add OnnxStableDiffusionUpscalePipeline pipeline (#2158)

* [Onnx] add Stable Diffusion Upscale pipeline

* add a test for the OnnxStableDiffusionUpscalePipeline

* check for VAE config before adjusting scaling factor

* update test assertions, lint fixes

* run fix-copies target

* switch test checkpoint to one hosted on huggingface

* partially restore attention mask

* reshape embeddings after running text encoder

* add longer nightly test for ONNX upscale pipeline

* use package import to fix tests

* fix scheduler compatibility and class labels dtype

* use more precise type

* remove LMS from fast tests

* lookup latent and timestamp types

* add docs for ONNX upscaling, rename lookup table

* replace deprecated pipeline names in ONNX docs

docs/source/en/optimization/onnx.mdx

@@ -21,13 +21,13 @@ specific language governing permissions and limitations under the License.
 
 ## Stable Diffusion Inference
 
-The snippet below demonstrates how to use the ONNX runtime. You need to use `StableDiffusionOnnxPipeline` instead of `StableDiffusionPipeline`. You also need to download the weights from the `onnx` branch of the repository, and indicate the runtime provider you want to use.
+The snippet below demonstrates how to use the ONNX runtime. You need to use `OnnxStableDiffusionPipeline` instead of `StableDiffusionPipeline`. You also need to download the weights from the `onnx` branch of the repository, and indicate the runtime provider you want to use.
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from diffusers import StableDiffusionOnnxPipeline
+from diffusers import OnnxStableDiffusionPipeline
 
