Performance Improve for Qwen Image Edit (huggingface#12190)

* fix(qwen-image-edit):
- update condition reshaping logic to improve editing performance

* fix(qwen-image-edit):
- remove _auto_resize

Author: naykun

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py

@@ -62,25 +62,6 @@
         >>> image.save("qwenimage_edit.png")
         ```
 """
-PREFERRED_QWENIMAGE_RESOLUTIONS = [
-    (672, 1568),
-    (688, 1504),
-    (720, 1456),
-    (752, 1392),
-    (800, 1328),
-    (832, 1248),
-    (880, 1184),
-    (944, 1104),
-    (1024, 1024),
-    (1104, 944),
-    (1184, 880),
-    (1248, 832),
-    (1328, 800),
-    (1392, 752),
-    (1456, 720),
-    (1504, 688),
-    (1568, 672),
-]
 
 
 # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
@@ -565,7 +546,6 @@ def __call__(
         callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
         callback_on_step_end_tensor_inputs: List[str] = ["latents"],
         max_sequence_length: int = 512,
-        _auto_resize: bool = True,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -684,18 +664,9 @@ def __call__(
         device = self._execution_device
         # 3. Preprocess image
         if image is not None and not (isinstance(image, torch.Tensor) and image.size(1) == self.latent_channels):
-            img = image[0] if isinstance(image, list) else image
-            image_height, image_width = self.image_processor.get_default_height_width(img)
-            aspect_ratio = image_width / image_height
-            if _auto_resize:
-                _, image_width, image_height = min(
-                    (abs(aspect_ratio - w / h), w, h) for w, h in PREFERRED_QWENIMAGE_RESOLUTIONS
-                )
-            image_width = image_width // multiple_of * multiple_of
-            image_height = image_height // multiple_of * multiple_of
-            image = self.image_processor.resize(image, image_height, image_width)
+            image = self.image_processor.resize(image, calculated_height, calculated_width)
             prompt_image = image
-            image = self.image_processor.preprocess(image, image_height, image_width)
+            image = self.image_processor.preprocess(image, calculated_height, calculated_width)
             image = image.unsqueeze(2)
 
         has_neg_prompt = negative_prompt is not None or (
@@ -712,9 +683,6 @@ def __call__(
             max_sequence_length=max_sequence_length,
         )
         if do_true_cfg:
-            # negative image is the same size as the original image, but all pixels are white
-            # negative_image = Image.new("RGB", (image.width, image.height), (255, 255, 255))
-
             negative_prompt_embeds, negative_prompt_embeds_mask = self.encode_prompt(
                 image=prompt_image,
                 prompt=negative_prompt,
@@ -741,7 +709,7 @@ def __call__(
         img_shapes = [
             [
                 (1, height // self.vae_scale_factor // 2, width // self.vae_scale_factor // 2),
-                (1, image_height // self.vae_scale_factor // 2, image_width // self.vae_scale_factor // 2),
+                (1, calculated_height // self.vae_scale_factor // 2, calculated_width // self.vae_scale_factor // 2),
             ]
         ] * batch_size