Skip to content

Pipeline

Pipelines, variances to the sklearn.pipeline.Pipeline object.

sklego.pipeline.DebugPipeline

Bases: Pipeline

A pipeline that has a log statement in between each step, useful for debugging purposes.

See sklearn.pipeline.Pipeline for all other parameters other than log_callback.

Note

This implementation is a hack on the original sklearn Pipeline. It aims to have the same behaviour as the original sklearn Pipeline, while changing minimal amount of code.

The log statement is added by overwriting the cache method of the memory, such that the function called in the cache is wrapped with a functions that calls the log callback function (log_callback).

This hack will break when:

  • The sklearn pipeline initialization function is changed.
  • The memory is used differently in the fit.
  • The joblib.memory.Memory changes behaviour of the cache method.
  • The joblib.memory.Memory starts using a _cache method.

Parameters:

Name Type Description Default
log_callback Callable | None

The callback function that logs information in between each intermediate step. If set to "default", default_log_callback is used.

 None

Examples:

# Set-up for example
import logging
import sys

import numpy as np
import pandas as pd

from sklearn.base import BaseEstimator, TransformerMixin
from sklego.pipeline import DebugPipeline

logging.basicConfig(
    format=("[%(funcName)s:%(lineno)d] - %(message)s"),
    level=logging.INFO,
    stream=sys.stdout,
    )

# DebugPipeline set-up
n_samples, n_features = 3, 5
X = np.zeros((n_samples, n_features))
y = np.arange(n_samples)

class Adder(TransformerMixin, BaseEstimator):
    def __init__(self, value):
        self._value = value

    def fit(self, X, y=None):
        return self

    def transform(self, X):
        return X + self._value

    def __repr__(self):
        return f"Adder(value={self._value})"

steps = [
    ("add_1", Adder(value=1)),
    ("add_10", Adder(value=10)),
    ("add_100", Adder(value=100)),
    ("add_1000", Adder(value=1000)),
]

# The DebugPipeline behaves the sames as the Sklearn pipeline.
pipe = DebugPipeline(steps)

_ = pipe.fit(X, y=y)
print(pipe.transform(X))
# [[1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]]

# But it has the option to set a `log_callback`, that logs in between each step.
pipe = DebugPipeline(steps, log_callback="default")

_ = pipe.fit(X, y=y)
# [default_log_callback:34] - [Adder(value=1)] shape=(3, 5) time=0s
# [default_log_callback:34] - [Adder(value=10)] shape=(3, 5) time=0s
# [default_log_callback:34] - [Adder(value=100)] shape=(3, 5) time=0s

print(pipe.transform(X))
# [[1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]]

# It is possible to set the `log_callback` later then initialisation.
pipe = DebugPipeline(steps)
pipe.log_callback = "default"

_ = pipe.fit(X, y=y)
# [default_log_callback:34] - [Adder(value=1)] shape=(3, 5) time=0s
# [default_log_callback:34] - [Adder(value=10)] shape=(3, 5) time=0s
# [default_log_callback:34] - [Adder(value=100)] shape=(3, 5) time=0s

print(pipe.transform(X))
# [[1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]]

# It is possible to define your own `log_callback` function.
def log_callback(output, execution_time, **kwargs):
    '''My custom `log_callback` function

    Parameters
    output : tuple[np.ndarray | pd.DataFrame, estimator | transformer]
        The output of the step and a step in the pipeline.
    execution_time : float
        The execution time of the step.

    Note
    The **kwargs are for arguments that are not used in this callback.
    '''
    logger = logging.getLogger(__name__)
    step_result, step = output
    logger.info(
        f"[{step}] shape={step_result.shape} "
        f"nbytes={step_result.nbytes} time={int(execution_time)}s")

pipe.log_callback = log_callback

_ = pipe.fit(X, y=y)
# [log_callback:20] - [Adder(value=1)] shape=(3, 5) nbytes=120 time=0s
# [log_callback:20] - [Adder(value=10)] shape=(3, 5) nbytes=120 time=0s
# [log_callback:20] - [Adder(value=100)] shape=(3, 5) nbytes=120 time=0s

print(pipe.transform(X))
# [[1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]]

# Remove the `log_callback` when you want to stop logging.
pipe.log_callback = None

_ = pipe.fit(X, y=y)
print(pipe.transform(X))
# [[1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]
#  [1111. 1111. 1111. 1111. 1111.]]

