Merge changes

.github/workflows/benchmark.yml

@@ -7,6 +7,7 @@ on:
 
 env:
   DIFFUSERS_IS_CI: yes
+  HF_HUB_ENABLE_HF_TRANSFER: 1
   HF_HOME: /mnt/cache
   OMP_NUM_THREADS: 8
   MKL_NUM_THREADS: 8

.github/workflows/pr_tests.yml

@@ -22,6 +22,7 @@ concurrency:
 
 env:
   DIFFUSERS_IS_CI: yes
+  HF_HUB_ENABLE_HF_TRANSFER: 1
   OMP_NUM_THREADS: 4
   MKL_NUM_THREADS: 4
   PYTEST_TIMEOUT: 60

.github/workflows/push_tests.yml

@@ -14,6 +14,7 @@ env:
   DIFFUSERS_IS_CI: yes
   OMP_NUM_THREADS: 8
   MKL_NUM_THREADS: 8
+  HF_HUB_ENABLE_HF_TRANSFER: 1
   PYTEST_TIMEOUT: 600
   PIPELINE_USAGE_CUTOFF: 50000
 

.github/workflows/push_tests_fast.yml

@@ -18,6 +18,7 @@ env:
   HF_HOME: /mnt/cache
   OMP_NUM_THREADS: 8
   MKL_NUM_THREADS: 8
+  HF_HUB_ENABLE_HF_TRANSFER: 1
   PYTEST_TIMEOUT: 600
   RUN_SLOW: no
 

.github/workflows/push_tests_mps.yml

@@ -13,6 +13,7 @@ env:
   HF_HOME: /mnt/cache
   OMP_NUM_THREADS: 8
   MKL_NUM_THREADS: 8
+  HF_HUB_ENABLE_HF_TRANSFER: 1
   PYTEST_TIMEOUT: 600
   RUN_SLOW: no
 

docker/diffusers-flax-cpu/Dockerfile

@@ -43,6 +43,7 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         numpy==1.26.4 \
         scipy \
         tensorboard \
-        transformers
+        transformers \
+        hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-flax-tpu/Dockerfile

@@ -45,6 +45,7 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         numpy==1.26.4 \
         scipy \
         tensorboard \
-        transformers
+        transformers \
+        hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-onnxruntime-cpu/Dockerfile

@@ -43,6 +43,7 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         numpy==1.26.4 \
         scipy \
         tensorboard \
-        transformers
+        transformers \
+        hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-onnxruntime-cuda/Dockerfile

@@ -44,6 +44,7 @@ RUN python3.10 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         numpy==1.26.4 \
         scipy \
         tensorboard \
-        transformers
+        transformers \
+        hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-pytorch-compile-cuda/Dockerfile

@@ -44,6 +44,7 @@ RUN python3.10 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
     numpy==1.26.4 \
     scipy \
     tensorboard \
-    transformers
+    transformers \
+    hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-pytorch-cpu/Dockerfile

@@ -44,6 +44,7 @@ RUN python3.10 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         numpy==1.26.4 \
         scipy \
         tensorboard \
-        transformers matplotlib
+        transformers matplotlib  \
+        hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-pytorch-cuda/Dockerfile

@@ -45,6 +45,7 @@ RUN python3.10 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
     scipy \
     tensorboard \
     transformers \
-    pytorch-lightning
+    pytorch-lightning  \
+    hf_transfer
 
 CMD ["/bin/bash"]

docker/diffusers-pytorch-xformers-cuda/Dockerfile

@@ -45,6 +45,7 @@ RUN python3.10 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         scipy \
         tensorboard \
         transformers \
-        xformers
+        xformers  \
+        hf_transfer
 
 CMD ["/bin/bash"]

docs/source/en/_toctree.yml

@@ -56,7 +56,7 @@
   - local: using-diffusers/overview_techniques
     title: Overview
   - local: training/distributed_inference
-    title: Distributed inference with multiple GPUs
+    title: Distributed inference
   - local: using-diffusers/merge_loras
     title: Merge LoRAs
   - local: using-diffusers/scheduler_features

docs/source/en/api/pipelines/pag.md

@@ -55,6 +55,9 @@ Since RegEx is supported as a way for matching layer identifiers, it is crucial
 
