Add DatasetPipeline to work on xarray datasets

Fixes #15 (closed)

bob/pipelines/tests/test_wrappers.py

@@ -49,16 +49,10 @@ class DummyTransformer(TransformerMixin, BaseEstimator):
     https://github.com/scikit-learn-contrib/project-
     template/blob/master/skltemplate/_template.py."""
 
-    def __init__(self, picklable=True, i=None, **kwargs):
+    def __init__(self, i=None, **kwargs):
         super().__init__(**kwargs)
-        self.picklable = picklable
         self.i = i
 
-        if not picklable:
-            import bob.core
-
-            self.rng = bob.core.random.mt19937()
-
     def fit(self, X, y=None):
         return self
 
@@ -221,7 +215,7 @@ def _build_estimator(path, i):
     return transformer
 
 
-def _build_transformer(path, i, picklable=True):
+def _build_transformer(path, i):
 
     features_dir = os.path.join(path, f"transformer{i}")
     estimator = mario.wrap(
@@ -288,7 +282,9 @@ def test_checkpoint_fit_transform_pipeline():
             transformer = ("1", _build_transformer(d, 1))
             pipeline = Pipeline([fitter, transformer])
             if dask_enabled:
-                pipeline = mario.wrap(["dask"], pipeline, fit_tag=[(1, "GPU")], npartitions=1)
+                pipeline = mario.wrap(
+                    ["dask"], pipeline, fit_tag=[(1, "GPU")], npartitions=1
+                )
                 pipeline = pipeline.fit(samples)
                 tags = mario.dask_tags(pipeline)
 
@@ -331,7 +327,7 @@ def test_checkpoint_fit_transform_pipeline_with_dask_non_pickle():
             fitter = ("0", _build_estimator(d, 0))
             transformer = (
                 "1",
-                _build_transformer(d, 1, picklable=False),
+                _build_transformer(d, 1),
             )
 
             pipeline = Pipeline([fitter, transformer])