Euler A redesign merge

.github/workflows/pr_tests.yml

@@ -21,7 +21,7 @@ jobs:
     runs-on: [ self-hosted, docker-gpu ]
     container:
       image: python:3.7
-      options: --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+      options: --shm-size "16gb" --ipc host -v /mnt/hf_cache:/mnt/cache/
 
     steps:
     - name: Checkout diffusers

.github/workflows/push_tests.yml

@@ -15,14 +15,10 @@ env:
 jobs:
   run_tests_single_gpu:
     name: Diffusers tests
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [ single-gpu ]
-    runs-on: [ self-hosted, docker-gpu, '${{ matrix.machine_type }}' ]
+    runs-on: [ self-hosted, docker-gpu, single-gpu ]
     container:
       image: nvcr.io/nvidia/pytorch:22.07-py3
-      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/hf_cache:/mnt/cache
 
     steps:
     - name: Checkout diffusers
@@ -66,14 +62,10 @@ jobs:
 
   run_examples_single_gpu:
     name: Examples tests
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [ single-gpu ]
-    runs-on: [ self-hosted, docker-gpu, '${{ matrix.machine_type }}' ]
+    runs-on: [ self-hosted, docker-gpu, single-gpu ]
     container:
       image: nvcr.io/nvidia/pytorch:22.07-py3
-      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/hf_cache:/mnt/cache
 
     steps:
     - name: Checkout diffusers

README.md

@@ -74,17 +74,18 @@ You need to accept the model license before downloading or using the Stable Diff
 
 ### Text-to-Image generation with Stable Diffusion
 
+We recommend using the model in [half-precision (`fp16`)](https://pytorch.org/blog/accelerating-training-on-nvidia-gpus-with-pytorch-automatic-mixed-precision/) as it gives almost always the same results as full
+precision while being roughly twice as fast and requiring half the amount of GPU RAM.
+
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline
 
-pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True)
+pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", torch_type=torch.float16, revision="fp16")
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 ```
 
 **Note**: If you don't want to use the token, you can also simply download the model weights
@@ -104,30 +105,27 @@ pipe = StableDiffusionPipeline.from_pretrained("./stable-diffusion-v1-4")
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 ```
 
-If you are limited by GPU memory, you might want to consider using the model in `fp16` as 
-well as chunking the attention computation.
+If you are limited by GPU memory, you might want to consider chunking the attention computation in addition 
+to using `fp16`.
 The following snippet should result in less than 4GB VRAM.
 
 ```python
 pipe = StableDiffusionPipeline.from_pretrained(
     "CompVis/stable-diffusion-v1-4", 
     revision="fp16", 
     torch_dtype=torch.float16,
-    use_auth_token=True
 )
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
 pipe.enable_attention_slicing()
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 ```
 
-Finally, if you wish to use a different scheduler, you can simply instantiate
+If you wish to use a different scheduler, you can simply instantiate
 it before the pipeline and pass it to `from_pretrained`.
 
 ```python
@@ -144,13 +142,29 @@ pipe = StableDiffusionPipeline.from_pretrained(
     revision="fp16", 
     torch_dtype=torch.float16,
     scheduler=lms,
-    use_auth_token=True
 )
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
+
+image.save("astronaut_rides_horse.png")
+```
+
+If you want to run Stable Diffusion on CPU or you want to have maximum precision on GPU, 
+please run the model in the default *full-precision* setting:
+
+```python
+# make sure you're logged in with `huggingface-cli login`
+from diffusers import StableDiffusionPipeline
+
+pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4")
+
+# disable the following line if you run on CPU
+pipe = pipe.to("cuda")
+
+prompt = "a photo of an astronaut riding a horse on mars"
+image = pipe(prompt).images[0]  
 
 image.save("astronaut_rides_horse.png")
 ```
@@ -160,7 +174,6 @@ image.save("astronaut_rides_horse.png")
 The `StableDiffusionImg2ImgPipeline` lets you pass a text prompt and an initial image to condition the generation of new images.
 
