Merge branch 'main' of https://github.com/huggingface/diffusers

README.md

@@ -345,7 +345,8 @@ Textual Inversion is a technique for capturing novel concepts from a small numbe
 
 ## Stable Diffusion Community Pipelines
 
-The release of Stable Diffusion as an open source model has fostered a lot of interesting ideas and experimentation. Our [Community Examples folder](https://github.com/huggingface/diffusers/tree/main/examples/community) contains many ideas worth exploring, like interpolating to create animated videos, using CLIP Guidance for additional prompt fidelity, term weighting, and much more! [Take a look](https://huggingface.co/docs/diffusers/using-diffusers/custom_pipeline_overview) and [contribute your own](https://huggingface.co/docs/diffusers/using-diffusers/contribute_pipeline).
+The release of Stable Diffusion as an open source model has fostered a lot of interesting ideas and experimentation. 
+Our [Community Examples folder](https://github.com/huggingface/diffusers/tree/main/examples/community) contains many ideas worth exploring, like interpolating to create animated videos, using CLIP Guidance for additional prompt fidelity, term weighting, and much more! [Take a look](https://huggingface.co/docs/diffusers/using-diffusers/custom_pipeline_overview) and [contribute your own](https://huggingface.co/docs/diffusers/using-diffusers/contribute_pipeline).
 
 ## Other Examples
 
@@ -394,10 +395,14 @@ image.save("ddpm_generated_image.png")
 - [Unconditional Latent Diffusion](https://huggingface.co/CompVis/ldm-celebahq-256)
 - [Unconditional Diffusion with continuous scheduler](https://huggingface.co/google/ncsnpp-ffhq-1024)
 
-**Other Notebooks**:
+**Other Image Notebooks**:
 * [image-to-image generation with Stable Diffusion](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/image_2_image_using_diffusers.ipynb) ![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg),
 * [tweak images via repeated Stable Diffusion seeds](https://colab.research.google.com/github/pcuenca/diffusers-examples/blob/main/notebooks/stable-diffusion-seeds.ipynb) ![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg),
 
+**Diffusers for Other Modalities**:
+* [Molecule conformation generation](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/geodiff_molecule_conformation.ipynb) ![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg),
+* [Model-based reinforcement learning](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/reinforcement_learning_with_diffusers.ipynb) ![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg),
+
 ### Web Demos
 If you just want to play around with some web demos, you can try out the following 🚀 Spaces:
 | Model                          	| Hugging Face Spaces                                                                                                                                               	|

docs/source/_toctree.yml

@@ -31,6 +31,14 @@
     - local: using-diffusers/contribute_pipeline
       title: "How to contribute a Pipeline"
     title: "Pipelines for Inference"
+  - sections:
+    - local: using-diffusers/rl
+      title: "Reinforcement Learning"
+    - local: using-diffusers/audio
+      title: "Audio"
+    - local: using-diffusers/other-modalities
+      title: "Other Modalities"
+    title: "Taking Diffusers Beyond Images"
   title: "Using Diffusers"
 - sections:
   - local: optimization/fp16
@@ -107,4 +115,8 @@
     - local: api/pipelines/repaint
       title: "RePaint"
     title: "Pipelines"
+  - sections:
+    - local: api/experimental/rl
+      title: "RL Planning"
+    title: "Experimental Features"
   title: "API"

docs/source/api/experimental/rl.mdx

+<!--Copyright 2022 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.
+-->
+
+# TODO
+
+Coming soon!
\ No newline at end of file

docs/source/using-diffusers/audio.mdx

+<!--Copyright 2022 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.
+-->
+
+# Using Diffusers for audio
+
+The [`DanceDiffusionPipeline`] can be used to generate audio rapidly!
+More coming soon!
\ No newline at end of file

docs/source/using-diffusers/other-modalities.mdx

+<!--Copyright 2022 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.
+-->
+
+# Using Diffusers with other modalities
+
+Diffusers is in the process of expanding to modalities other than images.
+
+Currently, one example is for [molecule conformation](https://www.nature.com/subjects/molecular-conformation#:~:text=Definition,to%20changes%20in%20their%20environment.) generation.
+* Generate conformations in Colab [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/geodiff_molecule_conformation.ipynb)
+
+More coming soon!
\ No newline at end of file

docs/source/using-diffusers/rl.mdx

+<!--Copyright 2022 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.
+-->
+
+# Using Diffusers for reinforcement learning
+
+Support for one RL model and related pipelines is included in the `experimental` source of diffusers.
+
+To try some of this in colab, please look at the following example:
+* Model-based reinforcement learning on Colab [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/reinforcement_learning_with_diffusers.ipynb) ![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)

