Clean pipeline using Neuraxle

[guillaume-chevalier]

Should do something that looks like this to clean the project by using Neuraxle:

deep_learning_seq_classif_pipeline = EpochRepeater(Pipeline([
    Windower(),  # TODO: not sure where this must be plugged into the pipeline.
    AutoRegress(),  # TODO: AutoRegress might be inside NormalizeThenUnNormalizeWrapper, not sure.
    NormalizeThenUnNormalizeWrapper(
        TrainOnlyWrapper(DataShuffler(seed=42)),
        MiniBatchSequentialPipeline([
            ForEachDataInput(Pipeline([
                ToNumpy(np_dtype=np.float32),
                DefaultValuesFiller(0.0),
            ])),
            ForecastModel(n_stacked=2, n_residual=3),
        ], batch_size=32),
    )
]), epochs=200, fit_only=True)

Where ForecastModel contains the actual TensorFlow code.

[guillaume-chevalier]

Similar issue: guillaume-chevalier/LSTM-Human-Activity-Recognition#33

[guillaume-chevalier]

Solved with #17