changed control flow so that errors caused by sample_weights or prios will be raised and will not result into falling back to the next input shape case.
At the moment, if you pass data as a 3D tensor (case 1), but either the priors or the sample_weight would be invalid, you will not get the error caused by these parameters, but rather will get an "obscure" error related the the shape of X itself, as the code falls back to assuming the input is a list of 2D arrays (which it is not).
This PR changes the control flow to "select" the input type case only based on errors raised during validation of X itself, but will correctly raise all errors caused by the validation of sample_weight or priors
changed control flow so that errors caused by sample_weights or prios will be raised and will not result into falling back to the next input shape case.
At the moment, if you pass data as a 3D tensor (case 1), but either the priors or the sample_weight would be invalid, you will not get the error caused by these parameters, but rather will get an "obscure" error related the the shape of X itself, as the code falls back to assuming the input is a list of 2D arrays (which it is not).
This PR changes the control flow to "select" the input type case only based on errors raised during validation of X itself, but will correctly raise all errors caused by the validation of
sample_weightorpriors