examples/community/README.md

@@ -370,7 +370,7 @@ def dummy(images, **kwargs):
 pipe.safety_checker = dummy
 
 images = []
-generator = th.Generator("cuda").manual_seed(0)
+generator = torch.Generator("cuda").manual_seed(0)
 
 seed = 0
 prompt = "a forest | a camel"
@@ -399,6 +399,7 @@ import requests
 from PIL import Image
 from io import BytesIO
 import torch
+import os
 from diffusers import DiffusionPipeline, DDIMScheduler
 has_cuda = torch.cuda.is_available()
 device = torch.device('cpu' if not has_cuda else 'cuda')
@@ -423,6 +424,7 @@ res = pipe.train(
     num_inference_steps=50,
     generator=generator)
 res = pipe(alpha=1)
+os.makedirs("imagic", exist_ok=True)
 image = res.images[0]
 image.save('./imagic/imagic_image_alpha_1.png')
 res = pipe(alpha=1.5)
@@ -652,4 +654,4 @@ prompt = "a cup"  # the masked out region will be replaced with this
 
 with autocast("cuda"):
     image = pipe(image=image, text=text, prompt=prompt).images[0]
-```
\ No newline at end of file
+```

examples/dreambooth/train_dreambooth.py

@@ -472,7 +472,7 @@ def main(args):
         eps=args.adam_epsilon,
     )
 
-    noise_scheduler = DDPMScheduler.from_config("CompVis/stable-diffusion-v1-4", subfolder="scheduler")
+    noise_scheduler = DDPMScheduler.from_config(args.pretrained_model_name_or_path, subfolder="scheduler")
 
     train_dataset = DreamBoothDataset(
         instance_data_root=args.instance_data_dir,

examples/text_to_image/train_text_to_image.py

@@ -372,7 +372,7 @@ def main():
         weight_decay=args.adam_weight_decay,
         eps=args.adam_epsilon,
     )
-    noise_scheduler = DDPMScheduler.from_config("CompVis/stable-diffusion-v1-4", subfolder="scheduler")
+    noise_scheduler = DDPMScheduler.from_config(args.pretrained_model_name_or_path, subfolder="scheduler")
 
     # Get the datasets: you can either provide your own training and evaluation files (see below)
     # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
@@ -605,7 +605,7 @@ def main():
             vae=vae,
             unet=unet,
             tokenizer=tokenizer,
-            scheduler=PNDMScheduler.from_config("CompVis/stable-diffusion-v1-4", subfolder="scheduler"),
+            scheduler=PNDMScheduler.from_config(args.pretrained_model_name_or_path, subfolder="scheduler"),
             safety_checker=StableDiffusionSafetyChecker.from_pretrained("CompVis/stable-diffusion-safety-checker"),
             feature_extractor=CLIPFeatureExtractor.from_pretrained("openai/clip-vit-base-patch32"),
         )

examples/textual_inversion/textual_inversion.py

@@ -441,7 +441,7 @@ def main():
         eps=args.adam_epsilon,
     )
 
-    noise_scheduler = DDPMScheduler.from_config("CompVis/stable-diffusion-v1-4", subfolder="scheduler")
+    noise_scheduler = DDPMScheduler.from_config(args.pretrained_model_name_or_path, subfolder="scheduler")
 
     train_dataset = TextualInversionDataset(
         data_root=args.train_data_dir,
@@ -574,7 +574,7 @@ def main():
             vae=vae,
             unet=unet,
             tokenizer=tokenizer,
-            scheduler=PNDMScheduler.from_config("CompVis/stable-diffusion-v1-4", subfolder="scheduler"),
+            scheduler=PNDMScheduler.from_config(args.pretrained_model_name_or_path, subfolder="scheduler"),
             safety_checker=StableDiffusionSafetyChecker.from_pretrained("CompVis/stable-diffusion-safety-checker"),
             feature_extractor=CLIPFeatureExtractor.from_pretrained("openai/clip-vit-base-patch32"),
         )

src/diffusers/__init__.py

@@ -82,6 +82,7 @@ if is_torch_available() and is_transformers_available() and is_onnx_available():
     from .pipelines import (
         OnnxStableDiffusionImg2ImgPipeline,
         OnnxStableDiffusionInpaintPipeline,
+        OnnxStableDiffusionInpaintPipelineLegacy,
         OnnxStableDiffusionPipeline,
         StableDiffusionOnnxPipeline,
     )

src/diffusers/pipelines/__init__.py

@@ -30,6 +30,7 @@ if is_transformers_available() and is_onnx_available():
     from .stable_diffusion import (
         OnnxStableDiffusionImg2ImgPipeline,
         OnnxStableDiffusionInpaintPipeline,
+        OnnxStableDiffusionInpaintPipelineLegacy,
         OnnxStableDiffusionPipeline,
         StableDiffusionOnnxPipeline,
     )