-pipe = StableDiffusionOnnxPipeline.from_pretrained(
+pipe = OnnxStableDiffusionPipeline.from_pretrained(
     "runwayml/stable-diffusion-v1-5",
     revision="onnx",
     provider="CUDAExecutionProvider",
@@ -37,6 +37,37 @@ prompt = "a photo of an astronaut riding a horse on mars"
 image = pipe(prompt).images[0]
 ```
 
+The snippet below demonstrates how to use the ONNX runtime with the Stable Diffusion upscaling pipeline.
+
+```python
+from diffusers import OnnxStableDiffusionPipeline, OnnxStableDiffusionUpscalePipeline
+
+prompt = "a photo of an astronaut riding a horse on mars"
+steps = 50
+
+txt2img = OnnxStableDiffusionPipeline.from_pretrained(
+    "runwayml/stable-diffusion-v1-5",
+    revision="onnx",
+    provider="CUDAExecutionProvider",
+)
+small_image = txt2img(
+    prompt,
+    num_inference_steps=steps,
+).images[0]
+
+generator = torch.manual_seed(0)
+upscale = OnnxStableDiffusionUpscalePipeline.from_pretrained(
+    "ssube/stable-diffusion-x4-upscaler-onnx",
+    provider="CUDAExecutionProvider",
+)
+large_image = upscale(
+    prompt,
+    small_image,
+    generator=generator,
+    num_inference_steps=steps,
+).images[0]
+```
+
 ## Known Issues
 
 - Generating multiple prompts in a batch seems to take too much memory. While we look into it, you may need to iterate instead of batching.

src/diffusers/__init__.py

@@ -158,6 +158,7 @@ else:
         OnnxStableDiffusionInpaintPipeline,
         OnnxStableDiffusionInpaintPipelineLegacy,
         OnnxStableDiffusionPipeline,
+        OnnxStableDiffusionUpscalePipeline,
         StableDiffusionOnnxPipeline,
     )
 

src/diffusers/pipelines/__init__.py

@@ -93,6 +93,7 @@ else:
         OnnxStableDiffusionInpaintPipeline,
         OnnxStableDiffusionInpaintPipelineLegacy,
         OnnxStableDiffusionPipeline,
+        OnnxStableDiffusionUpscalePipeline,
         StableDiffusionOnnxPipeline,
     )
 

src/diffusers/pipelines/stable_diffusion/__init__.py

@@ -104,6 +104,7 @@ else:
     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
+    from .pipeline_onnx_stable_diffusion_upscale import OnnxStableDiffusionUpscalePipeline
 
 if is_transformers_available() and is_flax_available():
     import flax

src/diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_upscale.py

+from logging import getLogger
+from typing import Any, Callable, List, Optional, Union
+
+import numpy as np
+import PIL
+import torch
+
+from ...schedulers import DDPMScheduler
+from ..onnx_utils import ORT_TO_NP_TYPE, OnnxRuntimeModel
+from ..pipeline_utils import ImagePipelineOutput
+from . import StableDiffusionUpscalePipeline
+
+
+logger = getLogger(__name__)
+
+
+NUM_LATENT_CHANNELS = 4
+NUM_UNET_INPUT_CHANNELS = 7
+
+ORT_TO_PT_TYPE = {
+    "float16": torch.float16,
+    "float32": torch.float32,
+}
+
+
+def preprocess(image):
+    if isinstance(image, torch.Tensor):
+        return image
+    elif isinstance(image, PIL.Image.Image):
+        image = [image]
+
+    if isinstance(image[0], PIL.Image.Image):
+        w, h = image[0].size
+        w, h = map(lambda x: x - x % 64, (w, h))  # resize to integer multiple of 32
+
+        image = [np.array(i.resize((w, h)))[None, :] for i in image]
+        image = np.concatenate(image, axis=0)
+        image = np.array(image).astype(np.float32) / 255.0
+        image = image.transpose(0, 3, 1, 2)
+        image = 2.0 * image - 1.0
+        image = torch.from_numpy(image)
+    elif isinstance(image[0], torch.Tensor):
+        image = torch.cat(image, dim=0)
+
+    return image
+
+
+class OnnxStableDiffusionUpscalePipeline(StableDiffusionUpscalePipeline):
+    def __init__(
+        self,
+        vae: OnnxRuntimeModel,
+        text_encoder: OnnxRuntimeModel,
+        tokenizer: Any,
+        unet: OnnxRuntimeModel,
+        low_res_scheduler: DDPMScheduler,
+        scheduler: Any,
+        max_noise_level: int = 350,
+    ):
+        super().__init__(vae, text_encoder, tokenizer, unet, low_res_scheduler, scheduler, max_noise_level)
+
+    def __call__(
+        self,
+        prompt: Union[str, List[str]],
+        image: Union[torch.FloatTensor, PIL.Image.Image, List[PIL.Image.Image]],
+        num_inference_steps: int = 75,
+        guidance_scale: float = 9.0,
+        noise_level: int = 20,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        num_images_per_prompt: Optional[int] = 1,
+        eta: float = 0.0,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        latents: Optional[torch.FloatTensor] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback_steps: Optional[int] = 1,
+    ):
+        # 1. Check inputs
+        self.check_inputs(prompt, image, noise_level, callback_steps)
+
+        # 2. Define call parameters
+        batch_size = 1 if isinstance(prompt, str) else len(prompt)
+        device = self._execution_device
+        # 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
+
+        # 3. Encode input prompt
+        text_embeddings = self._encode_prompt(
+            prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt
+        )
+
+        latents_dtype = ORT_TO_PT_TYPE[str(text_embeddings.dtype)]
+
+        # 4. Preprocess image
+        image = preprocess(image)
+        image = image.cpu()
+
+        # 5. set timesteps
+        self.scheduler.set_timesteps(num_inference_steps, device=device)
+        timesteps = self.scheduler.timesteps
+
+        # 5. Add noise to image
