diffusers 0.30.2__py3-none-any.whl → 0.31.0__py3-none-any.whl

diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py

@@ -1643,9 +1643,21 @@ class LEditsPPPipelineStableDiffusionXL(
 
 # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    """
-    Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
-    Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf). See Section 3.4
+    r"""
+    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
+    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
+    Flawed](https://arxiv.org/pdf/2305.08891.pdf).
+
+    Args:
+        noise_cfg (`torch.Tensor`):
+            The predicted noise tensor for the guided diffusion process.
+        noise_pred_text (`torch.Tensor`):
+            The predicted noise tensor for the text-guided diffusion process.
+        guidance_rescale (`float`, *optional*, defaults to 0.0):
+            A rescale factor applied to the noise predictions.
+
+    Returns:
+        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
     """
     std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
     std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)

diffusers/pipelines/lumina/pipeline_lumina.py

@@ -54,7 +54,7 @@ EXAMPLE_DOC_STRING = """
 
         >>> pipe = LuminaText2ImgPipeline.from_pretrained(
         ...     "Alpha-VLLM/Lumina-Next-SFT-diffusers", torch_dtype=torch.bfloat16
-        ... ).cuda()
+        ... )
         >>> # Enable memory optimizations.
         >>> pipe.enable_model_cpu_offload()
 
@@ -73,7 +73,7 @@ def retrieve_timesteps(
     sigmas: Optional[List[float]] = None,
     **kwargs,
 ):
-    """
+    r"""
     Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
     custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
 

diffusers/pipelines/pag/__init__.py

@@ -23,13 +23,16 @@ except OptionalDependencyNotAvailable:
     _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
 else:
     _import_structure["pipeline_pag_controlnet_sd"] = ["StableDiffusionControlNetPAGPipeline"]
+    _import_structure["pipeline_pag_controlnet_sd_inpaint"] = ["StableDiffusionControlNetPAGInpaintPipeline"]
     _import_structure["pipeline_pag_controlnet_sd_xl"] = ["StableDiffusionXLControlNetPAGPipeline"]
+    _import_structure["pipeline_pag_controlnet_sd_xl_img2img"] = ["StableDiffusionXLControlNetPAGImg2ImgPipeline"]
     _import_structure["pipeline_pag_hunyuandit"] = ["HunyuanDiTPAGPipeline"]
     _import_structure["pipeline_pag_kolors"] = ["KolorsPAGPipeline"]
     _import_structure["pipeline_pag_pixart_sigma"] = ["PixArtSigmaPAGPipeline"]
     _import_structure["pipeline_pag_sd"] = ["StableDiffusionPAGPipeline"]
     _import_structure["pipeline_pag_sd_3"] = ["StableDiffusion3PAGPipeline"]
     _import_structure["pipeline_pag_sd_animatediff"] = ["AnimateDiffPAGPipeline"]
+    _import_structure["pipeline_pag_sd_img2img"] = ["StableDiffusionPAGImg2ImgPipeline"]
     _import_structure["pipeline_pag_sd_xl"] = ["StableDiffusionXLPAGPipeline"]
     _import_structure["pipeline_pag_sd_xl_img2img"] = ["StableDiffusionXLPAGImg2ImgPipeline"]
     _import_structure["pipeline_pag_sd_xl_inpaint"] = ["StableDiffusionXLPAGInpaintPipeline"]
@@ -43,13 +46,16 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from ...utils.dummy_torch_and_transformers_objects import *
     else:
         from .pipeline_pag_controlnet_sd import StableDiffusionControlNetPAGPipeline
+        from .pipeline_pag_controlnet_sd_inpaint import StableDiffusionControlNetPAGInpaintPipeline
         from .pipeline_pag_controlnet_sd_xl import StableDiffusionXLControlNetPAGPipeline
+        from .pipeline_pag_controlnet_sd_xl_img2img import StableDiffusionXLControlNetPAGImg2ImgPipeline
         from .pipeline_pag_hunyuandit import HunyuanDiTPAGPipeline
         from .pipeline_pag_kolors import KolorsPAGPipeline
         from .pipeline_pag_pixart_sigma import PixArtSigmaPAGPipeline
         from .pipeline_pag_sd import StableDiffusionPAGPipeline
         from .pipeline_pag_sd_3 import StableDiffusion3PAGPipeline
         from .pipeline_pag_sd_animatediff import AnimateDiffPAGPipeline
+        from .pipeline_pag_sd_img2img import StableDiffusionPAGImg2ImgPipeline
         from .pipeline_pag_sd_xl import StableDiffusionXLPAGPipeline
         from .pipeline_pag_sd_xl_img2img import StableDiffusionXLPAGImg2ImgPipeline
         from .pipeline_pag_sd_xl_inpaint import StableDiffusionXLPAGInpaintPipeline

diffusers/pipelines/pag/pag_utils.py

@@ -98,7 +98,9 @@ class PAGMixin:
         else:
             return self.pag_scale
 
-    def _apply_perturbed_attention_guidance(self, noise_pred, do_classifier_free_guidance, guidance_scale, t):
+    def _apply_perturbed_attention_guidance(
+        self, noise_pred, do_classifier_free_guidance, guidance_scale, t, return_pred_text=False
+    ):
         r"""
         Apply perturbed attention guidance to the noise prediction.
 
@@ -107,9 +109,11 @@ class PAGMixin:
             do_classifier_free_guidance (bool): Whether to apply classifier-free guidance.
             guidance_scale (float): The scale factor for the guidance term.
             t (int): The current time step.
+            return_pred_text (bool): Whether to return the text noise prediction.
 
         Returns:
-            torch.Tensor: The updated noise prediction tensor after applying perturbed attention guidance.
+            Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]: The updated noise prediction tensor after applying
+            perturbed attention guidance and the text noise prediction.
         """
         pag_scale = self._get_pag_scale(t)
         if do_classifier_free_guidance:
@@ -122,6 +126,8 @@ class PAGMixin:
         else:
             noise_pred_text, noise_pred_perturb = noise_pred.chunk(2)
             noise_pred = noise_pred_text + pag_scale * (noise_pred_text - noise_pred_perturb)
+        if return_pred_text:
+            return noise_pred, noise_pred_text
         return noise_pred
 
     def _prepare_perturbed_attention_guidance(self, cond, uncond, do_classifier_free_guidance):

diffusers/pipelines/pag/pipeline_pag_controlnet_sd.py

@@ -104,7 +104,7 @@ def retrieve_timesteps(
     sigmas: Optional[List[float]] = None,
     **kwargs,
 ):
-    """
+    r"""
     Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
     custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.