# Logging also works with FeatureUnion
from sklearn.pipeline import FeatureUnion
pipe_w_default_log_callback = DebugPipeline(steps, log_callback="default")
pipe_w_custom_log_callback = DebugPipeline(steps, log_callback=log_callback)

pipe_union = DebugPipeline([
    ("feature_union", FeatureUnion([
        ("pipe_w_default_log_callback", pipe_w_default_log_callback),
        ("pipe_w_custom_log_callback", pipe_w_custom_log_callback),
    ])),
    ("final_adder", Adder(10000))
], log_callback="default")

_ = pipe_union.fit(X, y=y)   # doctest:+ELLIPSIS
# [default_log_callback:34] - [Adder(value=1)] shape=(3, 5) time=0s
# [default_log_callback:34] - [Adder(value=10)] shape=(3, 5) time=0s
# [default_log_callback:34] - [Adder(value=100)] shape=(3, 5) time=0s
# [log_callback:20] - [Adder(value=1)] shape=(3, 5) nbytes=120 time=0s
# [log_callback:20] - [Adder(value=10)] shape=(3, 5) nbytes=120 time=0s
# [log_callback:20] - [Adder(value=100)] shape=(3, 5) nbytes=120 time=0s
# [default_log_callback:34] - [FeatureUnion(...)] shape=(3, 10) time=0s

print(pipe_union.transform(X))
# [[11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111.]
#  [11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111.]
#  [11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111.]]
Source code in sklego/pipeline.py 
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
class DebugPipeline(Pipeline):
    """A pipeline that has a log statement in between each step, useful for debugging purposes.

    See [`sklearn.pipeline.Pipeline`](https://scikit-learn.org/stable/modules/generated/sklearn.pipeline.Pipeline.html#sklearn.pipeline.Pipeline)
    for all other parameters other than `log_callback`.

    !!! note

        This implementation is a hack on the original sklearn Pipeline. It aims to have the same behaviour as the
        original sklearn Pipeline, while changing minimal amount of code.

        The log statement is added by overwriting the cache method of the memory, such that the function called in the
        cache is wrapped with a functions that calls the log callback function (`log_callback`).

        This hack will break when:

        - The sklearn pipeline initialization function is changed.
        - The memory is used differently in the fit.
        - The [`joblib.memory.Memory`](https://joblib.readthedocs.io/en/latest/generated/joblib.Memory.html)
            changes behaviour of the `cache` method.
        - The [`joblib.memory.Memory`](https://joblib.readthedocs.io/en/latest/generated/joblib.Memory.html)
            starts using a `_cache` method.

    Parameters
    ----------
    log_callback : Callable | None, default=None
        The callback function that logs information in between each intermediate step.
        If set to `"default"`, `default_log_callback` is used.

    Examples
    --------
    ```py
    # Set-up for example
    import logging
    import sys

    import numpy as np
    import pandas as pd

    from sklearn.base import BaseEstimator, TransformerMixin
    from sklego.pipeline import DebugPipeline

    logging.basicConfig(
        format=("[%(funcName)s:%(lineno)d] - %(message)s"),
        level=logging.INFO,
        stream=sys.stdout,
        )

    # DebugPipeline set-up
    n_samples, n_features = 3, 5
    X = np.zeros((n_samples, n_features))
    y = np.arange(n_samples)

    class Adder(TransformerMixin, BaseEstimator):
        def __init__(self, value):
            self._value = value

        def fit(self, X, y=None):
            return self

        def transform(self, X):
            return X + self._value

        def __repr__(self):
            return f"Adder(value={self._value})"

    steps = [
        ("add_1", Adder(value=1)),
        ("add_10", Adder(value=10)),
        ("add_100", Adder(value=100)),
        ("add_1000", Adder(value=1000)),
    ]

    # The DebugPipeline behaves the sames as the Sklearn pipeline.
    pipe = DebugPipeline(steps)

    _ = pipe.fit(X, y=y)
    print(pipe.transform(X))
    # [[1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]]

    # But it has the option to set a `log_callback`, that logs in between each step.
    pipe = DebugPipeline(steps, log_callback="default")

    _ = pipe.fit(X, y=y)
    # [default_log_callback:34] - [Adder(value=1)] shape=(3, 5) time=0s
    # [default_log_callback:34] - [Adder(value=10)] shape=(3, 5) time=0s
    # [default_log_callback:34] - [Adder(value=100)] shape=(3, 5) time=0s

    print(pipe.transform(X))
    # [[1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]]

    # It is possible to set the `log_callback` later then initialisation.
    pipe = DebugPipeline(steps)
    pipe.log_callback = "default"

