Improve post-processing performance

* Use multiplication instead of division
* Avoid splitting and re-stacking tensors to reduce memory bandwidth and CPU-GPU syncs

Author: soof-golan

src/diffusers/image_processor.py

@@ -236,7 +236,7 @@ def denormalize(images: Union[np.ndarray, torch.Tensor]) -> Union[np.ndarray, to
             `np.ndarray` or `torch.Tensor`:
                 The denormalized image array.
         """
-        return (images / 2 + 0.5).clamp(0, 1)
+        return (images * 0.5 + 0.5).clamp(0, 1)
 
     @staticmethod
     def convert_to_rgb(image: PIL.Image.Image) -> PIL.Image.Image:
@@ -537,6 +537,27 @@ def binarize(self, image: PIL.Image.Image) -> PIL.Image.Image:
 
         return image
 
+    def _denormalize_conditionally(self, images: torch.Tensor, do_denormalize: Optional[List[bool]] = None) -> torch.Tensor:
+        r"""
+        Denormalize a batch of images based on a condition list.
+
+        Args:
+            images (`torch.Tensor`):
+                The input image tensor.
+            do_denormalize (`Optional[List[bool]`, *optional*, defaults to `None`):
+                A list of booleans indicating whether to denormalize each image in the batch. If `None`, will use the
+                value of `do_normalize` in the `VaeImageProcessor` config.
+        """
+        if do_denormalize is None:
+            return self.denormalize(images) if self.config.do_normalize else images
+
+        # De-normalizing a batch and selectively torch.stack'ing the results turns out to be
+        # significantly faster than performing a lot of smaller denormalizations
+        denormalized = self.denormalize(images)
+        return torch.stack([
+            denormalized[i] if do_denormalize[i] else images[i] for i in range(images.shape[0])
+        ])
+
     def get_default_height_width(
         self,
         image: Union[PIL.Image.Image, np.ndarray, torch.Tensor],
@@ -752,12 +773,7 @@ def postprocess(
         if output_type == "latent":
             return image
 
-        if do_denormalize is None:
-            do_denormalize = [self.config.do_normalize] * image.shape[0]
-
-        image = torch.stack(
-            [self.denormalize(image[i]) if do_denormalize[i] else image[i] for i in range(image.shape[0])]
-        )
+        image = self._denormalize_conditionally(image, do_denormalize)
 
         if output_type == "pt":
             return image
@@ -966,12 +982,7 @@ def postprocess(
             deprecate("Unsupported output_type", "1.0.0", deprecation_message, standard_warn=False)
             output_type = "np"
 
-        if do_denormalize is None:
-            do_denormalize = [self.config.do_normalize] * image.shape[0]
-
-        image = torch.stack(
-            [self.denormalize(image[i]) if do_denormalize[i] else image[i] for i in range(image.shape[0])]
-        )
+        image = self._denormalize_conditionally(image, do_denormalize)
 
         image = self.pt_to_numpy(image)