add tests

Author: eaidova

optimum/exporters/openvino/model_configs.py

@@ -131,10 +131,11 @@ def init_model_configs():
         "transformers",
         "Qwen2VLForConditionalGeneration",
     )
-    TasksManager._CUSTOM_CLASSES[("pt", "qwen2-5-vl", "image-text-to-text")] = (
-        "transformers",
-        "Qwen2_5_VLForConditionalGeneration",
-    )
+    if is_transformers_version(">", "4.48.99"):
+        TasksManager._CUSTOM_CLASSES[("pt", "qwen2-5-vl", "image-text-to-text")] = (
+            "transformers",
+            "Qwen2_5_VLForConditionalGeneration",
+        )
     TasksManager._TRANSFORMERS_TASKS_TO_MODEL_LOADERS[
         "image-text-to-text"
     ] = TasksManager._TRANSFORMERS_TASKS_TO_MODEL_LOADERS["text-generation"]

optimum/intel/openvino/modeling_visual_language.py

@@ -2658,8 +2658,7 @@ def get_vision_embeddings(self, pixel_values, grid_thw, **kwargs):
         window_index, cu_window_seqlens = self.get_window_index(grid_thw)
         cu_window_seqlens = torch.tensor(
             cu_window_seqlens,
-            device=hidden_states.device,
-            dtype=grid_thw.dtype if torch.jit.is_tracing() else torch.int32,
+            dtype=torch.int32,
         )
         cu_window_seqlens = torch.unique_consecutive(cu_window_seqlens)
         cu_seqlens = torch.repeat_interleave(grid_thw[:, 1] * grid_thw[:, 2], grid_thw[:, 0]).cumsum(

tests/openvino/test_modeling.py

@@ -2115,8 +2115,11 @@ class OVModelForVisualCausalLMIntegrationTest(unittest.TestCase):
         SUPPORTED_ARCHITECTURES += ["llava_next", "nanollava"]
     if is_transformers_version(">=", "4.45.0"):
         SUPPORTED_ARCHITECTURES += ["minicpmv", "internvl2", "phi3_v", "qwen2_vl", "maira2"]
+    
+    if is_transformers_version(">=", "4.48.0"):
+        SUPPORTED_ARCHITECTURES += ["qwen2_5_vl"]
     TASK = "image-text-to-text"
-    REMOTE_CODE_MODELS = ["internvl2", "minicpmv", "nanollava", "phi3_v", "maira2"]
+    REMOTE_CODE_MODELS = ["internvl2", "minicpmv", "nanollava", "phi3_v", "maira2", "qwen2_5_vl"]
 
     IMAGE = Image.open(
         requests.get(

tests/openvino/utils_tests.py

@@ -126,6 +126,7 @@
     "qwen2": "fxmarty/tiny-dummy-qwen2",
     "qwen2-moe": "katuni4ka/tiny-random-qwen1.5-moe",
     "qwen2_vl": "katuni4ka/tiny-random-qwen2vl",
+    "qwen2_5_vl": "katuni4ka/tiny-random-qwen2.5-vl",
     "resnet": "hf-internal-testing/tiny-random-resnet",
     "roberta": "hf-internal-testing/tiny-random-roberta",
     "roformer": "hf-internal-testing/tiny-random-roformer",