 ## StableDiffusionControlNetPAGPipeline
 [[autodoc]] StableDiffusionControlNetPAGPipeline
+
+## StableDiffusionControlNetPAGInpaintPipeline
+[[autodoc]] StableDiffusionControlNetPAGInpaintPipeline
 	- all
 	- __call__
 

docs/source/en/api/pipelines/text_to_video_zero.md

@@ -40,6 +40,7 @@ To generate a video from prompt, run the following Python code:
 ```python
 import torch
 from diffusers import TextToVideoZeroPipeline
+import imageio
 
 model_id = "stable-diffusion-v1-5/stable-diffusion-v1-5"
 pipe = TextToVideoZeroPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda")

docs/source/en/api/schedulers/overview.md

@@ -45,6 +45,15 @@ Many schedulers are implemented from the [k-diffusion](https://github.com/crowso
 | N/A                 | [`DEISMultistepScheduler`]          |                                                                                                               |
 | N/A                 | [`UniPCMultistepScheduler`]         |                                                                                                               |
 
+## Noise schedules and schedule types
+| A1111/k-diffusion        | 🤗 Diffusers                                                               |
+|--------------------------|----------------------------------------------------------------------------|
+| Karras                   | init with `use_karras_sigmas=True`                                         |
+| sgm_uniform              | init with `timestep_spacing="trailing"`                                    |
+| simple                   | init with `timestep_spacing="trailing"`                                    |
+| exponential              | init with `timestep_spacing="linspace"`, `use_exponential_sigmas=True`     |
+| beta                     | init with `timestep_spacing="linspace"`, `use_beta_sigmas=True`            |
+
 All schedulers are built from the base [`SchedulerMixin`] class which implements low level utilities shared by all schedulers.
 
 ## SchedulerMixin

docs/source/en/community_projects.md

@@ -75,4 +75,8 @@ Happy exploring, and thank you for being part of the Diffusers community!
     <td><a href="https://github.com/cumulo-autumn/StreamDiffusion"> StreamDiffusion </a></td>
     <td>A Pipeline-Level Solution for Real-Time Interactive Generation</td>
   </tr>
+  <tr style="border-top: 2px solid black">
+    <td><a href="https://github.com/Netwrck/stable-diffusion-server"> Stable Diffusion Server </a></td>
+    <td>A server configured for Inpainting/Generation/img2img with one stable diffusion model</td>
+  </tr>
 </table>

docs/source/en/optimization/coreml.md

@@ -95,7 +95,7 @@ print(f"Model downloaded at {model_path}")
 Once you have downloaded a snapshot of the model, you can test it using Apple's Python script.
 
