Merge branch 'main' into safety_checker

Author: suzukimain

.github/workflows/push_tests.yml

@@ -81,7 +81,7 @@ jobs:
       - name: Environment
         run: |
           python utils/print_env.py
-      - name: Slow PyTorch CUDA checkpoint tests on Ubuntu
+      - name: PyTorch CUDA checkpoint tests on Ubuntu
         env:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
           # https://pytorch.org/docs/stable/notes/randomness.html#avoiding-nondeterministic-algorithms
@@ -184,7 +184,7 @@ jobs:
       run: |
         python utils/print_env.py
 
-    - name: Run slow Flax TPU tests
+    - name: Run Flax TPU tests
       env:
         HF_TOKEN: ${{ secrets.HF_TOKEN }}
       run: |
@@ -232,7 +232,7 @@ jobs:
       run: |
         python utils/print_env.py
 
-    - name: Run slow ONNXRuntime CUDA tests
+    - name: Run ONNXRuntime CUDA tests
       env:
         HF_TOKEN: ${{ secrets.HF_TOKEN }}
       run: |

docs/source/en/_toctree.yml

@@ -188,6 +188,8 @@
       title: Metal Performance Shaders (MPS)
     - local: optimization/habana
       title: Habana Gaudi
+    - local: optimization/neuron
+      title: AWS Neuron
     title: Optimized hardware
   title: Accelerate inference and reduce memory
 - sections:

docs/source/en/optimization/neuron.md

@@ -0,0 +1,61 @@
+<!--Copyright 2024 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.
+-->
+
+# AWS Neuron
+
+Diffusers functionalities are available on [AWS Inf2 instances](https://aws.amazon.com/ec2/instance-types/inf2/), which are EC2 instances powered by [Neuron machine learning accelerators](https://aws.amazon.com/machine-learning/inferentia/). These instances aim to provide better compute performance (higher throughput, lower latency) with good cost-efficiency, making them good candidates for AWS users to deploy diffusion models to production.
+
+[Optimum Neuron](https://huggingface.co/docs/optimum-neuron/en/index) is the interface between Hugging Face libraries and AWS Accelerators, including AWS [Trainium](https://aws.amazon.com/machine-learning/trainium/) and AWS [Inferentia](https://aws.amazon.com/machine-learning/inferentia/). It supports many of the features in Diffusers with similar APIs, so it is easier to learn if you're already familiar with Diffusers. Once you have created an AWS Inf2 instance, install Optimum Neuron.
+
+```bash
+python -m pip install --upgrade-strategy eager optimum[neuronx]
+```
+
+<Tip>
+
+We provide pre-built [Hugging Face Neuron Deep Learning AMI](https://aws.amazon.com/marketplace/pp/prodview-gr3e6yiscria2) (DLAMI) and Optimum Neuron containers for Amazon SageMaker. It's recommended to correctly set up your environment.
+
+</Tip>
+
+The example below demonstrates how to generate images with the Stable Diffusion XL model on an inf2.8xlarge instance (you can switch to cheaper inf2.xlarge instances once the model is compiled). To generate some images, use the [`~optimum.neuron.NeuronStableDiffusionXLPipeline`] class, which is similar to the [`StableDiffusionXLPipeline`] class in Diffusers.
+
+Unlike Diffusers, you need to compile models in the pipeline to the Neuron format, `.neuron`. Launch the following command to export the model to the `.neuron` format.
+
+```bash
+optimum-cli export neuron --model stabilityai/stable-diffusion-xl-base-1.0 \
+  --batch_size 1 \
+  --height 1024 `# height in pixels of generated image, eg. 768, 1024` \
+  --width 1024 `# width in pixels of generated image, eg. 768, 1024` \
+  --num_images_per_prompt 1 `# number of images to generate per prompt, defaults to 1` \
+  --auto_cast matmul `# cast only matrix multiplication operations` \
+  --auto_cast_type bf16 `# cast operations from FP32 to BF16` \
+  sd_neuron_xl/
+```
+
+Now generate some images with the pre-compiled SDXL model.
+
+```python
+>>> from optimum.neuron import NeuronStableDiffusionXLPipeline
+
+>>> stable_diffusion_xl = NeuronStableDiffusionXLPipeline.from_pretrained("sd_neuron_xl/")
+>>> prompt = "a pig with wings flying in floating US dollar banknotes in the air, skyscrapers behind, warm color palette, muted colors, detailed, 8k"
+>>> image = stable_diffusion_xl(prompt).images[0]
+```
+
+<img
+  src="https://huggingface.co/datasets/Jingya/document_images/resolve/main/optimum/neuron/sdxl_pig.png"
+  width="256"
+  height="256"
+  alt="peggy generated by sdxl on inf2"
+/>
+
+Feel free to check out more guides and examples on different use cases from the Optimum Neuron [documentation](https://huggingface.co/docs/optimum-neuron/en/inference_tutorials/stable_diffusion#generate-images-with-stable-diffusion-models-on-aws-inferentia)!