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -81,7 +81,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.__init__
     def __init__(
         self,
         vae: AutoencoderKL,
@@ -148,7 +147,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
             feature_extractor=feature_extractor,
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.enable_attention_slicing
     def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
         r"""
         Enable sliced attention computation.
@@ -168,7 +166,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
             slice_size = self.unet.config.attention_head_dim // 2
         self.unet.set_attention_slice(slice_size)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.disable_attention_slicing
     def disable_attention_slicing(self):
         r"""
         Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
@@ -177,7 +174,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
         # set slice_size = `None` to disable `attention slicing`
         self.enable_attention_slicing(None)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.enable_sequential_cpu_offload
     def enable_sequential_cpu_offload(self, gpu_id=0):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
@@ -196,7 +192,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
                 cpu_offload(cpu_offloaded_model, device)
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline._execution_device
     def _execution_device(self):
         r"""
         Returns the device on which the pipeline's models will be executed. After calling
@@ -214,7 +209,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
                 return torch.device(module._hf_hook.execution_device)
         return self.device
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.enable_xformers_memory_efficient_attention
     def enable_xformers_memory_efficient_attention(self):
         r"""
         Enable memory efficient attention as implemented in xformers.
@@ -227,14 +221,12 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
         """
         self.unet.set_use_memory_efficient_attention_xformers(True)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.disable_xformers_memory_efficient_attention
     def disable_xformers_memory_efficient_attention(self):
         r"""
         Disable memory efficient attention as implemented in xformers.
         """
         self.unet.set_use_memory_efficient_attention_xformers(False)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline._encode_prompt
     def _encode_prompt(self, prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt):
         r"""
         Encodes the prompt into text encoder hidden states.
@@ -340,7 +332,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
 
         return text_embeddings
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is not None:
             safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(device)
@@ -351,7 +342,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
             has_nsfw_concept = None
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.decode_latents
     def decode_latents(self, latents):
         latents = 1 / 0.18215 * latents
         image = self.vae.decode(latents).sample
@@ -360,7 +350,6 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.AltDiffusionPipeline.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/stable_diffusion/__init__.py

@@ -39,6 +39,7 @@ if is_transformers_available() and is_onnx_available():
     from .pipeline_onnx_stable_diffusion import OnnxStableDiffusionPipeline, StableDiffusionOnnxPipeline
     from .pipeline_onnx_stable_diffusion_img2img import OnnxStableDiffusionImg2ImgPipeline
     from .pipeline_onnx_stable_diffusion_inpaint import OnnxStableDiffusionInpaintPipeline
+    from .pipeline_onnx_stable_diffusion_inpaint_legacy import OnnxStableDiffusionInpaintPipelineLegacy
 
 if is_transformers_available() and is_flax_available():
     import flax

src/diffusers/pipelines/stable_diffusion/pipeline_flax_stable_diffusion.py

@@ -165,6 +165,7 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
         guidance_scale: float = 7.5,
         latents: Optional[jnp.array] = None,
         debug: bool = False,
+        neg_prompt_ids: jnp.array = None,
     ):
         if height % 8 != 0 or width % 8 != 0:
             raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")
@@ -177,10 +178,14 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
         batch_size = prompt_ids.shape[0]
 
         max_length = prompt_ids.shape[-1]
-        uncond_input = self.tokenizer(
-            [""] * batch_size, padding="max_length", max_length=max_length, return_tensors="np"
-        )
-        uncond_embeddings = self.text_encoder(uncond_input.input_ids, params=params["text_encoder"])[0]
+
+        if neg_prompt_ids is None:
+            uncond_input = self.tokenizer(
+                [""] * batch_size, padding="max_length", max_length=max_length, return_tensors="np"
+            ).input_ids
+        else:
+            uncond_input = neg_prompt_ids
+        uncond_embeddings = self.text_encoder(uncond_input, params=params["text_encoder"])[0]
         context = jnp.concatenate([uncond_embeddings, text_embeddings])
 
         latents_shape = (batch_size, self.unet.in_channels, height // 8, width // 8)
@@ -251,6 +256,7 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
         return_dict: bool = True,
         jit: bool = False,
         debug: bool = False,
+        neg_prompt_ids: jnp.array = None,
         **kwargs,
     ):
         r"""
@@ -298,11 +304,30 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
         """
         if jit:
             images = _p_generate(
-                self, prompt_ids, params, prng_seed, num_inference_steps, height, width, guidance_scale, latents, debug
+                self,
+                prompt_ids,
+                params,
+                prng_seed,
+                num_inference_steps,
+                height,
+                width,
+                guidance_scale,
+                latents,
+                debug,
+                neg_prompt_ids,
             )
         else:
             images = self._generate(
-                prompt_ids, params, prng_seed, num_inference_steps, height, width, guidance_scale, latents, debug
+                prompt_ids,
+                params,
+                prng_seed,
+                num_inference_steps,
+                height,
+                width,
+                guidance_scale,
+                latents,
+                debug,
+                neg_prompt_ids,
             )
 