    _ = pipe.fit(X, y=y)
    # [default_log_callback:34] - [Adder(value=1)] shape=(3, 5) time=0s
    # [default_log_callback:34] - [Adder(value=10)] shape=(3, 5) time=0s
    # [default_log_callback:34] - [Adder(value=100)] shape=(3, 5) time=0s

    print(pipe.transform(X))
    # [[1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]]

    # It is possible to define your own `log_callback` function.
    def log_callback(output, execution_time, **kwargs):
        '''My custom `log_callback` function

        Parameters
        output : tuple[np.ndarray | pd.DataFrame, estimator | transformer]
            The output of the step and a step in the pipeline.
        execution_time : float
            The execution time of the step.

        Note
        The **kwargs are for arguments that are not used in this callback.
        '''
        logger = logging.getLogger(__name__)
        step_result, step = output
        logger.info(
            f"[{step}] shape={step_result.shape} "
            f"nbytes={step_result.nbytes} time={int(execution_time)}s")

    pipe.log_callback = log_callback

    _ = pipe.fit(X, y=y)
    # [log_callback:20] - [Adder(value=1)] shape=(3, 5) nbytes=120 time=0s
    # [log_callback:20] - [Adder(value=10)] shape=(3, 5) nbytes=120 time=0s
    # [log_callback:20] - [Adder(value=100)] shape=(3, 5) nbytes=120 time=0s

    print(pipe.transform(X))
    # [[1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]]

    # Remove the `log_callback` when you want to stop logging.
    pipe.log_callback = None

    _ = pipe.fit(X, y=y)
    print(pipe.transform(X))
    # [[1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]
    #  [1111. 1111. 1111. 1111. 1111.]]

    # Logging also works with FeatureUnion
    from sklearn.pipeline import FeatureUnion
    pipe_w_default_log_callback = DebugPipeline(steps, log_callback="default")
    pipe_w_custom_log_callback = DebugPipeline(steps, log_callback=log_callback)

    pipe_union = DebugPipeline([
        ("feature_union", FeatureUnion([
            ("pipe_w_default_log_callback", pipe_w_default_log_callback),
            ("pipe_w_custom_log_callback", pipe_w_custom_log_callback),
        ])),
        ("final_adder", Adder(10000))
    ], log_callback="default")

    _ = pipe_union.fit(X, y=y)   # doctest:+ELLIPSIS
    # [default_log_callback:34] - [Adder(value=1)] shape=(3, 5) time=0s
    # [default_log_callback:34] - [Adder(value=10)] shape=(3, 5) time=0s
    # [default_log_callback:34] - [Adder(value=100)] shape=(3, 5) time=0s
    # [log_callback:20] - [Adder(value=1)] shape=(3, 5) nbytes=120 time=0s
    # [log_callback:20] - [Adder(value=10)] shape=(3, 5) nbytes=120 time=0s
    # [log_callback:20] - [Adder(value=100)] shape=(3, 5) nbytes=120 time=0s
    # [default_log_callback:34] - [FeatureUnion(...)] shape=(3, 10) time=0s

    print(pipe_union.transform(X))
    # [[11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111.]
    #  [11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111.]
    #  [11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111. 11111.]]
    ```
    """

    def __init__(self, steps, memory=None, verbose=False, *, log_callback=None):
        self.log_callback = log_callback
        super().__init__(steps=steps, memory=memory, verbose=verbose)

    @property
    def memory(self):
        # When no log callback function is given, change nothing.
        # Or, if the memory cache was changed, set it back to its original.
        if self._log_callback is None:
            if hasattr(self._memory, "_cache"):
                self._memory.cache = self._memory._cache
            return self._memory

        self._memory = check_memory(self._memory)

        # Overwrite cache function of memory such that it logs the
        # output when the function is called
        if not hasattr(self._memory, "_cache"):
            self._memory._cache = self._memory.cache
        self._memory.cache = _cache_with_function_log_statement(self._log_callback).__get__(
            self._memory, self._memory.__class__
        )
        return self._memory

    @memory.setter
    def memory(self, memory):
        self._memory = memory

    @property
    def log_callback(self):
        return self._log_callback

    @log_callback.setter
    def log_callback(self, func):
        self._log_callback = func
        if self._log_callback == "default":
            self._log_callback = default_log_callback

sklego.pipeline.make_debug_pipeline(*steps, **kwargs)

Construct a DebugPipeline from the given estimators.

