try next runner if load fails

ollama/engine.py

@@ -39,12 +39,18 @@ def load(model_name, models={}):
             for model_type in cls.model_types()
         }
 
+        while len(runners) > 0:
+            try:
                 best_match, _ = process.extractOne(model_path, runners.keys())
                 model = runners.get(best_match, LlamaCppRunner)
-
-        models.update({model_name: model(model_path, best_match)})
-
+                runner = model(model_path, best_match)
+                models.update({model_name: runner})
                 return models.get(model_name)
+            except Exception:
+                # try the next runner
+                runners.pop(best_match)
+
+        raise Exception("failed to load model", model_path, model_name)
 
 
 def unload(model_name, models={}):