         if self.safety_checker is not None:
@@ -333,10 +358,29 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
 # TODO: maybe use a config dict instead of so many static argnums
 @partial(jax.pmap, static_broadcasted_argnums=(0, 4, 5, 6, 7, 9))
 def _p_generate(
-    pipe, prompt_ids, params, prng_seed, num_inference_steps, height, width, guidance_scale, latents, debug
+    pipe,
+    prompt_ids,
+    params,
+    prng_seed,
+    num_inference_steps,
+    height,
+    width,
+    guidance_scale,
+    latents,
+    debug,
+    neg_prompt_ids,
 ):
     return pipe._generate(
-        prompt_ids, params, prng_seed, num_inference_steps, height, width, guidance_scale, latents, debug
+        prompt_ids,
+        params,
+        prng_seed,
+        num_inference_steps,
+        height,
+        width,
+        guidance_scale,
+        latents,
+        debug,
+        neg_prompt_ids,
     )
 
 

src/diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_inpaint.py

@@ -408,9 +408,9 @@ class OnnxStableDiffusionInpaintPipeline(DiffusionPipeline):
             # expand the latents if we are doing classifier free guidance
             latent_model_input = np.concatenate([latents] * 2) if do_classifier_free_guidance else latents
             # concat latents, mask, masked_image_latnets in the channel dimension
-            latent_model_input = np.concatenate([latent_model_input, mask, masked_image_latents], axis=1)
             latent_model_input = self.scheduler.scale_model_input(torch.from_numpy(latent_model_input), t)
             latent_model_input = latent_model_input.cpu().numpy()
+            latent_model_input = np.concatenate([latent_model_input, mask, masked_image_latents], axis=1)
 
