.11239) | Unconditional Image Generation |
 | [ddim](./api/pipelines/ddim) | [Denoising Diffusion Implicit Models](https://arxiv.org/abs/2010.02502) | Unconditional Image Generation |
+| [if](./if) | [**IF**](./api/pipelines/if) | Image Generation |
+| [if_img2img](./if) | [**IF**](./api/pipelines/if) | Image-to-Image Generation |
+| [if_inpainting](./if) | [**IF**](./api/pipelines/if) | Image-to-Image Generation |
 | [latent_diffusion](./api/pipelines/latent_diffusion) | [High-Resolution Image Synthesis with Latent Diffusion Models](https://arxiv.org/abs/2112.10752)| Text-to-Image Generation |
 | [latent_diffusion](./api/pipelines/latent_diffusion) | [High-Resolution Image Synthesis with Latent Diffusion Models](https://arxiv.org/abs/2112.10752)| Super Resolution Image-to-Image |
 | [latent_diffusion_uncond](./api/pipelines/latent_diffusion_uncond) | [High-Resolution Image Synthesis with Latent Diffusion Models](https://arxiv.org/abs/2112.10752) | Unconditional Image Generation |

src/diffusers/__init__.py

@@ -114,6 +114,12 @@ else:
         AltDiffusionPipeline,
         AudioLDMPipeline,
         CycleDiffusionPipeline,
+        IFImg2ImgPipeline,
+        IFImg2ImgSuperResolutionPipeline,
+        IFInpaintingPipeline,
+        IFInpaintingSuperResolutionPipeline,
+        IFPipeline,
+        IFSuperResolutionPipeline,
         LDMTextToImagePipeline,
         PaintByExamplePipeline,
         SemanticStableDiffusionPipeline,

src/diffusers/configuration_utils.py

@@ -109,6 +109,7 @@ class ConfigMixin:
         # TODO: remove this when we remove the deprecation warning, and the `kwargs` argument,
         # or solve in a more general way.
         kwargs.pop("kwargs", None)
+
         if not hasattr(self, "_internal_dict"):
             internal_dict = kwargs
         else:
@@ -550,6 +551,9 @@ class ConfigMixin:
             return value
 
         config_dict = {k: to_json_saveable(v) for k, v in config_dict.items()}
+        # Don't save "_ignore_files"
+        config_dict.pop("_ignore_files", None)
+
         return json.dumps(config_dict, indent=2, sort_keys=True) + "\n"
 
     def to_json_file(self, json_file_path: Union[str, os.PathLike]):

src/diffusers/models/embeddings.py

@@ -377,3 +377,69 @@ class CombinedTimestepLabelEmbeddings(nn.Module):
         conditioning = timesteps_emb + class_labels  # (N, D)
 