examples/advanced_diffusion_training/train_dreambooth_lora_flux_advanced.py

@@ -2198,8 +2198,8 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
 
                 latent_image_ids = FluxPipeline._prepare_latent_image_ids(
                     model_input.shape[0],
-                    model_input.shape[2],
-                    model_input.shape[3],
+                    model_input.shape[2] // 2,
+                    model_input.shape[3] // 2,
                     accelerator.device,
                     weight_dtype,
                 )
@@ -2253,8 +2253,8 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                 )[0]
                 model_pred = FluxPipeline._unpack_latents(
                     model_pred,
-                    height=int(model_input.shape[2] * vae_scale_factor / 2),
-                    width=int(model_input.shape[3] * vae_scale_factor / 2),
+                    height=model_input.shape[2] * vae_scale_factor,
+                    width=model_input.shape[3] * vae_scale_factor,
                     vae_scale_factor=vae_scale_factor,
                 )
 

examples/controlnet/train_controlnet_flux.py

@@ -1256,8 +1256,8 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
 
                 latent_image_ids = FluxControlNetPipeline._prepare_latent_image_ids(
                     batch_size=pixel_latents_tmp.shape[0],
-                    height=pixel_latents_tmp.shape[2],
-                    width=pixel_latents_tmp.shape[3],
+                    height=pixel_latents_tmp.shape[2] // 2,
+                    width=pixel_latents_tmp.shape[3] // 2,
                     device=pixel_values.device,
                     dtype=pixel_values.dtype,
                 )

examples/dreambooth/train_dreambooth_flux.py

@@ -1540,12 +1540,12 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                 model_input = (model_input - vae.config.shift_factor) * vae.config.scaling_factor
                 model_input = model_input.to(dtype=weight_dtype)
 
-                vae_scale_factor = 2 ** (len(vae.config.block_out_channels))
+                vae_scale_factor = 2 ** (len(vae.config.block_out_channels) - 1)
 
                 latent_image_ids = FluxPipeline._prepare_latent_image_ids(
                     model_input.shape[0],
-                    model_input.shape[2],
-                    model_input.shape[3],
+                    model_input.shape[2] // 2,
+                    model_input.shape[3] // 2,
                     accelerator.device,
                     weight_dtype,
                 )
@@ -1601,8 +1601,8 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                 # upscaling height & width as discussed in https://github.com/huggingface/diffusers/pull/9257#discussion_r1731108042
                 model_pred = FluxPipeline._unpack_latents(
                     model_pred,
-                    height=int(model_input.shape[2] * vae_scale_factor / 2),
-                    width=int(model_input.shape[3] * vae_scale_factor / 2),
+                    height=model_input.shape[2] * vae_scale_factor,
+                    width=model_input.shape[3] * vae_scale_factor,
                     vae_scale_factor=vae_scale_factor,
                 )
 

examples/dreambooth/train_dreambooth_lora_flux.py

@@ -1645,12 +1645,12 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                 model_input = (model_input - vae_config_shift_factor) * vae_config_scaling_factor
                 model_input = model_input.to(dtype=weight_dtype)
 