             # predict the noise residual
             timestep = np.array([t], dtype=timestep_dtype)

src/diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_inpaint_legacy.py

+import inspect
+from typing import Callable, List, Optional, Union
+
+import numpy as np
+import torch
+
+import PIL
+from transformers import CLIPFeatureExtractor, CLIPTokenizer
+
+from ...configuration_utils import FrozenDict
+from ...onnx_utils import OnnxRuntimeModel
+from ...pipeline_utils import DiffusionPipeline
+from ...schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
+from ...utils import deprecate, logging
+from . import StableDiffusionPipelineOutput
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+def preprocess(image):
+    w, h = image.size
+    w, h = map(lambda x: x - x % 32, (w, h))  # resize to integer multiple of 32
+    image = image.resize((w, h), resample=PIL.Image.LANCZOS)
+    image = np.array(image).astype(np.float32) / 255.0
+    image = image[None].transpose(0, 3, 1, 2)
+    return 2.0 * image - 1.0
+
+
+def preprocess_mask(mask):
+    mask = mask.convert("L")
+    w, h = mask.size
+    w, h = map(lambda x: x - x % 32, (w, h))  # resize to integer multiple of 32
+    mask = mask.resize((w // 8, h // 8), resample=PIL.Image.NEAREST)
+    mask = np.array(mask).astype(np.float32) / 255.0
+    mask = np.tile(mask, (4, 1, 1))
+    mask = mask[None].transpose(0, 1, 2, 3)  # what does this step do?
+    mask = 1 - mask  # repaint white, keep black
+    return mask
+
+
+class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
+    r"""
+    Pipeline for text-guided image inpainting using Stable Diffusion. This is a *legacy feature* for Onnx pipelines to
+    provide compatibility with StableDiffusionInpaintPipelineLegacy and may be removed in the future.
+
+    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 latents. Can be one of
+            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
+        safety_checker ([`StableDiffusionSafetyChecker`]):
+            Classification module that estimates whether generated images could be considered offensive or harmful.
+            Please, refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for details.
+        feature_extractor ([`CLIPFeatureExtractor`]):
+            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
+    """
+    vae_encoder: OnnxRuntimeModel
+    vae_decoder: OnnxRuntimeModel
+    text_encoder: OnnxRuntimeModel
+    tokenizer: CLIPTokenizer
+    unet: OnnxRuntimeModel
+    scheduler: Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler]
+    safety_checker: OnnxRuntimeModel
+    feature_extractor: CLIPFeatureExtractor
+
+    def __init__(
+        self,
+        vae_encoder: OnnxRuntimeModel,
+        vae_decoder: OnnxRuntimeModel,
+        text_encoder: OnnxRuntimeModel,
+        tokenizer: CLIPTokenizer,
+        unet: OnnxRuntimeModel,
+        scheduler: Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler],
+        safety_checker: OnnxRuntimeModel,
+        feature_extractor: CLIPFeatureExtractor,
+    ):
+        super().__init__()
+
+        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+            deprecation_message = (
+                f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
+                f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
+                "to update the config accordingly as leaving `steps_offset` might led to incorrect results"
+                " in future versions. If you have downloaded this checkpoint from the Hugging Face Hub,"
+                " it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`"
+                " file"
+            )
+            deprecate("steps_offset!=1", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(scheduler.config)
+            new_config["steps_offset"] = 1
+            scheduler._internal_dict = FrozenDict(new_config)
+
+        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True:
+            deprecation_message = (
+                f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
+                " `clip_sample` should be set to False in the configuration file. Please make sure to update the"
+                " config accordingly as not setting `clip_sample` in the config might lead to incorrect results in"
+                " future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it would be very"
+                " nice if you could open a Pull request for the `scheduler/scheduler_config.json` file"
+            )
+            deprecate("clip_sample not set", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(scheduler.config)
+            new_config["clip_sample"] = False
+            scheduler._internal_dict = FrozenDict(new_config)
+
+        if safety_checker is None:
+            logger.warning(
+                f"You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure"
+                " that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered"
+                " results in services or applications open to the public. Both the diffusers team and Hugging Face"
+                " strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling"
+                " it only for use-cases that involve analyzing network behavior or auditing its results. For more"
+                " information, please have a look at https://github.com/huggingface/diffusers/pull/254 ."
+            )
+
+        self.register_modules(
+            vae_encoder=vae_encoder,
+            vae_decoder=vae_decoder,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            unet=unet,
+            scheduler=scheduler,
+            safety_checker=safety_checker,
+            feature_extractor=feature_extractor,
+        )
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_onnx_stable_diffusion.OnnxStableDiffusionPipeline._encode_prompt
+    def _encode_prompt(self, prompt, num_images_per_prompt, do_classifier_free_guidance, negative_prompt):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+            prompt (`str` or `list(int)`):
+                prompt to be encoded
+            num_images_per_prompt (`int`):
+                number of images that should be generated per prompt
+            do_classifier_free_guidance (`bool`):
+                whether to use classifier free guidance or not
+            negative_prompt (`str` or `List[str]`):
+                The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
+                if `guidance_scale` is less than `1`).
+        """
+        batch_size = len(prompt) if isinstance(prompt, list) else 1
+
+        # get prompt text embeddings
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=self.tokenizer.model_max_length,
+            truncation=True,
+            return_tensors="np",
+        )
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer(prompt, padding="max_length", return_tensors="np").input_ids
+
+        if not np.array_equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1])
+            logger.warning(
+                "The following part of your input was truncated because CLIP can only handle sequences up to"
+                f" {self.tokenizer.model_max_length} tokens: {removed_text}"
+            )
+
+        text_embeddings = self.text_encoder(input_ids=text_input_ids.astype(np.int32))[0]
+        text_embeddings = np.repeat(text_embeddings, num_images_per_prompt, axis=0)
+
+        # get unconditional embeddings for classifier free guidance
+        if do_classifier_free_guidance:
+            uncond_tokens: List[str]
+            if negative_prompt is None:
+                uncond_tokens = [""] * batch_size
+            elif type(prompt) is not type(negative_prompt):
+                raise TypeError(
+                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
+                    f" {type(prompt)}."
+                )
+            elif isinstance(negative_prompt, str):
+                uncond_tokens = [negative_prompt] * batch_size
+            elif batch_size != len(negative_prompt):
+                raise ValueError(
+                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
+                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
+                    " the batch size of `prompt`."
+                )
+            else:
+                uncond_tokens = negative_prompt
+
+            max_length = text_input_ids.shape[-1]
+            uncond_input = self.tokenizer(
+                uncond_tokens,
+                padding="max_length",
+                max_length=max_length,
+                truncation=True,
+                return_tensors="np",
+            )
+            uncond_embeddings = self.text_encoder(input_ids=uncond_input.input_ids.astype(np.int32))[0]
+            uncond_embeddings = np.repeat(uncond_embeddings, num_images_per_prompt, axis=0)
+
+            # For classifier free guidance, we need to do two forward passes.
+            # Here we concatenate the unconditional and text embeddings into a single batch
+            # to avoid doing two forward passes
+            text_embeddings = np.concatenate([uncond_embeddings, text_embeddings])
+
+        return text_embeddings
+
+    def __call__(
+        self,
+        prompt: Union[str, List[str]],
+        init_image: Union[np.ndarray, PIL.Image.Image],
