Fix OwlViT torchscript tests (#18347)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/owlvit/modeling_owlvit.py

@@ -1153,7 +1153,6 @@ class OwlViTClassPredictionHead(nn.Module):
 
 class OwlViTForObjectDetection(OwlViTPreTrainedModel):
     config_class = OwlViTConfig
-    main_input_name = "pixel_values"
 
     def __init__(self, config: OwlViTConfig):
         super().__init__(config)
@@ -1246,8 +1245,8 @@ class OwlViTForObjectDetection(OwlViTPreTrainedModel):
 
     def image_text_embedder(
         self,
-        pixel_values: torch.FloatTensor,
         input_ids: torch.Tensor,
+        pixel_values: torch.FloatTensor,
         attention_mask: torch.Tensor,
         output_attentions: Optional[bool] = None,
     ) -> torch.FloatTensor:
@@ -1284,8 +1283,8 @@ class OwlViTForObjectDetection(OwlViTPreTrainedModel):
     @replace_return_docstrings(output_type=OwlViTObjectDetectionOutput, config_class=OwlViTConfig)
     def forward(
         self,
-        pixel_values: torch.FloatTensor,
         input_ids: torch.Tensor,
+        pixel_values: torch.FloatTensor,
         attention_mask: Optional[torch.Tensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
@@ -1338,8 +1337,8 @@ class OwlViTForObjectDetection(OwlViTPreTrainedModel):
 
         if output_hidden_states:
             outputs = self.owlvit(
-                pixel_values=pixel_values,
                 input_ids=input_ids,
+                pixel_values=pixel_values,
                 attention_mask=attention_mask,
                 output_attentions=output_attentions,
                 output_hidden_states=output_hidden_states,
@@ -1350,8 +1349,8 @@ class OwlViTForObjectDetection(OwlViTPreTrainedModel):
 
         # Embed images and text queries
         feature_map, query_embeds = self.image_text_embedder(
-            pixel_values=pixel_values,
             input_ids=input_ids,
+            pixel_values=pixel_values,
             attention_mask=attention_mask,
             output_attentions=output_attentions,
         )
@@ -1374,7 +1373,7 @@ class OwlViTForObjectDetection(OwlViTPreTrainedModel):
         pred_boxes = self.box_predictor(image_feats, feature_map)
 
         if not return_dict:
-            return (
+            output = (
                 pred_logits,
                 pred_boxes,
                 query_embeds,
@@ -1383,6 +1382,8 @@ class OwlViTForObjectDetection(OwlViTPreTrainedModel):
                 text_model_last_hidden_states,
                 vision_model_last_hidden_states,
             )
+            output = tuple(x for x in output if x is not None)
+            return output
 
         return OwlViTObjectDetectionOutput(
             image_embeds=feature_map,