 ```shell
-python -m python_coreml_stable_diffusion.pipeline --prompt "a photo of an astronaut riding a horse on mars" -i models/coreml-stable-diffusion-v1-4_original_packages -o </path/to/output/image> --compute-unit CPU_AND_GPU --seed 93
+python -m python_coreml_stable_diffusion.pipeline --prompt "a photo of an astronaut riding a horse on mars" -i ./models/coreml-stable-diffusion-v1-4_original_packages/original/packages -o </path/to/output/image> --compute-unit CPU_AND_GPU --seed 93
 ```
 
 Pass the path of the downloaded checkpoint with `-i` flag to the script. `--compute-unit` indicates the hardware you want to allow for inference. It must be one of the following options: `ALL`, `CPU_AND_GPU`, `CPU_ONLY`, `CPU_AND_NE`. You may also provide an optional output path, and a seed for reproducibility.

docs/source/en/training/distributed_inference.md

@@ -10,7 +10,7 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
 specific language governing permissions and limitations under the License.
 -->
 
-# Distributed inference with multiple GPUs
+# Distributed inference
 
 On distributed setups, you can run inference across multiple GPUs with 🤗 [Accelerate](https://huggingface.co/docs/accelerate/index) or [PyTorch Distributed](https://pytorch.org/tutorials/beginner/dist_overview.html), which is useful for generating with multiple prompts in parallel.
 
@@ -109,3 +109,131 @@ torchrun run_distributed.py --nproc_per_node=2
 
 > [!TIP]
 > You can use `device_map` within a [`DiffusionPipeline`] to distribute its model-level components on multiple devices. Refer to the [Device placement](../tutorials/inference_with_big_models#device-placement) guide to learn more.
+
+## Model sharding
+
+Modern diffusion systems such as [Flux](../api/pipelines/flux) are very large and have multiple models. For example, [Flux.1-Dev](https://hf.co/black-forest-labs/FLUX.1-dev) is made up of two text encoders - [T5-XXL](https://hf.co/google/t5-v1_1-xxl) and [CLIP-L](https://hf.co/openai/clip-vit-large-patch14) - a [diffusion transformer](../api/models/flux_transformer), and a [VAE](../api/models/autoencoderkl). With a model this size, it can be challenging to run inference on consumer GPUs.
+
+Model sharding is a technique that distributes models across GPUs when the models don't fit on a single GPU. The example below assumes two 16GB GPUs are available for inference.
+
+Start by computing the text embeddings with the text encoders. Keep the text encoders on two GPUs by setting `device_map="balanced"`. The `balanced` strategy evenly distributes the model on all available GPUs. Use the `max_memory` parameter to allocate the maximum amount of memory for each text encoder on each GPU.
+
+> [!TIP]
+> **Only** load the text encoders for this step! The diffusion transformer and VAE are loaded in a later step to preserve memory.
+
+```py
+from diffusers import FluxPipeline
+import torch
+
+prompt = "a photo of a dog with cat-like look"
+
+pipeline = FluxPipeline.from_pretrained(
+    "black-forest-labs/FLUX.1-dev",
+    transformer=None,
+    vae=None,
+    device_map="balanced",
+    max_memory={0: "16GB", 1: "16GB"},
+    torch_dtype=torch.bfloat16
+)
+with torch.no_grad():
+    print("Encoding prompts.")
+    prompt_embeds, pooled_prompt_embeds, text_ids = pipeline.encode_prompt(
+        prompt=prompt, prompt_2=None, max_sequence_length=512
+    )
+```
+
+Once the text embeddings are computed, remove them from the GPU to make space for the diffusion transformer.
+
+```py
+import gc 
+
+def flush():
+    gc.collect()
+    torch.cuda.empty_cache()
+    torch.cuda.reset_max_memory_allocated()
+    torch.cuda.reset_peak_memory_stats()
+
+del pipeline.text_encoder
+del pipeline.text_encoder_2
+del pipeline.tokenizer
+del pipeline.tokenizer_2
+del pipeline
+
+flush()
+```
+
+Load the diffusion transformer next which has 12.5B parameters. This time, set `device_map="auto"` to automatically distribute the model across two 16GB GPUs. The `auto` strategy is backed by [Accelerate](https://hf.co/docs/accelerate/index) and available as a part of the [Big Model Inference](https://hf.co/docs/accelerate/concept_guides/big_model_inference) feature. It starts by distributing a model across the fastest device first (GPU) before moving to slower devices like the CPU and hard drive if needed. The trade-off of storing model parameters on slower devices is slower inference latency.
