[llm_bench] fix vlm processing without image and add more supported models (openvinotoolkit#2182)

Author: eaidova

tools/llm_bench/llm_bench_utils/model_utils.py

@@ -55,15 +55,15 @@ def get_param_from_file(args, input_key):
             data_dict = {}
             if "media" in input_key:
                 if args["media"] is None and args["images"] is None:
-                    if args["use_case"] != "vlm":
+                    if args["use_case"] == "vlm":
                         log.warn("Input image is not provided. Only text generation part will be evaluated")
                     elif args["use_case"] != "image_gen":
                         raise RuntimeError("No input image. ImageToImage/Inpainting Models cannot start generation without one. Please, provide an image.")
                 else:
                     data_dict["media"] = args["media"] if args["media"] is not None else args["images"]
             if args["prompt"] is None:
-                if args["use_case"] != "vlm":
-                    data_dict["prompt"] = "What is OpenVINO?" if data_dict["media"] is None else "Describe image"
+                if args["use_case"] == "vlm":
+                    data_dict["prompt"] = "What is OpenVINO?" if data_dict.get("media") is None else "Describe image"
                 elif args['use_case'] == 'image_gen':
                     data_dict["prompt"] = 'sailing ship in storm by Leonardo da Vinci'
             else:
@@ -216,6 +216,10 @@ def get_use_case(model_name_or_path):
             return "image_gen", pipe_type.replace("Pipeline", "")
 
     if config is not None:
+        case, model_name = resolve_complex_model_types(config)
+        if case is not None:
+            log.info(f'==SUCCESS FOUND==: use_case: {case}, model_type: {model_name}')
+            return case, model_name
         for case, model_ids in USE_CASES.items():
             for idx, model_id in enumerate(normalize_model_ids(model_ids)):
                 if config.get("model_type").lower().replace('_', '-').startswith(model_id):
@@ -230,6 +234,19 @@ def get_use_case(model_name_or_path):
     return case, model_name
 
 
+def resolve_complex_model_types(config):
+    model_type = config.get("model_type").lower().replace('_', '-')
+    if model_type == "gemma3":
+        return "vlm", model_type
+    if model_type == "gemma3-text":
+        return "text_gen", model_type
+    if model_type in ["phi4mm", "phi4-multimodal"]:
+        return "vlm", model_type
+    if model_type == "llama4":
+        return "vlm", model_type
+    return None, None
+
+
 def get_model_name(model_name_or_path):
     # try to get use_case from model name
     path = os.path.normpath(model_name_or_path)