+        mask_image: Union[np.ndarray, PIL.Image.Image],
+        strength: float = 0.8,
+        num_inference_steps: Optional[int] = 50,
+        guidance_scale: Optional[float] = 7.5,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        num_images_per_prompt: Optional[int] = 1,
+        eta: Optional[float] = 0.0,
+        generator: Optional[np.random.RandomState] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
+        callback_steps: Optional[int] = 1,
+        **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.
+            init_image (`nd.ndarray` 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 (`nd.ndarray` 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. If `mask_image` is a
+                PIL image, it will be converted to a single channel (luminance) before use. If it's a tensor, it should
+                contain one color channel (L) instead of 3, so the expected shape would be `(B, H, W, 1)`.uu
+            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.
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
+                if `guidance_scale` is less than `1`).
+            num_images_per_prompt (`int`, *optional*, defaults to 1):
+                The number of images to generate per prompt.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (?) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
+            generator (`np.random.RandomState`, *optional*):
+                A np.random.RandomState 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 `np.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
+                plain tuple.
+            callback (`Callable`, *optional*):
+                A function that will be called every `callback_steps` steps during inference. The function will be
+                called with the following arguments: `callback(step: int, timestep: int, latents: np.ndarray)`.
+            callback_steps (`int`, *optional*, defaults to 1):
+                The frequency at which the `callback` function will be called. If not specified, the callback will be
+                called at every step.
+
+        Returns:
+            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
+            [`~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):
+            batch_size = len(prompt)
+        else:
+            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
+
+        if strength < 0 or strength > 1:
+            raise ValueError(f"The value of strength should in [0.0, 1.0] but is {strength}")
+
+        if (callback_steps is None) or (
+            callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0)
+        ):
+            raise ValueError(
+                f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
+                f" {type(callback_steps)}."
+            )
+
+        if generator is None:
+            generator = np.random
+
+        # set timesteps
+        self.scheduler.set_timesteps(num_inference_steps)
+
+        if isinstance(init_image, PIL.Image.Image):
+            init_image = preprocess(init_image)
+
+        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
+        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
+        # corresponds to doing no classifier free guidance.
+        do_classifier_free_guidance = guidance_scale > 1.0
+
+        text_embeddings = self._encode_prompt(
+            prompt, num_images_per_prompt, do_classifier_free_guidance, negative_prompt
+        )
+
+        latents_dtype = text_embeddings.dtype
+        init_image = init_image.astype(latents_dtype)
+
+        # encode the init image into latents and scale the latents
+        init_latents = self.vae_encoder(sample=init_image)[0]
+        init_latents = 0.18215 * init_latents
+
+        # Expand init_latents for batch_size and num_images_per_prompt
+        init_latents = np.concatenate([init_latents] * num_images_per_prompt, axis=0)
+        init_latents_orig = init_latents
+
+        # preprocess mask
+        if not isinstance(mask_image, np.ndarray):
+            mask_image = preprocess_mask(mask_image)
+        mask_image = mask_image.astype(latents_dtype)
+        mask = np.concatenate([mask_image] * num_images_per_prompt, axis=0)
+
+        # check sizes
+        if not mask.shape == init_latents.shape:
+            raise ValueError("The mask and init_image should be the same size!")
+
+        # get the original timestep using init_timestep
+        offset = self.scheduler.config.get("steps_offset", 0)
+        init_timestep = int(num_inference_steps * strength) + offset
+        init_timestep = min(init_timestep, num_inference_steps)
+
+        timesteps = self.scheduler.timesteps.numpy()[-init_timestep]
+        timesteps = np.array([timesteps] * batch_size * num_images_per_prompt)
+
+        # add noise to latents using the timesteps
+        noise = generator.randn(*init_latents.shape).astype(latents_dtype)
+        init_latents = self.scheduler.add_noise(
+            torch.from_numpy(init_latents), torch.from_numpy(noise), torch.from_numpy(timesteps)
+        )
+        init_latents = init_latents.numpy()
+
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (?) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to ? in DDIM paper: https://arxiv.org/abs/2010.02502
+        # and should be between [0, 1]
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        latents = init_latents
+
+        t_start = max(num_inference_steps - init_timestep + offset, 0)
+        timesteps = self.scheduler.timesteps[t_start:].numpy()
+
+        for i, t in enumerate(self.progress_bar(timesteps)):
+            # expand the latents if we are doing classifier free guidance
+            latent_model_input = np.concatenate([latents] * 2) if do_classifier_free_guidance else latents
+            latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
+
+            # predict the noise residual
+            noise_pred = self.unet(
+                sample=latent_model_input, timestep=np.array([t]), encoder_hidden_states=text_embeddings
+            )[0]
+
+            # perform guidance
+            if do_classifier_free_guidance:
+                noise_pred_uncond, noise_pred_text = np.split(noise_pred, 2)
+                noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
+
+            # compute the previous noisy sample x_t -> x_t-1
+            latents = self.scheduler.step(
+                torch.from_numpy(noise_pred), t, torch.from_numpy(latents), **extra_step_kwargs
+            ).prev_sample
+
+            latents = latents.numpy()
+
+            init_latents_proper = self.scheduler.add_noise(
+                torch.from_numpy(init_latents_orig), torch.from_numpy(noise), torch.from_numpy(np.array([t]))
+            )
+
+            init_latents_proper = init_latents_proper.numpy()
+
+            latents = (init_latents_proper * mask) + (latents * (1 - mask))
+
+            # call the callback, if provided
+            if callback is not None and i % callback_steps == 0:
+                callback(i, t, latents)
+
+        latents = 1 / 0.18215 * latents
+        # image = self.vae_decoder(latent_sample=latents)[0]
+        # it seems likes there is a strange result for using half-precision vae decoder if batchsize>1
+        image = np.concatenate(
+            [self.vae_decoder(latent_sample=latents[i : i + 1])[0] for i in range(latents.shape[0])]
+        )
+
+        image = np.clip(image / 2 + 0.5, 0, 1)
+        image = image.transpose((0, 2, 3, 1))
+
+        if self.safety_checker is not None:
+            safety_checker_input = self.feature_extractor(
+                self.numpy_to_pil(image), return_tensors="np"
+            ).pixel_values.astype(image.dtype)
+            # There will throw an error if use safety_checker batchsize>1
+            images, has_nsfw_concept = [], []
+            for i in range(image.shape[0]):
+                image_i, has_nsfw_concept_i = self.safety_checker(
+                    clip_input=safety_checker_input[i : i + 1], images=image[i : i + 1]
+                )
+                images.append(image_i)
+                has_nsfw_concept.append(has_nsfw_concept_i[0])
+            image = np.concatenate(images)
+        else:
+            has_nsfw_concept = None
+
+        if output_type == "pil":
+            image = self.numpy_to_pil(image)
+
+        if not return_dict:
+            return (image, has_nsfw_concept)
+
+        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -35,16 +35,93 @@ logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
 def prepare_mask_and_masked_image(image, mask):
-    image = np.array(image.convert("RGB"))
-    image = image[None].transpose(0, 3, 1, 2)
-    image = torch.from_numpy(image).to(dtype=torch.float32) / 127.5 - 1.0
+    """
+    Prepares a pair (image, mask) to be consumed by the Stable Diffusion pipeline.
+    This means that those inputs will be converted to ``torch.Tensor`` with
+    shapes ``batch x channels x height x width`` where ``channels`` is ``3`` for
+    the ``image`` and ``1`` for the ``mask``.
+
+    The ``image`` will be converted to ``torch.float32`` and normalized to be in
+    ``[-1, 1]``. The ``mask`` will be binarized (``mask > 0.5``) and cast to
+    ``torch.float32`` too.
 