         return conditioning
+
+
+class TextTimeEmbedding(nn.Module):
+    def __init__(self, encoder_dim: int, time_embed_dim: int, num_heads: int = 64):
+        super().__init__()
+        self.norm1 = nn.LayerNorm(encoder_dim)
+        self.pool = AttentionPooling(num_heads, encoder_dim)
+        self.proj = nn.Linear(encoder_dim, time_embed_dim)
+        self.norm2 = nn.LayerNorm(time_embed_dim)
+
+    def forward(self, hidden_states):
+        hidden_states = self.norm1(hidden_states)
+        hidden_states = self.pool(hidden_states)
+        hidden_states = self.proj(hidden_states)
+        hidden_states = self.norm2(hidden_states)
+        return hidden_states
+
+
+class AttentionPooling(nn.Module):
+    # Copied from https://github.com/deep-floyd/IF/blob/2f91391f27dd3c468bf174be5805b4cc92980c0b/deepfloyd_if/model/nn.py#L54
+
+    def __init__(self, num_heads, embed_dim, dtype=None):
+        super().__init__()
+        self.dtype = dtype
+        self.positional_embedding = nn.Parameter(torch.randn(1, embed_dim) / embed_dim**0.5)
+        self.k_proj = nn.Linear(embed_dim, embed_dim, dtype=self.dtype)
+        self.q_proj = nn.Linear(embed_dim, embed_dim, dtype=self.dtype)
+        self.v_proj = nn.Linear(embed_dim, embed_dim, dtype=self.dtype)
+        self.num_heads = num_heads
+        self.dim_per_head = embed_dim // self.num_heads
+
+    def forward(self, x):
+        bs, length, width = x.size()
+
+        def shape(x):
+            # (bs, length, width) --> (bs, length, n_heads, dim_per_head)
+            x = x.view(bs, -1, self.num_heads, self.dim_per_head)
+            # (bs, length, n_heads, dim_per_head) --> (bs, n_heads, length, dim_per_head)
+            x = x.transpose(1, 2)
+            # (bs, n_heads, length, dim_per_head) --> (bs*n_heads, length, dim_per_head)
+            x = x.reshape(bs * self.num_heads, -1, self.dim_per_head)
+            # (bs*n_heads, length, dim_per_head) --> (bs*n_heads, dim_per_head, length)
+            x = x.transpose(1, 2)
+            return x
+
+        class_token = x.mean(dim=1, keepdim=True) + self.positional_embedding.to(x.dtype)
+        x = torch.cat([class_token, x], dim=1)  # (bs, length+1, width)
+
+        # (bs*n_heads, class_token_length, dim_per_head)
+        q = shape(self.q_proj(class_token))
+        # (bs*n_heads, length+class_token_length, dim_per_head)
+        k = shape(self.k_proj(x))
+        v = shape(self.v_proj(x))
+
+        # (bs*n_heads, class_token_length, length+class_token_length):
+        scale = 1 / math.sqrt(math.sqrt(self.dim_per_head))
+        weight = torch.einsum("bct,bcs->bts", q * scale, k * scale)  # More stable with f16 than dividing afterwards
+        weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
+
+        # (bs*n_heads, dim_per_head, class_token_length)
+        a = torch.einsum("bts,bcs->bct", weight, v)
+
+        # (bs, length+1, width)
+        a = a.reshape(bs, -1, 1).transpose(1, 2)
+
+        return a[:, 0, :]  # cls_token

src/diffusers/models/modeling_utils.py

@@ -15,6 +15,7 @@
 # limitations under the License.
 
 import inspect
+import itertools
 import os
 from functools import partial
 from typing import Any, Callable, List, Optional, Tuple, Union
@@ -60,7 +61,8 @@ if is_safetensors_available():
 
 def get_parameter_device(parameter: torch.nn.Module):
     try:
-        return next(parameter.parameters()).device
+        parameters_and_buffers = itertools.chain(parameter.parameters(), parameter.buffers())
+        return next(parameters_and_buffers).device
     except StopIteration:
         # For torch.nn.DataParallel compatibility in PyTorch 1.5
 
@@ -75,7 +77,8 @@ def get_parameter_device(parameter: torch.nn.Module):
 
 def get_parameter_dtype(parameter: torch.nn.Module):
     try:
-        return next(parameter.parameters()).dtype
+        parameters_and_buffers = itertools.chain(parameter.parameters(), parameter.buffers())
+        return next(parameters_and_buffers).dtype
     except StopIteration:
         # For torch.nn.DataParallel compatibility in PyTorch 1.5
 

