Release: v0.27.2

* fix

* fix

* add a tests

* fix

---------

Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
Co-authored-by: yiyixuxu <yixu310@gmail,com>

setup.py

@@ -249,7 +249,7 @@ version_range_max = max(sys.version_info[1], 10) + 1
 
 setup(
     name="diffusers",
-    version="0.27.1",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
+    version="0.27.2",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
     description="State-of-the-art diffusion in PyTorch and JAX.",
     long_description=open("README.md", "r", encoding="utf-8").read(),
     long_description_content_type="text/markdown",

src/diffusers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.27.1"
+__version__ = "0.27.2"
 
 from typing import TYPE_CHECKING
 

src/diffusers/models/unets/unet_2d_condition.py

@@ -1178,6 +1178,7 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin,
         # we're popping the `scale` instead of getting it because otherwise `scale` will be propagated
         # to the internal blocks and will raise deprecation warnings. this will be confusing for our users.
         if cross_attention_kwargs is not None:
+            cross_attention_kwargs = cross_attention_kwargs.copy()
             lora_scale = cross_attention_kwargs.pop("scale", 1.0)
         else:
             lora_scale = 1.0

src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py

@@ -528,15 +528,12 @@ class StableDiffusionInpaintPipelineLegacy(
                     f" {negative_prompt_embeds.shape}."
                 )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.StableDiffusionImg2ImgPipeline.get_timesteps
     def get_timesteps(self, num_inference_steps, strength, device):
         # get the original timestep using init_timestep
         init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
 
         t_start = max(num_inference_steps - init_timestep, 0)
         timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
-        if hasattr(self.scheduler, "set_begin_index"):
-            self.scheduler.set_begin_index(t_start * self.scheduler.order)
 
         return timesteps, num_inference_steps - t_start
 

src/diffusers/pipelines/stable_cascade/pipeline_stable_cascade.py

@@ -100,8 +100,10 @@ class StableCascadeDecoderPipeline(DiffusionPipeline):
         )
         self.register_to_config(latent_dim_scale=latent_dim_scale)
 
-    def prepare_latents(self, image_embeddings, num_images_per_prompt, dtype, device, generator, latents, scheduler):
-        batch_size, channels, height, width = image_embeddings.shape
+    def prepare_latents(
+        self, batch_size, image_embeddings, num_images_per_prompt, dtype, device, generator, latents, scheduler
+    ):
+        _, channels, height, width = image_embeddings.shape
         latents_shape = (
             batch_size * num_images_per_prompt,
             4,
@@ -383,7 +385,19 @@ class StableCascadeDecoderPipeline(DiffusionPipeline):
         )
         if isinstance(image_embeddings, list):
             image_embeddings = torch.cat(image_embeddings, dim=0)
-        batch_size = image_embeddings.shape[0]
+
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        # Compute the effective number of images per prompt
+        # We must account for the fact that the image embeddings from the prior can be generated with num_images_per_prompt > 1
+        # This results in a case where a single prompt is associated with multiple image embeddings
+        # Divide the number of image embeddings by the batch size to determine if this is the case.
+        num_images_per_prompt = num_images_per_prompt * (image_embeddings.shape[0] // batch_size)
 
         # 2. Encode caption
         if prompt_embeds is None and negative_prompt_embeds is None:
@@ -417,7 +431,7 @@ class StableCascadeDecoderPipeline(DiffusionPipeline):
 
         # 5. Prepare latents
         latents = self.prepare_latents(
-            image_embeddings, num_images_per_prompt, dtype, device, generator, latents, self.scheduler
+            batch_size, image_embeddings, num_images_per_prompt, dtype, device, generator, latents, self.scheduler
         )
 
         # 6. Run denoising loop

src/diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py

@@ -716,15 +716,12 @@ class StableDiffusionDiffEditPipeline(
                     f" `source_negative_prompt_embeds` {source_negative_prompt_embeds.shape}."
                 )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.StableDiffusionImg2ImgPipeline.get_timesteps
     def get_timesteps(self, num_inference_steps, strength, device):
         # get the original timestep using init_timestep
         init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
 
         t_start = max(num_inference_steps - init_timestep, 0)
         timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
-        if hasattr(self.scheduler, "set_begin_index"):
-            self.scheduler.set_begin_index(t_start * self.scheduler.order)
 
         return timesteps, num_inference_steps - t_start
 

src/diffusers/schedulers/scheduling_consistency_models.py

@@ -434,7 +434,11 @@ class CMStochasticIterativeScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_deis_multistep.py