 ```python
-from torch import autocast
 import requests
 import torch
 from PIL import Image
@@ -175,10 +188,9 @@ pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
     model_id_or_path,
     revision="fp16", 
     torch_dtype=torch.float16,
-    use_auth_token=True
 )
 # or download via git clone https://huggingface.co/CompVis/stable-diffusion-v1-4
-# and pass `model_id_or_path="./stable-diffusion-v1-4"` without having to use `use_auth_token=True`.
+# and pass `model_id_or_path="./stable-diffusion-v1-4"`.
 pipe = pipe.to(device)
 
 # let's download an initial image
@@ -190,8 +202,7 @@ init_image = init_image.resize((768, 512))
 
 prompt = "A fantasy landscape, trending on artstation"
 
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
+images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
 
 images[0].save("fantasy_landscape.png")
 ```
@@ -204,7 +215,6 @@ The `StableDiffusionInpaintPipeline` lets you edit specific parts of an image by
 ```python
 from io import BytesIO
 
-from torch import autocast
 import torch
 import requests
 import PIL
@@ -227,15 +237,13 @@ pipe = StableDiffusionInpaintPipeline.from_pretrained(
     model_id_or_path,
     revision="fp16", 
     torch_dtype=torch.float16,
-    use_auth_token=True
 )
 # or download via git clone https://huggingface.co/CompVis/stable-diffusion-v1-4
-# and pass `model_id_or_path="./stable-diffusion-v1-4"` without having to use `use_auth_token=True`.
+# and pass `model_id_or_path="./stable-diffusion-v1-4"`.
 pipe = pipe.to(device)
 
 prompt = "a cat sitting on a bench"
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
+images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
 
 images[0].save("cat_on_bench.png")
 ```
@@ -258,7 +266,6 @@ If you want to run the code yourself 💻, you can try out:
 - [Text-to-Image Latent Diffusion](https://huggingface.co/CompVis/ldm-text2im-large-256)
 ```python
 # !pip install diffusers transformers
-from torch import autocast
 from diffusers import DiffusionPipeline
 
 device = "cuda"
@@ -270,16 +277,14 @@ ldm = ldm.to(device)
 
 # run pipeline in inference (sample random noise and denoise)
 prompt = "A painting of a squirrel eating a burger"
-with autocast(device):
-    image = ldm([prompt], num_inference_steps=50, eta=0.3, guidance_scale=6).images[0]
+image = ldm([prompt], num_inference_steps=50, eta=0.3, guidance_scale=6).images[0]
 
 # save image
 image.save("squirrel.png")
 ```
 - [Unconditional Diffusion with discrete scheduler](https://huggingface.co/google/ddpm-celebahq-256)
 ```python
 # !pip install diffusers
-from torch import autocast
 from diffusers import DDPMPipeline, DDIMPipeline, PNDMPipeline
 
 model_id = "google/ddpm-celebahq-256"
@@ -290,8 +295,7 @@ ddpm = DDPMPipeline.from_pretrained(model_id)  # you can replace DDPMPipeline wi
 ddpm.to(device)
 
 # run pipeline in inference (sample random noise and denoise)
-with autocast("cuda"):
-    image = ddpm().images[0]
+image = ddpm().images[0]
 
 # save image
 image.save("ddpm_generated_image.png")
@@ -377,3 +381,16 @@ This library concretizes previous work by many different authors and would not h
 - @yang-song's Score-VE and Score-VP implementations, available [here](https://github.com/yang-song/score_sde_pytorch)
 
 We also want to thank @heejkoo for the very helpful overview of papers, code and resources on diffusion models, available [here](https://github.com/heejkoo/Awesome-Diffusion-Models) as well as @crowsonkb and @rromb for useful discussions and insights.
+
+## Citation
+
+```bibtex
+@misc{von-platen-etal-2022-diffusers,
+  author = {Patrick von Platen and Suraj Patil and Anton Lozhkov and Pedro Cuenca and Nathan Lambert and Kashif Rasul and Mishig Davaadorj and Thomas Wolf},
+  title = {Diffusers: State-of-the-art diffusion models},
+  year = {2022},
+  publisher = {GitHub},
+  journal = {GitHub repository},
+  howpublished = {\url{https://github.com/huggingface/diffusers}}
+}
+```

docs/source/_toctree.yml

@@ -12,6 +12,8 @@
       title: "Loading Pipelines, Models, and Schedulers"
     - local: using-diffusers/configuration
       title: "Configuring Pipelines, Models, and Schedulers"
+    - local: using-diffusers/custom_pipelines
+      title: "Loading and Creating Custom Pipelines"
     title: "Loading"
   - sections:
     - local: using-diffusers/unconditional_image_generation

docs/source/api/pipelines/overview.mdx

@@ -98,15 +98,13 @@ logic including pre-processing, an unrolled diffusion loop, and post-processing
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline, LMSDiscreteScheduler
 
-pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True)
+pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4")
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]
+image = pipe(prompt).images[0]
 
 image.save("astronaut_rides_horse.png")
 ```
@@ -116,7 +114,6 @@ image.save("astronaut_rides_horse.png")
 The `StableDiffusionImg2ImgPipeline` lets you pass a text prompt and an initial image to condition the generation of new images.
 
 ```python
-from torch import autocast
 import requests
 from PIL import Image
 from io import BytesIO
@@ -126,7 +123,7 @@ from diffusers import StableDiffusionImg2ImgPipeline
 # load the pipeline
 device = "cuda"
 pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
-    "CompVis/stable-diffusion-v1-4", revision="fp16", torch_dtype=torch.float16, use_auth_token=True
+    "CompVis/stable-diffusion-v1-4", revision="fp16", torch_dtype=torch.float16
 ).to(device)
 
 # let's download an initial image
@@ -138,8 +135,7 @@ init_image = init_image.resize((768, 512))
 
 prompt = "A fantasy landscape, trending on artstation"
 
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
+images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
 
 images[0].save("fantasy_landscape.png")
 ```
@@ -157,7 +153,6 @@ The `StableDiffusionInpaintPipeline` lets you edit specific parts of an image by
 ```python
 from io import BytesIO
 
-from torch import autocast
 import requests
 import PIL
 
@@ -177,12 +172,11 @@ mask_image = download_image(mask_url).resize((512, 512))
 
 device = "cuda"
 pipe = StableDiffusionInpaintPipeline.from_pretrained(
-    "CompVis/stable-diffusion-v1-4", revision="fp16", torch_dtype=torch.float16, use_auth_token=True
+    "CompVis/stable-diffusion-v1-4", revision="fp16", torch_dtype=torch.float16
 ).to(device)
 
 prompt = "a cat sitting on a bench"
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
+images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
 
 images[0].save("cat_on_bench.png")
 ```

docs/source/api/schedulers.mdx

@@ -36,16 +36,15 @@ This allows for rapid experimentation and cleaner abstractions in the code, wher
 To this end, the design of schedulers is such that:
 
 - Schedulers can be used interchangeably between diffusion models in inference to find the preferred trade-off between speed and generation quality.
-- Schedulers are currently by default in PyTorch, but are designed to be framework independent (partial Numpy support currently exists).
+- Schedulers are currently by default in PyTorch, but are designed to be framework independent (partial Jax support currently exists).
 
 
 ## API
 
 The core API for any new scheduler must follow a limited structure.
 - Schedulers should provide one or more `def step(...)` functions that should be called to update the generated sample iteratively.
 - Schedulers should provide a `set_timesteps(...)` method that configures the parameters of a schedule function for a specific inference task.
-- Schedulers should be framework-agnostic, but provide a simple functionality to convert the scheduler into a specific framework, such as PyTorch
-with a `set_format(...)` method.
+- Schedulers should be framework-specific.
 
 The base class [`SchedulerMixin`] implements low level utilities used by multiple schedulers.
 

docs/source/index.mdx

@@ -35,7 +35,7 @@ available a colab notebook to directly try them out.
 | Pipeline | Paper | Tasks | Colab
 |---|---|:---:|:---:|
 | [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 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/training_example.ipynb)
+| [ddim](./api/pipelines/ddim) | [**Denoising Diffusion Implicit Models**](https://arxiv.org/abs/2010.02502) | Unconditional 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_uncond](./api/pipelines/latent_diffusion_uncond) | [**High-Resolution Image Synthesis with Latent Diffusion Models**](https://arxiv.org/abs/2112.10752) | Unconditional Image Generation | 
 | [pndm](./api/pipelines/pndm) | [**Pseudo Numerical Methods for Diffusion Models on Manifolds**](https://arxiv.org/abs/2202.09778) | Unconditional Image Generation |