-                vae_scale_factor = 2 ** (len(vae_config_block_out_channels))
+                vae_scale_factor = 2 ** (len(vae_config_block_out_channels) - 1)
 
                 latent_image_ids = FluxPipeline._prepare_latent_image_ids(
                     model_input.shape[0],
-                    model_input.shape[2],
-                    model_input.shape[3],
+                    model_input.shape[2] // 2,
+                    model_input.shape[3] // 2,
                     accelerator.device,
                     weight_dtype,
                 )
@@ -1704,8 +1704,8 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                 )[0]
                 model_pred = FluxPipeline._unpack_latents(
                     model_pred,
-                    height=int(model_input.shape[2] * vae_scale_factor / 2),
-                    width=int(model_input.shape[3] * vae_scale_factor / 2),
+                    height=model_input.shape[2] * vae_scale_factor,
+                    width=model_input.shape[3] * vae_scale_factor,
                     vae_scale_factor=vae_scale_factor,
                 )
 

examples/dreambooth/train_dreambooth_lora_sd3.py

@@ -86,6 +86,15 @@ def save_model_card(
     validation_prompt=None,
     repo_folder=None,
 ):
+    if "large" in base_model:
+        model_variant = "SD3.5-Large"
+        license_url = "https://huggingface.co/stabilityai/stable-diffusion-3.5-large/blob/main/LICENSE.md"
+        variant_tags = ["sd3.5-large", "sd3.5", "sd3.5-diffusers"]
+    else:
+        model_variant = "SD3"
+        license_url = "https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/LICENSE.md"
+        variant_tags = ["sd3", "sd3-diffusers"]
+
     widget_dict = []
     if images is not None:
         for i, image in enumerate(images):
@@ -95,7 +104,7 @@ def save_model_card(
             )
 
     model_description = f"""
-# SD3 DreamBooth LoRA - {repo_id}
+# {model_variant} DreamBooth LoRA - {repo_id}
 
 <Gallery />
 
@@ -120,7 +129,7 @@ def save_model_card(
 ```py
 from diffusers import AutoPipelineForText2Image
 import torch
-pipeline = AutoPipelineForText2Image.from_pretrained('stabilityai/stable-diffusion-3-medium-diffusers', torch_dtype=torch.float16).to('cuda')
+pipeline = AutoPipelineForText2Image.from_pretrained({base_model}, torch_dtype=torch.float16).to('cuda')
 pipeline.load_lora_weights('{repo_id}', weight_name='pytorch_lora_weights.safetensors')
 image = pipeline('{validation_prompt if validation_prompt else instance_prompt}').images[0]
 ```
@@ -135,7 +144,7 @@ def save_model_card(
 
 ## License
 
-Please adhere to the licensing terms as described [here](https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/LICENSE).
+Please adhere to the licensing terms as described [here]({license_url}).
 """
     model_card = load_or_create_model_card(
         repo_id_or_path=repo_id,
@@ -151,11 +160,11 @@ def save_model_card(
         "diffusers-training",
         "diffusers",
         "lora",
-        "sd3",
-        "sd3-diffusers",
         "template:sd-lora",
     ]
 
+    tags += variant_tags
+
     model_card = populate_model_card(model_card, tags=tags)
     model_card.save(os.path.join(repo_folder, "README.md"))
 

examples/dreambooth/train_dreambooth_sd3.py

@@ -77,6 +77,15 @@ def save_model_card(
     validation_prompt=None,
     repo_folder=None,
 ):
+    if "large" in base_model:
+        model_variant = "SD3.5-Large"
+        license_url = "https://huggingface.co/stabilityai/stable-diffusion-3.5-large/blob/main/LICENSE.md"
+        variant_tags = ["sd3.5-large", "sd3.5", "sd3.5-diffusers"]
+    else:
+        model_variant = "SD3"
+        license_url = "https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/LICENSE.md"
+        variant_tags = ["sd3", "sd3-diffusers"]
+
     widget_dict = []
     if images is not None:
         for i, image in enumerate(images):
@@ -86,7 +95,7 @@ def save_model_card(
             )
 
     model_description = f"""
-# SD3 DreamBooth - {repo_id}
+# {model_variant} DreamBooth - {repo_id}
 
 <Gallery />
 
@@ -113,7 +122,7 @@ def save_model_card(
 
 ## License
 
-Please adhere to the licensing terms as described `[here](https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/LICENSE)`.
+Please adhere to the licensing terms as described `[here]({license_url})`.
 """
     model_card = load_or_create_model_card(
         repo_id_or_path=repo_id,
@@ -128,10 +137,9 @@ def save_model_card(
         "text-to-image",
         "diffusers-training",
         "diffusers",
-        "sd3",
-        "sd3-diffusers",
         "template:sd-lora",
     ]
+    tags += variant_tags
 
     model_card = populate_model_card(model_card, tags=tags)
     model_card.save(os.path.join(repo_folder, "README.md"))