@@ -768,10 +768,14 @@ class DEISMultistepScheduler(SchedulerMixin, ConfigMixin):
             schedule_timesteps = self.timesteps.to(original_samples.device)
             timesteps = timesteps.to(original_samples.device)
 
-        # begin_index is None when the scheduler is used for training
+        # begin_index is None when the scheduler is used for training or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_dpmsolver_multistep.py

@@ -1011,10 +1011,14 @@ class DPMSolverMultistepScheduler(SchedulerMixin, ConfigMixin):
             schedule_timesteps = self.timesteps.to(original_samples.device)
             timesteps = timesteps.to(original_samples.device)
 
-        # begin_index is None when the scheduler is used for training
+        # begin_index is None when the scheduler is used for training or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_dpmsolver_sde.py

@@ -543,7 +543,11 @@ class DPMSolverSDEScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_dpmsolver_singlestep.py

@@ -961,10 +961,14 @@ class DPMSolverSinglestepScheduler(SchedulerMixin, ConfigMixin):
             schedule_timesteps = self.timesteps.to(original_samples.device)
             timesteps = timesteps.to(original_samples.device)
 
-        # begin_index is None when the scheduler is used for training
+        # begin_index is None when the scheduler is used for training or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_edm_dpmsolver_multistep.py

@@ -669,7 +669,11 @@ class EDMDPMSolverMultistepScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_edm_euler.py

@@ -367,7 +367,11 @@ class EDMEulerScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_euler_ancestral_discrete.py

@@ -467,7 +467,11 @@ class EulerAncestralDiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_euler_discrete.py

@@ -562,7 +562,11 @@ class EulerDiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_heun_discrete.py

@@ -468,7 +468,11 @@ class HeunDiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_k_dpm_2_ancestral_discrete.py

@@ -494,7 +494,11 @@ class KDPM2AncestralDiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_k_dpm_2_discrete.py

@@ -469,7 +469,11 @@ class KDPM2DiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_lms_discrete.py

@@ -461,7 +461,11 @@ class LMSDiscreteScheduler(SchedulerMixin, ConfigMixin):
         # self.begin_index is None when scheduler is used for training, or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

src/diffusers/schedulers/scheduling_unipc_multistep.py

@@ -862,10 +862,14 @@ class UniPCMultistepScheduler(SchedulerMixin, ConfigMixin):
             schedule_timesteps = self.timesteps.to(original_samples.device)
             timesteps = timesteps.to(original_samples.device)
 
-        # begin_index is None when the scheduler is used for training
+        # begin_index is None when the scheduler is used for training or pipeline does not implement set_begin_index
         if self.begin_index is None:
             step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps]
+        elif self.step_index is not None:
+            # add_noise is called after first denoising step (for inpainting)
+            step_indices = [self.step_index] * timesteps.shape[0]
         else:
+            # add noise is called bevore first denoising step to create inital latent(img2img)
             step_indices = [self.begin_index] * timesteps.shape[0]
 
         sigma = sigmas[step_indices].flatten()

tests/lora/test_lora_layers_peft.py

@@ -158,7 +158,7 @@ class PeftLoraLoaderMixinTests:
 
         pipeline_inputs = {
             "prompt": "A painting of a squirrel eating a burger",
-            "num_inference_steps": 2,
+            "num_inference_steps": 5,
             "guidance_scale": 6.0,
             "output_type": "np",
         }
@@ -589,7 +589,7 @@ class PeftLoraLoaderMixinTests:
                 **inputs, generator=torch.manual_seed(0), cross_attention_kwargs={"scale": 0.5}
             ).images
             self.assertTrue(
-                not np.allclose(output_lora, output_lora_scale, atol=1e-3, rtol=1e-3),
+                not np.allclose(output_lora, output_lora_scale, atol=1e-4, rtol=1e-4),
                 "Lora + scale should change the output",
             )
 
@@ -1300,6 +1300,11 @@ class StableDiffusionLoRATests(PeftLoraLoaderMixinTests, unittest.TestCase):
         pipe.load_lora_weights(lora_id)
         pipe = pipe.to("cuda")
 
