Fix depth pipeline "input/weight type should be the same" error at fp16 (huggingface#8321)

Fix "input/weight type should be the same"

Co-authored-by: YiYi Xu <[email protected]>

Author: jonahclarsen

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py

@@ -546,7 +546,7 @@ def prepare_depth_map(self, image, depth_map, batch_size, do_classifier_free_gui
 
         if depth_map is None:
             pixel_values = self.feature_extractor(images=image, return_tensors="pt").pixel_values
-            pixel_values = pixel_values.to(device=device)
+            pixel_values = pixel_values.to(device=device, dtype=dtype)
             # The DPT-Hybrid model uses batch-norm layers which are not compatible with fp16.
             # So we use `torch.autocast` here for half precision inference.
             if torch.backends.mps.is_available():