-    mask = np.array(mask.convert("L"))
-    mask = mask.astype(np.float32) / 255.0
-    mask = mask[None, None]
-    mask[mask < 0.5] = 0
-    mask[mask >= 0.5] = 1
-    mask = torch.from_numpy(mask)
+    Args:
+        image (Union[np.array, PIL.Image, torch.Tensor]): The image to inpaint.
+            It can be a ``PIL.Image``, or a ``height x width x 3`` ``np.array``
+            or a ``channels x height x width`` ``torch.Tensor`` or a
+            ``batch x channels x height x width`` ``torch.Tensor``.
+        mask (_type_): The mask to apply to the image, i.e. regions to inpaint.
+            It can be a ``PIL.Image``, or a ``height x width`` ``np.array`` or
+            a ``1 x height x width`` ``torch.Tensor`` or a
+            ``batch x 1 x height x width`` ``torch.Tensor``.
+
+
+    Raises:
+        ValueError: ``torch.Tensor`` images should be in the ``[-1, 1]`` range.
+        ValueError: ``torch.Tensor`` mask should be in the ``[0, 1]`` range.
+        ValueError: ``mask`` and ``image`` should have the same spatial dimensions.
+        TypeError: ``mask`` is a ``torch.Tensor`` but ``image`` is not
+            (ot the other way around).
+
+    Returns:
+        tuple[torch.Tensor]: The pair (mask, masked_image) as ``torch.Tensor`` with 4
+            dimensions: ``batch x channels x height x width``.
+    """
+    if isinstance(image, torch.Tensor):
+        if not isinstance(mask, torch.Tensor):
+            raise TypeError(f"`image` is a torch.Tensor but `mask` (type: {type(mask)} is not")
+
+        # Batch single image
+        if image.ndim == 3:
+            assert image.shape[0] == 3, "Image outside a batch should be of shape (3, H, W)"
+            image = image.unsqueeze(0)
+
+        # Batch and add channel dim for single mask
+        if mask.ndim == 2:
+            mask = mask.unsqueeze(0).unsqueeze(0)
+
+        # Batch single mask or add channel dim
+        if mask.ndim == 3:
+            # Single batched mask, no channel dim or single mask not batched but channel dim
+            if mask.shape[0] == 1:
+                mask = mask.unsqueeze(0)
+
+            # Batched masks no channel dim
+            else:
+                mask = mask.unsqueeze(1)
+
+        assert image.ndim == 4 and mask.ndim == 4, "Image and Mask must have 4 dimensions"
+        assert image.shape[-2:] == mask.shape[-2:], "Image and Mask must have the same spatial dimensions"
+        assert image.shape[0] == mask.shape[0], "Image and Mask must have the same batch size"
+
+        # Check image is in [-1, 1]
+        if image.min() < -1 or image.max() > 1:
+            raise ValueError("Image should be in [-1, 1] range")
+
+        # Check mask is in [0, 1]
+        if mask.min() < 0 or mask.max() > 1:
+            raise ValueError("Mask should be in [0, 1] range")
+
+        # Binarize mask
+        mask[mask < 0.5] = 0
+        mask[mask >= 0.5] = 1
+
+        # Image as float32
+        image = image.to(dtype=torch.float32)
+    elif isinstance(mask, torch.Tensor):
+        raise TypeError(f"`mask` is a torch.Tensor but `image` (type: {type(image)} is not")
+    else:
+        if isinstance(image, PIL.Image.Image):
+            image = np.array(image.convert("RGB"))
+        image = image[None].transpose(0, 3, 1, 2)
+        image = torch.from_numpy(image).to(dtype=torch.float32) / 127.5 - 1.0
+        if isinstance(mask, PIL.Image.Image):
+            mask = np.array(mask.convert("L"))
+            mask = mask.astype(np.float32) / 255.0
+        mask = mask[None, None]
+        mask[mask < 0.5] = 0
+        mask[mask >= 0.5] = 1
+        mask = torch.from_numpy(mask)
 