+        self.assertTrue(
+            self.check_if_lora_correctly_set(pipe.unet),
+            "Lora not correctly set in UNet",
+        )
+
         self.assertTrue(
             self.check_if_lora_correctly_set(pipe.text_encoder),
             "Lora not correctly set in text encoder 2",

tests/pipelines/stable_cascade/test_stable_cascade_decoder.py

@@ -33,6 +33,7 @@ from diffusers.utils.testing_utils import (
     slow,
     torch_device,
 )
+from diffusers.utils.torch_utils import randn_tensor
 
 from ..test_pipelines_common import PipelineTesterMixin
 
@@ -246,6 +247,66 @@ class StableCascadeDecoderPipelineFastTests(PipelineTesterMixin, unittest.TestCa
 
         assert np.abs(decoder_output_prompt.images - decoder_output_prompt_embeds.images).max() < 1e-5
 
+    def test_stable_cascade_decoder_single_prompt_multiple_image_embeddings(self):
+        device = "cpu"
+        components = self.get_dummy_components()
+
+        pipe = StableCascadeDecoderPipeline(**components)
+        pipe.set_progress_bar_config(disable=None)
+
+        prior_num_images_per_prompt = 2
+        decoder_num_images_per_prompt = 2
+        prompt = ["a cat"]
+        batch_size = len(prompt)
+
+        generator = torch.Generator(device)
+        image_embeddings = randn_tensor(
+            (batch_size * prior_num_images_per_prompt, 4, 4, 4), generator=generator.manual_seed(0)
+        )
+        decoder_output = pipe(
+            image_embeddings=image_embeddings,
+            prompt=prompt,
+            num_inference_steps=1,
+            output_type="np",
+            guidance_scale=0.0,
+            generator=generator.manual_seed(0),
+            num_images_per_prompt=decoder_num_images_per_prompt,
+        )
+
+        assert decoder_output.images.shape[0] == (
+            batch_size * prior_num_images_per_prompt * decoder_num_images_per_prompt
+        )
+
+    def test_stable_cascade_decoder_single_prompt_multiple_image_embeddings_with_guidance(self):
+        device = "cpu"
+        components = self.get_dummy_components()
+
+        pipe = StableCascadeDecoderPipeline(**components)
+        pipe.set_progress_bar_config(disable=None)
+
+        prior_num_images_per_prompt = 2
+        decoder_num_images_per_prompt = 2
+        prompt = ["a cat"]
+        batch_size = len(prompt)
+
+        generator = torch.Generator(device)
+        image_embeddings = randn_tensor(
+            (batch_size * prior_num_images_per_prompt, 4, 4, 4), generator=generator.manual_seed(0)
+        )
+        decoder_output = pipe(
+            image_embeddings=image_embeddings,
+            prompt=prompt,
+            num_inference_steps=1,
+            output_type="np",
+            guidance_scale=2.0,
+            generator=generator.manual_seed(0),
+            num_images_per_prompt=decoder_num_images_per_prompt,
+        )
+
+        assert decoder_output.images.shape[0] == (
+            batch_size * prior_num_images_per_prompt * decoder_num_images_per_prompt
+        )
+
 
 @slow
 @require_torch_gpu

tests/pipelines/stable_diffusion/test_stable_diffusion_inpaint.py

@@ -29,6 +29,7 @@ from diffusers import (
     AutoencoderKL,
     DDIMScheduler,
     DPMSolverMultistepScheduler,
+    EulerAncestralDiscreteScheduler,
     LCMScheduler,
     LMSDiscreteScheduler,
     PNDMScheduler,
@@ -557,6 +558,29 @@ class StableDiffusionSimpleInpaintPipelineFastTests(StableDiffusionInpaintPipeli
         image_slice2 = images[1, -3:, -3:, -1]
         assert np.abs(image_slice1.flatten() - image_slice2.flatten()).max() > 1e-2
 
+    def test_stable_diffusion_inpaint_euler(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        components = self.get_dummy_components(time_cond_proj_dim=256)
+        sd_pipe = StableDiffusionInpaintPipeline(**components)
+        sd_pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(sd_pipe.scheduler.config)
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device, output_pil=False)
+        half_dim = inputs["image"].shape[2] // 2
+        inputs["mask_image"][0, 0, :half_dim, :half_dim] = 0
+
+        inputs["num_inference_steps"] = 4
+        image = sd_pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 64, 64, 3)
+
+        expected_slice = np.array(
+            [[0.6387283, 0.5564158, 0.58631873, 0.5539942, 0.5494673, 0.6461868, 0.5251618, 0.5497595, 0.5508756]]
+        )
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-4
+
 
 @slow
 @require_torch_gpu