Fix model_type coercion in Engine parser

Adjust `Engine.from_model_type` to normalize non-string `model_type` values by converting either `.value` or the original object with `str(...)` in a single step. This avoids eager fallback conversion inside `getattr` and makes enum-like inputs more reliably coercible.

Author: C-Achard

dlclivegui/temp/engine.py

@@ -26,6 +26,12 @@ def is_tensorflow_model_dir_path(model_path: str | Path) -> bool:
 
     @classmethod
     def from_model_type(cls, model_type: str) -> "Engine":
+        if not isinstance(model_type, str):
+            try:
+                model_type = str(getattr(model_type, "value", model_type))
+            except Exception as e:
+                raise ValueError(f"Could not convert model_type to string: {model_type}") from e
+
         if model_type.lower() == "pytorch":
             return cls.PYTORCH
         elif model_type.lower() in ("tensorflow", "base", "tensorrt", "lite"):