+
+```py
+from diffusers import FluxTransformer2DModel
+import torch 
+
+transformer = FluxTransformer2DModel.from_pretrained(
+    "black-forest-labs/FLUX.1-dev", 
+    subfolder="transformer",
+    device_map="auto",
+    torch_dtype=torch.bfloat16
+)
+```
+
+> [!TIP]
+> At any point, you can try `print(pipeline.hf_device_map)` to see how the various models are distributed across devices. This is useful for tracking the device placement of the models.
+
+Add the transformer model to the pipeline for denoising, but set the other model-level components like the text encoders and VAE to `None` because you don't need them yet.
+
+```py
+pipeline = FluxPipeline.from_pretrained(
+    "black-forest-labs/FLUX.1-dev", ,
+    text_encoder=None,
+    text_encoder_2=None,
+    tokenizer=None,
+    tokenizer_2=None,
+    vae=None,
+    transformer=transformer,
+    torch_dtype=torch.bfloat16
+)
+
+print("Running denoising.")
+height, width = 768, 1360
+latents = pipeline(
+    prompt_embeds=prompt_embeds,
+    pooled_prompt_embeds=pooled_prompt_embeds,
+    num_inference_steps=50,
+    guidance_scale=3.5,
+    height=height,
+    width=width,
+    output_type="latent",
+).images
+```
+
+Remove the pipeline and transformer from memory as they're no longer needed.
+
+```py
+del pipeline.transformer
+del pipeline
+
+flush()
+```
+
+Finally, decode the latents with the VAE into an image. The VAE is typically small enough to be loaded on a single GPU.
+
+```py
+from diffusers import AutoencoderKL
+from diffusers.image_processor import VaeImageProcessor
+import torch 
+
+vae = AutoencoderKL.from_pretrained(ckpt_id, subfolder="vae", torch_dtype=torch.bfloat16).to("cuda")
+vae_scale_factor = 2 ** (len(vae.config.block_out_channels))
+image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor)
+
+with torch.no_grad():
+    print("Running decoding.")
+    latents = FluxPipeline._unpack_latents(latents, height, width, vae_scale_factor)
+    latents = (latents / vae.config.scaling_factor) + vae.config.shift_factor
+
+    image = vae.decode(latents, return_dict=False)[0]
+    image = image_processor.postprocess(image, output_type="pil")
+    image[0].save("split_transformer.png")
+```
+
+By selectively loading and unloading the models you need at a given stage and sharding the largest models across multiple GPUs, it is possible to run inference with large models on consumer GPUs.

docs/source/en/using-diffusers/callback.md

@@ -171,14 +171,13 @@ def latents_to_rgb(latents):
     weights = (
         (60, -60, 25, -70),
         (60,  -5, 15, -50),
-        (60,  10, -5, -35)
+        (60,  10, -5, -35),
     )
 
     weights_tensor = torch.t(torch.tensor(weights, dtype=latents.dtype).to(latents.device))
     biases_tensor = torch.tensor((150, 140, 130), dtype=latents.dtype).to(latents.device)
     rgb_tensor = torch.einsum("...lxy,lr -> ...rxy", latents, weights_tensor) + biases_tensor.unsqueeze(-1).unsqueeze(-1)
-    image_array = rgb_tensor.clamp(0, 255)[0].byte().cpu().numpy()
-    image_array = image_array.transpose(1, 2, 0)
+    image_array = rgb_tensor.clamp(0, 255).byte().cpu().numpy().transpose(1, 2, 0)
 
     return Image.fromarray(image_array)
 ```
@@ -189,7 +188,7 @@ def latents_to_rgb(latents):
 def decode_tensors(pipe, step, timestep, callback_kwargs):
     latents = callback_kwargs["latents"]
 
-    image = latents_to_rgb(latents)
+    image = latents_to_rgb(latents[0])
     image.save(f"{step}.png")
 
     return callback_kwargs

examples/cogvideo/train_cogvideox_lora.py

@@ -38,10 +38,7 @@
 from diffusers.models.embeddings import get_3d_rotary_pos_embed
 from diffusers.optimization import get_scheduler
 from diffusers.pipelines.cogvideo.pipeline_cogvideox import get_resize_crop_region_for_grid
-from diffusers.training_utils import (
-    cast_training_params,
-    clear_objs_and_retain_memory,
-)
+from diffusers.training_utils import cast_training_params, free_memory
 from diffusers.utils import check_min_version, convert_unet_state_dict_to_peft, export_to_video, is_wandb_available
 from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
 from diffusers.utils.torch_utils import is_compiled_module
@@ -726,7 +723,8 @@ def log_validation(
                 }
             )
 
-    clear_objs_and_retain_memory([pipe])
+    del pipe
+    free_memory()
 
     return videos