Hi there, The Pipeline of ml package is really a great feature and we use it in our every day task. But we have some use case where we need a Pipeline of Transformers only and the problem is that there's not train phase in that case. For example, we have a pipeline of image analytics with the following step : ImageDecoder -> FeatureExtraction -> Maching Learning algo. Some times, the machine learning algorithms are not part of the pipeline but we still need the ImageDecoder + FeatureExtraction steps. The FeatureExtraction step has no train phase (FFT, SIFT, PreTrained CNN, ...) and it this case calling .fit method in order to get the pipeline model has no meaningful signification. How can we handle this case correctly ?
Cheers, Jao