     masked_image = image * (mask < 0.5)
 
@@ -586,9 +663,8 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
             latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
 
             # concat latents, mask, masked_image_latents in the channel dimension
-            latent_model_input = torch.cat([latent_model_input, mask, masked_image_latents], dim=1)
-
             latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
+            latent_model_input = torch.cat([latent_model_input, mask, masked_image_latents], dim=1)
 
             # predict the noise residual
             noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=text_embeddings).sample

src/diffusers/schedulers/scheduling_lms_discrete.py

@@ -243,19 +243,18 @@ class LMSDiscreteScheduler(SchedulerMixin, ConfigMixin):
         timesteps: torch.FloatTensor,
     ) -> torch.FloatTensor:
         # Make sure sigmas and timesteps have the same device and dtype as original_samples
-        self.sigmas = self.sigmas.to(device=original_samples.device, dtype=original_samples.dtype)
+        sigmas = self.sigmas.to(device=original_samples.device, dtype=original_samples.dtype)
         if original_samples.device.type == "mps" and torch.is_floating_point(timesteps):
             # mps does not support float64
-            self.timesteps = self.timesteps.to(original_samples.device, dtype=torch.float32)
+            schedule_timesteps = self.timesteps.to(original_samples.device, dtype=torch.float32)
             timesteps = timesteps.to(original_samples.device, dtype=torch.float32)
         else:
-            self.timesteps = self.timesteps.to(original_samples.device)
+            schedule_timesteps = self.timesteps.to(original_samples.device)
             timesteps = timesteps.to(original_samples.device)
 
-        schedule_timesteps = self.timesteps
         step_indices = [(schedule_timesteps == t).nonzero().item() for t in timesteps]
 
-        sigma = self.sigmas[step_indices].flatten()
+        sigma = sigmas[step_indices].flatten()
         while len(sigma.shape) < len(original_samples.shape):
             sigma = sigma.unsqueeze(-1)
 

src/diffusers/schedulers/scheduling_utils_flax.py

@@ -118,7 +118,10 @@ class FlaxSchedulerMixin:
 
         """
         config, kwargs = cls.load_config(
-            pretrained_model_name_or_path=pretrained_model_name_or_path, return_unused_kwargs=True, **kwargs
+            pretrained_model_name_or_path=pretrained_model_name_or_path,
+            subfolder=subfolder,
+            return_unused_kwargs=True,
+            **kwargs,
         )
         scheduler, unused_kwargs = cls.from_config(config, return_unused_kwargs=True, **kwargs)