Making Linearized a transformer without FIT

bob/pipelines/mixins.py

@@ -222,7 +222,6 @@ class SampleMixin:
             raise ValueError("Type for sample not supported %s" % type(samples))
 
     def fit(self, samples, y=None):
-
         # if the super method is not fittable,
         # there's no reason to stack those samples
         if hasattr(super(), "fit"):

bob/pipelines/transformers/linearize.py

@@ -4,21 +4,17 @@
 
 
 from bob.pipelines.mixins import CheckpointMixin, SampleMixin
-from sklearn.preprocessing import FunctionTransformer
+from sklearn.base import TransformerMixin
 import numpy as np
 
 
-def linearize(X):
-    X = np.asarray(X)
-    return np.reshape(X, (X.shape[0], -1))
-
-
-class Linearize(FunctionTransformer):
+class Linearize(TransformerMixin):
     """Extracts features by simply concatenating all elements of the data into one long vector.
     """
 
-    def __init__(self, **kwargs):
-        super().__init__(func=linearize, **kwargs)
+    def transform(self, X):
+        X = np.asarray(X)
+        return np.reshape(X, (X.shape[0], -1))
 
 
 class SampleLinearize(SampleMixin, Linearize):