+        noise_level = torch.tensor([noise_level], dtype=torch.long, device=device)
+        noise = torch.randn(image.shape, generator=generator, device=device, dtype=latents_dtype)
+        image = self.low_res_scheduler.add_noise(image, noise, noise_level)
+
+        batch_multiplier = 2 if do_classifier_free_guidance else 1
+        image = np.concatenate([image] * batch_multiplier * num_images_per_prompt)
+        noise_level = np.concatenate([noise_level] * image.shape[0])
+
+        # 6. Prepare latent variables
+        height, width = image.shape[2:]
+        latents = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            NUM_LATENT_CHANNELS,
+            height,
+            width,
+            latents_dtype,
+            device,
+            generator,
+            latents,
+        )
+
+        # 7. Check that sizes of image and latents match
+        num_channels_image = image.shape[1]
+        if NUM_LATENT_CHANNELS + num_channels_image != NUM_UNET_INPUT_CHANNELS:
+            raise ValueError(
+                "Incorrect configuration settings! The config of `pipeline.unet` expects"
+                f" {NUM_UNET_INPUT_CHANNELS} but received `num_channels_latents`: {NUM_LATENT_CHANNELS} +"
+                f" `num_channels_image`: {num_channels_image} "
+                f" = {NUM_LATENT_CHANNELS+num_channels_image}. Please verify the config of"
+                " `pipeline.unet` or your `image` input."
+            )
+
+        # 8. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
+        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
+
+        timestep_dtype = next(
+            (input.type for input in self.unet.model.get_inputs() if input.name == "timestep"), "tensor(float)"
+        )
+        timestep_dtype = ORT_TO_NP_TYPE[timestep_dtype]
+
+        # 9. Denoising loop
+        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(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
+
+                # concat latents, mask, masked_image_latents in the channel dimension
+                latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
+                latent_model_input = np.concatenate([latent_model_input, image], axis=1)
+
+                # timestep to tensor
+                timestep = np.array([t], dtype=timestep_dtype)
+
+                # predict the noise residual
+                noise_pred = self.unet(
+                    sample=latent_model_input,
+                    timestep=timestep,
+                    encoder_hidden_states=text_embeddings,
+                    class_labels=noise_level.astype(np.int64),
+                )[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, latents, **extra_step_kwargs
+                ).prev_sample
+
+                # call the callback, if provided
+                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
+                    progress_bar.update()
+                    if callback is not None and i % callback_steps == 0:
+                        callback(i, t, latents)
+
+        # 10. Post-processing
+        image = self.decode_latents(latents.float())
+
+        # 11. Convert to PIL
+        if output_type == "pil":
+            image = self.numpy_to_pil(image)
+
+        if not return_dict:
+            return (image,)
+
+        return ImagePipelineOutput(images=image)
+
+    def decode_latents(self, latents):
+        latents = 1 / 0.08333 * latents
+        image = self.vae(latent_sample=latents)[0]
+        image = np.clip(image / 2 + 0.5, 0, 1)
+        image = image.transpose((0, 2, 3, 1))
+        return image
+
+    def _encode_prompt(self, prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt):
+        batch_size = len(prompt) if isinstance(prompt, list) else 1
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=self.tokenizer.model_max_length,
+            truncation=True,
+            return_tensors="pt",
+        )
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.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}"
+            )
+
+        # if hasattr(text_inputs, "attention_mask"):
+        #     attention_mask = text_inputs.attention_mask.to(device)
+        # else:
+        #     attention_mask = None
+
+        # no positional arguments to text_encoder
+        text_embeddings = self.text_encoder(
+            input_ids=text_input_ids.int().to(device),
+            # attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
+
+        bs_embed, seq_len, _ = text_embeddings.shape
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        text_embeddings = text_embeddings.repeat(1, num_images_per_prompt)
+        text_embeddings = text_embeddings.reshape(bs_embed * num_images_per_prompt, seq_len, -1)
+
+        # 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]
+            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="pt",
+            )
+
+            # if hasattr(uncond_input, "attention_mask"):
+            #     attention_mask = uncond_input.attention_mask.to(device)
+            # else:
+            #     attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                input_ids=uncond_input.input_ids.int().to(device),
+                # attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
+
+            seq_len = uncond_embeddings.shape[1]
+            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
