Using Type annotations in functions annotated by @is_fairstep raises issues

[vemonet]

Hi, I am trying to use fairworkflows to describe a simple ML workflow, but I am facing a few issues

For example whenever I try to use a type annotation to annotate the input or return value of a function I get an error

• Type annotation in function parameters:

@is_fairstep(label='Create and fit classifier', is_script_task=True)
def fit_classifier(hyper_params: Dict[str, float], data, y):
    clf = RandomForestClassifier(
        n_jobs=hyper_params['n_jobs'],
        random_state=hyper_params['random_state'],
    )
    clf.fit(data, y)
    return clf

Will raise something like:

Traceback (most recent call last):
  File "my-model/src/my_model/train.py", line 51, in <module>
    def fit_classifier(hyper_params: Dict[str, float], data, y):
  File "site-packages/fairworkflows/fairstep.py", line 452, in _modify_function
    inputs = _extract_inputs_from_function(func, kwargs)
  File "site-packages/fairworkflows/fairstep.py", line 504, in _extract_inputs_from_function
    computational_type = argspec.annotations[arg].__name__
  File "/opt/conda/lib/python3.9/typing.py", line 711, in __getattr__
    raise AttributeError(attr)
AttributeError: __name__

• Type annotation for function return:

@is_fairstep(label='Evaluate the trained model', is_script_task=True)
def evaluate() -> Dict[str, float]:
    return {
        'precision': 0.85,
        'recall': 0.80,
        'accuracy': 0.85,
        'roc_auc': 0.90,
        'f1': 0.75,
        'average_precision': 0.85,
    }

will raise:

Traceback (most recent call last):
  File "my-model/src/my_model/train.py", line 61, in <module>
    def evaluate(model) -> Dict[str, float]:
  File "site-packages/fairworkflows/fairstep.py", line 453, in _modify_function
    outputs = _extract_outputs_from_function(func, kwargs)
  File "site-packages/fairworkflows/fairstep.py", line 551, in _extract_outputs_from_function
    computational_type = return_annotation.__name__ if return_annotation is not None else None
  File "/opt/conda/lib/python3.9/typing.py", line 711, in __getattr__
    raise AttributeError(attr)
AttributeError: __name__