tsai

tsai copied to clipboard

I use MiniRocketClassifier on a multivariate time series dataset

cls = MiniRocketClassifier()
cls.fit(X, y)

An error occurred

Traceback (most recent call last):
  File "e:\...\src\pipeline\classification\ROCKET.py", line 190, in <module>
    cls.fit(X_train, y_train)
  File "G:\Anaconda\envs\coolprop\Lib\site-packages\sklearn\base.py", line 1389, in wrapper
    return fit_method(estimator, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "G:\Anaconda\envs\coolprop\Lib\site-packages\sklearn\pipeline.py", line 646, in fit
    if not _routing_enabled() and self.transform_input is not None:
                                  ^^^^^^^^^^^^^^^^^^^^
AttributeError: 'MiniRocketClassifier' object has no attribute 'transform_input'

The same error occurred when I use RocketClassifier

File "e:\...\src\pipeline\classification\ROCKET.py", line 30, in fit
    cls.fit(X, y)
  File "G:\Anaconda\envs\coolprop\Lib\site-packages\sklearn\base.py", line 1389, in wrapper
    return fit_method(estimator, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "G:\Anaconda\envs\coolprop\Lib\site-packages\sklearn\pipeline.py", line 646, in fit
    if not _routing_enabled() and self.transform_input is not None:
                                  ^^^^^^^^^^^^^^^^^^^^
AttributeError: 'RocketClassifier' object has no attribute 'transform_input'

Package version:

• scikit-learn 1.6.1
• tsai 0.4.0

I wonder if this is caused by the incompatibility of the two library versions