+            uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt)
+            uncond_embeddings = uncond_embeddings.reshape(batch_size * num_images_per_prompt, seq_len, -1)
+
+            # 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

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py

@@ -88,21 +88,22 @@ class StableDiffusionUpscalePipeline(DiffusionPipeline):
     ):
         super().__init__()
 
-        # check if vae has a config attribute `scaling_factor` and if it is set to 0.08333, else set it to 0.08333 and deprecate
-        is_vae_scaling_factor_set_to_0_08333 = (
-            hasattr(vae.config, "scaling_factor") and vae.config.scaling_factor == 0.08333
-        )
-        if not is_vae_scaling_factor_set_to_0_08333:
-            deprecation_message = (
-                "The configuration file of the vae does not contain `scaling_factor` or it is set to"
-                f" {vae.config.scaling_factor}, which seems highly unlikely. If your checkpoint is a fine-tuned"
-                " version of `stabilityai/stable-diffusion-x4-upscaler` you should change 'scaling_factor' to 0.08333"
-                " Please make sure to update the config accordingly, as not doing so 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 `vae/config.json` file"
+        if hasattr(vae, "config"):
+            # check if vae has a config attribute `scaling_factor` and if it is set to 0.08333, else set it to 0.08333 and deprecate
+            is_vae_scaling_factor_set_to_0_08333 = (
+                hasattr(vae.config, "scaling_factor") and vae.config.scaling_factor == 0.08333
             )
-            deprecate("wrong scaling_factor", "1.0.0", deprecation_message, standard_warn=False)
-            vae.register_to_config(scaling_factor=0.08333)
+            if not is_vae_scaling_factor_set_to_0_08333:
+                deprecation_message = (
+                    "The configuration file of the vae does not contain `scaling_factor` or it is set to"
+                    f" {vae.config.scaling_factor}, which seems highly unlikely. If your checkpoint is a fine-tuned"
+                    " version of `stabilityai/stable-diffusion-x4-upscaler` you should change 'scaling_factor' to"
+                    " 0.08333 Please make sure to update the config accordingly, as not doing so 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 `vae/config.json` file"
+                )
+                deprecate("wrong scaling_factor", "1.0.0", deprecation_message, standard_warn=False)
+                vae.register_to_config(scaling_factor=0.08333)
 
         self.register_modules(
             vae=vae,

src/diffusers/utils/dummy_torch_and_transformers_and_onnx_objects.py

@@ -62,6 +62,21 @@ class OnnxStableDiffusionPipeline(metaclass=DummyObject):
         requires_backends(cls, ["torch", "transformers", "onnx"])
 
 
+class OnnxStableDiffusionUpscalePipeline(metaclass=DummyObject):
+    _backends = ["torch", "transformers", "onnx"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch", "transformers", "onnx"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers", "onnx"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers", "onnx"])
+
+
 class StableDiffusionOnnxPipeline(metaclass=DummyObject):
     _backends = ["torch", "transformers", "onnx"]
 

tests/pipelines/stable_diffusion/test_onnx_stable_diffusion_upscale.py

+# coding=utf-8
+# Copyright 2022 HuggingFace Inc.
+#
+# 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.
+
+import random
+import unittest
+
+import numpy as np
+import torch
+
+from diffusers import (
+    DPMSolverMultistepScheduler,
+    EulerAncestralDiscreteScheduler,
+    EulerDiscreteScheduler,
+    LMSDiscreteScheduler,
+    OnnxStableDiffusionUpscalePipeline,
+    PNDMScheduler,
+)
+from diffusers.utils import floats_tensor
+from diffusers.utils.testing_utils import (
+    is_onnx_available,
+    load_image,
+    nightly,
+    require_onnxruntime,
+    require_torch_gpu,
+)
+
+from ...test_pipelines_onnx_common import OnnxPipelineTesterMixin
+
+
+if is_onnx_available():
+    import onnxruntime as ort
+
+
+class OnnxStableDiffusionUpscalePipelineFastTests(OnnxPipelineTesterMixin, unittest.TestCase):
+    # TODO: is there an appropriate internal test set?
+    hub_checkpoint = "ssube/stable-diffusion-x4-upscaler-onnx"
+
+    def get_dummy_inputs(self, seed=0):
+        image = floats_tensor((1, 3, 128, 128), rng=random.Random(seed))
+        generator = torch.manual_seed(seed)
+        inputs = {
+            "prompt": "A painting of a squirrel eating a burger",
+            "image": image,
+            "generator": generator,
+            "num_inference_steps": 3,
+            "guidance_scale": 7.5,
+            "output_type": "numpy",
+        }
+        return inputs
+
+    def test_pipeline_default_ddpm(self):
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(self.hub_checkpoint, provider="CPUExecutionProvider")
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs()
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1].flatten()
+
+        # started as 128, should now be 512
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array(
+            [0.6974782, 0.68902093, 0.70135885, 0.7583618, 0.7804545, 0.7854912, 0.78667426, 0.78743863, 0.78070223]
+        )
+        assert np.abs(image_slice - expected_slice).max() < 1e-1