src/diffusers/models/unet_2d_condition.py

@@ -23,7 +23,7 @@ from ..configuration_utils import ConfigMixin, register_to_config
 from ..loaders import UNet2DConditionLoadersMixin
 from ..utils import BaseOutput, logging
 from .attention_processor import AttentionProcessor, AttnProcessor
-from .embeddings import GaussianFourierProjection, TimestepEmbedding, Timesteps
+from .embeddings import GaussianFourierProjection, TextTimeEmbedding, TimestepEmbedding, Timesteps
 from .modeling_utils import ModelMixin
 from .unet_2d_blocks import (
     CrossAttnDownBlock2D,
@@ -97,11 +97,16 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
         class_embed_type (`str`, *optional*, defaults to None):
             The type of class embedding to use which is ultimately summed with the time embeddings. Choose from `None`,
             `"timestep"`, `"identity"`, `"projection"`, or `"simple_projection"`.
+        addition_embed_type (`str`, *optional*, defaults to None):
+            Configures an optional embedding which will be summed with the time embeddings. Choose from `None` or
+            "text". "text" will use the `TextTimeEmbedding` layer.
         num_class_embeds (`int`, *optional*, defaults to None):
             Input dimension of the learnable embedding matrix to be projected to `time_embed_dim`, when performing
             class conditioning with `class_embed_type` equal to `None`.
         time_embedding_type (`str`, *optional*, default to `positional`):
             The type of position embedding to use for timesteps. Choose from `positional` or `fourier`.
+        time_embedding_dim (`int`, *optional*, default to `None`):
+            An optional override for the dimension of the projected time embedding.
         time_embedding_act_fn (`str`, *optional*, default to `None`):
             Optional activation function to use on the time embeddings only one time before they as passed to the rest
             of the unet. Choose from `silu`, `mish`, `gelu`, and `swish`.
@@ -155,12 +160,14 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
         dual_cross_attention: bool = False,
         use_linear_projection: bool = False,
         class_embed_type: Optional[str] = None,
+        addition_embed_type: Optional[str] = None,
         num_class_embeds: Optional[int] = None,
         upcast_attention: bool = False,
         resnet_time_scale_shift: str = "default",
         resnet_skip_time_act: bool = False,
         resnet_out_scale_factor: int = 1.0,
         time_embedding_type: str = "positional",
+        time_embedding_dim: Optional[int] = None,
         time_embedding_act_fn: Optional[str] = None,
         timestep_post_act: Optional[str] = None,
         time_cond_proj_dim: Optional[int] = None,
@@ -170,6 +177,7 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
         class_embeddings_concat: bool = False,
         mid_block_only_cross_attention: Optional[bool] = None,
         cross_attention_norm: Optional[str] = None,
+        addition_embed_type_num_heads=64,
     ):
         super().__init__()
 
@@ -214,7 +222,7 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
 
         # time
         if time_embedding_type == "fourier":
-            time_embed_dim = block_out_channels[0] * 2
+            time_embed_dim = time_embedding_dim or block_out_channels[0] * 2
             if time_embed_dim % 2 != 0:
                 raise ValueError(f"`time_embed_dim` should be divisible by 2, but is {time_embed_dim}.")
             self.time_proj = GaussianFourierProjection(
@@ -222,7 +230,7 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
             )
             timestep_input_dim = time_embed_dim
         elif time_embedding_type == "positional":
-            time_embed_dim = block_out_channels[0] * 4
+            time_embed_dim = time_embedding_dim or block_out_channels[0] * 4
 
             self.time_proj = Timesteps(block_out_channels[0], flip_sin_to_cos, freq_shift)
             timestep_input_dim = block_out_channels[0]
@@ -273,6 +281,18 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
         else:
             self.class_embedding = None
 
+        if addition_embed_type == "text":
+            if encoder_hid_dim is not None:
+                text_time_embedding_from_dim = encoder_hid_dim
+            else:
+                text_time_embedding_from_dim = cross_attention_dim
+
+            self.add_embedding = TextTimeEmbedding(
+                text_time_embedding_from_dim, time_embed_dim, num_heads=addition_embed_type_num_heads
+            )
+        elif addition_embed_type is not None:
+            raise ValueError(f"addition_embed_type: {addition_embed_type} must be None or 'text'.")
+
         if time_embedding_act_fn is None:
             self.time_embed_act = None
         elif time_embedding_act_fn == "swish":
@@ -684,6 +704,10 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
             else:
                 emb = emb + class_emb
 
+        if self.config.addition_embed_type == "text":
+            aug_emb = self.add_embedding(encoder_hidden_states)
+            emb = emb + aug_emb
+
         if self.time_embed_act is not None:
             emb = self.time_embed_act(emb)
 

src/diffusers/pipelines/__init__.py

@@ -44,6 +44,14 @@ except OptionalDependencyNotAvailable:
 else:
     from .alt_diffusion import AltDiffusionImg2ImgPipeline, AltDiffusionPipeline
     from .audioldm import AudioLDMPipeline
+    from .deepfloyd_if import (
+        IFImg2ImgPipeline,
+        IFImg2ImgSuperResolutionPipeline,
+        IFInpaintingPipeline,
+        IFInpaintingSuperResolutionPipeline,
+        IFPipeline,
+        IFSuperResolutionPipeline,
+    )
     from .latent_diffusion import LDMTextToImagePipeline
     from .paint_by_example import PaintByExamplePipeline
     from .semantic_stable_diffusion import SemanticStableDiffusionPipeline

