NAS ModelStatus failed after running

Question

I am running a neural architecture search (NAS) experiment on Google Colab. But when I exported the top model by running exp.export_top_models(), it says:

[2023-11-13 02:01:30] WARNING: Only 0 models are trained, but 1 top models are requested.
WARNING:nni.nas.strategy.base:Only 0 models are trained, but 1 top models are requested.
[]

I checked the source code and checked the output of list(exp.strategy.engine.list_models()), which returns

[SimplifiedModelSpace(model=SerializableObject(type=MyModelSpace), mutables=MutableDict({}), evaluator=FunctionalEvaluator(<function evaluate_model at 0x7fbbf3d324d0>, arguments={})), sample={'conv2': 0, 'dropout': 0.5, 'feature': 64}, status=ModelStatus.Failed),
 SimplifiedModelSpace(model=SerializableObject(type=MyModelSpace), mutables=MutableDict({}), evaluator=FunctionalEvaluator(<function evaluate_model at 0x7fbbf3d32560>, arguments={})), sample={'conv2': 1, 'dropout': 0.75, 'feature': 64}, status=ModelStatus.Failed),
 SimplifiedModelSpace(model=SerializableObject(type=MyModelSpace), mutables=MutableDict({}), evaluator=FunctionalEvaluator(<function evaluate_model at 0x7fbbf3daa560>, arguments={})), sample={'conv2': 1, 'dropout': 0.25, 'feature': 256}, status=ModelStatus.Failed)]

The model status are all failed. I wonder in what situations it would return a failed model, and what could go wrong? How shall I figure this out. Thanks.