+
+    def test_pipeline_pndm(self):
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(self.hub_checkpoint, provider="CPUExecutionProvider")
+        pipe.scheduler = PNDMScheduler.from_config(pipe.scheduler.config, skip_prk_steps=True)
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs()
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array(
+            [0.6898892, 0.59240556, 0.52499527, 0.58866215, 0.52258235, 0.52572715, 0.62414473, 0.6174387, 0.6214964]
+        )
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-1
+
+    def test_pipeline_dpm_multistep(self):
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(self.hub_checkpoint, provider="CPUExecutionProvider")
+        pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs()
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array(
+            [0.7659278, 0.76437664, 0.75579107, 0.7691116, 0.77666986, 0.7727672, 0.7758664, 0.7812226, 0.76942515]
+        )
+
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-1
+
+    def test_pipeline_euler(self):
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(self.hub_checkpoint, provider="CPUExecutionProvider")
+        pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config)
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs()
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array(
+            [0.6974782, 0.68902093, 0.70135885, 0.7583618, 0.7804545, 0.7854912, 0.78667426, 0.78743863, 0.78070223]
+        )
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-1
+
+    def test_pipeline_euler_ancestral(self):
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(self.hub_checkpoint, provider="CPUExecutionProvider")
+        pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs()
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array(
+            [0.77424496, 0.773601, 0.7645288, 0.7769598, 0.7772739, 0.7738688, 0.78187233, 0.77879584, 0.767043]
+        )
+
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-1
+
+
+@nightly
+@require_onnxruntime
+@require_torch_gpu
+class OnnxStableDiffusionUpscalePipelineIntegrationTests(unittest.TestCase):
+    @property
+    def gpu_provider(self):
+        return (
+            "CUDAExecutionProvider",
+            {
+                "gpu_mem_limit": "15000000000",  # 15GB
+                "arena_extend_strategy": "kSameAsRequested",
+            },
+        )
+
+    @property
+    def gpu_options(self):
+        options = ort.SessionOptions()
+        options.enable_mem_pattern = False
+        return options
+
+    def test_inference_default_ddpm(self):
+        init_image = load_image(
+            "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main"
+            "/img2img/sketch-mountains-input.jpg"
+        )
+        init_image = init_image.resize((128, 128))
+        # using the PNDM scheduler by default
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(
+            "ssube/stable-diffusion-x4-upscaler-onnx",
+            provider=self.gpu_provider,
+            sess_options=self.gpu_options,
+        )
+        pipe.set_progress_bar_config(disable=None)
+
+        prompt = "A fantasy landscape, trending on artstation"
+
+        generator = torch.manual_seed(0)
+        output = pipe(
+            prompt=prompt,
+            image=init_image,
+            guidance_scale=7.5,
+            num_inference_steps=10,
+            generator=generator,
+            output_type="np",
+        )
+        images = output.images
+        image_slice = images[0, 255:258, 383:386, -1]
+
+        assert images.shape == (1, 512, 512, 3)
+        expected_slice = np.array([0.4883, 0.4947, 0.4980, 0.4975, 0.4982, 0.4980, 0.5000, 0.5006, 0.4972])
+        # TODO: lower the tolerance after finding the cause of onnxruntime reproducibility issues
+
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 2e-2
+
+    def test_inference_k_lms(self):
+        init_image = load_image(
+            "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main"
+            "/img2img/sketch-mountains-input.jpg"
+        )
+        init_image = init_image.resize((128, 128))
+        lms_scheduler = LMSDiscreteScheduler.from_pretrained(
+            "ssube/stable-diffusion-x4-upscaler-onnx", subfolder="scheduler"
+        )
+        pipe = OnnxStableDiffusionUpscalePipeline.from_pretrained(
+            "ssube/stable-diffusion-x4-upscaler-onnx",
+            scheduler=lms_scheduler,
+            provider=self.gpu_provider,
+            sess_options=self.gpu_options,
+        )
+        pipe.set_progress_bar_config(disable=None)
+
+        prompt = "A fantasy landscape, trending on artstation"
+
+        generator = torch.manual_seed(0)
+        output = pipe(
+            prompt=prompt,
+            image=init_image,
+            guidance_scale=7.5,
+            num_inference_steps=20,
+            generator=generator,
+            output_type="np",
+        )
+        images = output.images
+        image_slice = images[0, 255:258, 383:386, -1]
+
+        assert images.shape == (1, 512, 512, 3)
+        expected_slice = np.array(
+            [0.50173753, 0.50223356, 0.502039, 0.50233036, 0.5023725, 0.5022601, 0.5018758, 0.50234085, 0.50241566]
+        )
+        # TODO: lower the tolerance after finding the cause of onnxruntime reproducibility issues
+
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 2e-2