src/diffusers/pipelines/deepfloyd_if/__init__.py

+from dataclasses import dataclass
+from typing import List, Optional, Union
+
+import numpy as np
+import PIL
+
+from ...utils import BaseOutput, OptionalDependencyNotAvailable, is_torch_available, is_transformers_available
+from .timesteps import (
+    fast27_timesteps,
+    smart27_timesteps,
+    smart50_timesteps,
+    smart100_timesteps,
+    smart185_timesteps,
+    super27_timesteps,
+    super40_timesteps,
+    super100_timesteps,
+)
+
+
+@dataclass
+class IFPipelineOutput(BaseOutput):
+    """
+    Args:
+    Output class for Stable Diffusion pipelines.
+        images (`List[PIL.Image.Image]` or `np.ndarray`)
+            List of denoised PIL images of length `batch_size` or numpy array of shape `(batch_size, height, width,
+            num_channels)`. PIL images or numpy array present the denoised images of the diffusion pipeline.
+        nsfw_detected (`List[bool]`)
+            List of flags denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content or a watermark. `None` if safety checking could not be performed.
+        watermark_detected (`List[bool]`)
+            List of flags denoting whether the corresponding generated image likely has a watermark. `None` if safety
+            checking could not be performed.
+    """
+
+    images: Union[List[PIL.Image.Image], np.ndarray]
+    nsfw_detected: Optional[List[bool]]
+    watermark_detected: Optional[List[bool]]
+
+
+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 .pipeline_if import IFPipeline
+    from .pipeline_if_img2img import IFImg2ImgPipeline
+    from .pipeline_if_img2img_superresolution import IFImg2ImgSuperResolutionPipeline
+    from .pipeline_if_inpainting import IFInpaintingPipeline
+    from .pipeline_if_inpainting_superresolution import IFInpaintingSuperResolutionPipeline
+    from .pipeline_if_superresolution import IFSuperResolutionPipeline
+    from .safety_checker import IFSafetyChecker
+    from .watermark import IFWatermarker

src/diffusers/pipelines/deepfloyd_if/safety_checker.py

+import numpy as np
+import torch
+import torch.nn as nn
+from transformers import CLIPConfig, CLIPVisionModelWithProjection, PreTrainedModel
+
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class IFSafetyChecker(PreTrainedModel):
+    config_class = CLIPConfig
+
+    _no_split_modules = ["CLIPEncoderLayer"]
+
+    def __init__(self, config: CLIPConfig):
+        super().__init__(config)
+
+        self.vision_model = CLIPVisionModelWithProjection(config.vision_config)
+
+        self.p_head = nn.Linear(config.vision_config.projection_dim, 1)
+        self.w_head = nn.Linear(config.vision_config.projection_dim, 1)
+
+    @torch.no_grad()
+    def forward(self, clip_input, images, p_threshold=0.5, w_threshold=0.5):
+        image_embeds = self.vision_model(clip_input)[0]
+
+        nsfw_detected = self.p_head(image_embeds)
+        nsfw_detected = nsfw_detected.flatten()
+        nsfw_detected = nsfw_detected > p_threshold
+        nsfw_detected = nsfw_detected.tolist()
+
+        if any(nsfw_detected):
+            logger.warning(
+                "Potential NSFW content was detected in one or more images. A black image will be returned instead."
+                " Try again with a different prompt and/or seed."
+            )
+
+        for idx, nsfw_detected_ in enumerate(nsfw_detected):
+            if nsfw_detected_:
+                images[idx] = np.zeros(images[idx].shape)
+
+        watermark_detected = self.w_head(image_embeds)
+        watermark_detected = watermark_detected.flatten()
+        watermark_detected = watermark_detected > w_threshold
+        watermark_detected = watermark_detected.tolist()
+
+        if any(watermark_detected):
+            logger.warning(
+                "Potential watermarked content was detected in one or more images. A black image will be returned instead."
+                " Try again with a different prompt and/or seed."
+            )
+
+        for idx, watermark_detected_ in enumerate(watermark_detected):
+            if watermark_detected_:
+                images[idx] = np.zeros(images[idx].shape)
+
+        return images, nsfw_detected, watermark_detected