fix tests

Author: IlyasMoutawwakil

optimum/intel/openvino/modeling_visual_language.py

@@ -1021,12 +1021,16 @@ def preprocess_inputs(
             else:
                 prompt = text
 
-        if (
-            getattr(processor, "patch_size", None) is None
-            and getattr(config, "vision_config", None) is not None
-            and getattr(config.vision_config, "patch_size", None) is not None
-        ):
-            processor.patch_size = config.vision_config.patch_size
+        if getattr(processor, "patch_size", None) is None:
+            if (
+                getattr(config, "vision_config", None) is not None
+                and getattr(config.vision_config, "patch_size", None) is not None
+            ):
+                processor.patch_size = config.vision_config.patch_size
+            else:
+                raise ValueError(
+                    "Processor does not have `patch_size` attribute. Please fix the processor or provide `patch_size` in the config."
+                )
 
         inputs = processor(images=image, text=prompt, return_tensors="pt")
         return inputs
@@ -1923,6 +1927,7 @@ def preprocess_inputs(
             input_ids = tokenizer(text, return_tensors="pt").input_ids
         attention_mask = torch.ones_like(input_ids, dtype=torch.int64)
         result = {"input_ids": input_ids, "attention_mask": attention_mask}
+
         if image is not None:
             result["images"] = processor(images=[image], return_tensors="pt")["pixel_values"]
         return result

tests/openvino/test_modeling.py

@@ -2311,17 +2311,17 @@ def test_generate_utils(self, model_arch):
 
     def get_preprocessors(self, model_arch):
         model_id = MODEL_NAMES[model_arch]
+        config = AutoConfig.from_pretrained(model_id, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS)
+
         if model_arch == "nanollava":
-            config = AutoConfig.from_pretrained(model_id, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS)
             processor = AutoProcessor.from_pretrained(
                 config.mm_vision_tower, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS
             )
             tokenizer = AutoTokenizer.from_pretrained(
                 model_id, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS
             )
-            preprocessors = {"processor": processor, "tokenizer": tokenizer}
+            preprocessors = {"processor": processor, "tokenizer": tokenizer, "config": config}
         elif model_arch == "internvl2":
-            config = AutoConfig.from_pretrained(model_id, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS)
             tokenizer = AutoTokenizer.from_pretrained(
                 model_id, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS
             )
@@ -2330,7 +2330,8 @@ def get_preprocessors(self, model_arch):
             processor = AutoProcessor.from_pretrained(
                 model_id, trust_remote_code=model_arch in self.REMOTE_CODE_MODELS
             )
-            preprocessors = {"processor": processor, "tokenizer": None}
+            preprocessors = {"processor": processor, "tokenizer": None, "config": config}
+
         return preprocessors
 
     @parameterized.expand(SUPPORTED_ARCHITECTURES)