This is a shorthand for the DebugPipeline constructor; it does not require, and does not permit, naming the estimators. Instead, their names will be set to the lowercase of their types automatically.

Parameters:

Name Type Description Default
*steps list

List of estimators to be included in the pipeline.

 ()
**kwargs dict

Additional keyword arguments passed to the DebugPipeline constructor. Possible arguments are memory, verbose and log_callback:

  • memory : str | object with the joblib.Memory interface, default=None

    Used to cache the fitted transformers of the pipeline. The last step will never be cached, even if it is a transformer. By default, no caching is performed. If a string is given, it is the path to the caching directory. Enabling caching triggers a clone of the transformers before fitting. Therefore, the transformer instance given to the pipeline cannot be inspected directly. Use the attribute named_steps or steps to inspect estimators within the pipeline. Caching the transformers is advantageous when fitting is time consuming.

  • verbose : bool, default=False

    If True, the time elapsed while fitting each step will be printed as it is completed.

  • log_callback : str | Callable | None, default=None.

    The callback function that logs information in between each intermediate step. If set to "default", default_log_callback is used.

 {}

Returns:

Type Description
DebugPipeline

Instance with given steps, memory, verbose and log_callback.

Examples:

from sklearn.naive_bayes import GaussianNB
from sklearn.preprocessing import StandardScaler

make_debug_pipeline(StandardScaler(), GaussianNB(priors=None))
# DebugPipeline(steps=[("standardscaler", StandardScaler()),
#                 ("gaussiannb", GaussianNB())])
See Also

sklego.pipeline.DebugPipeline : Class for creating a pipeline of transforms with a final estimator.

Source code in sklego/pipeline.py 
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
def make_debug_pipeline(*steps, **kwargs):
    """Construct a `DebugPipeline` from the given estimators.

    This is a shorthand for the `DebugPipeline` constructor; it does not require, and does not permit, naming the
    estimators. Instead, their names will be set to the lowercase of their types automatically.

    Parameters
    ----------
    *steps : list
        List of estimators to be included in the pipeline.
    **kwargs : dict
        Additional keyword arguments passed to the `DebugPipeline` constructor.
        Possible arguments are `memory`, `verbose` and `log_callback`:

        - `memory` : str | object with the joblib.Memory interface, default=None

            Used to cache the fitted transformers of the pipeline. The last step will never be cached, even if it is a
            transformer. By default, no caching is performed. If a string is given, it is the path to the caching
            directory. Enabling caching triggers a clone of the transformers before fitting. Therefore, the transformer
            instance given to the pipeline cannot be inspected directly. Use the attribute `named_steps` or `steps` to
            inspect estimators within the pipeline. Caching the transformers is advantageous when fitting is time
            consuming.

        - `verbose` : bool, default=False

            If True, the time elapsed while fitting each step will be printed as it is completed.

        - `log_callback` : str | Callable | None, default=None.

            The callback function that logs information in between each intermediate step. If set to `"default"`,
            `default_log_callback` is used.

    Returns
    -------
    DebugPipeline
        Instance with given steps, `memory`, `verbose` and `log_callback`.

    Examples
    --------
    ```py
    from sklearn.naive_bayes import GaussianNB
    from sklearn.preprocessing import StandardScaler

    make_debug_pipeline(StandardScaler(), GaussianNB(priors=None))
    # DebugPipeline(steps=[("standardscaler", StandardScaler()),
    #                 ("gaussiannb", GaussianNB())])
    ```

    See Also
    --------
    sklego.pipeline.DebugPipeline : Class for creating a pipeline of transforms with a final estimator.
    """
    memory = kwargs.pop("memory", None)
    verbose = kwargs.pop("verbose", False)
    log_callback = kwargs.pop("log_callback", None)
    if kwargs:
        raise TypeError('Unknown keyword arguments: "{}"'.format(list(kwargs.keys())[0]))
    return DebugPipeline(
        _name_estimators(steps),
        memory=memory,
        verbose=verbose,
        log_callback=log_callback,
    )

sklego.pipeline.default_log_callback(output, execution_time, **kwargs)

The default log callback which logs the step name, shape of the output and the execution time of the step.

Info

If you write your custom callback function the input is:

Parameter Type Description
func Callable[..., T] The function to be wrapped
input_args tuple The input arguments
input_kwargs dict The input key-word arguments
output T The output of the function
execution_time float The execution time of the step

Parameters:

Name Type Description Default
output tuple[ndarray | DataFrame, estimator | transformer]

The output of the step and a step in the pipeline.

 required
execution_time float

The execution time of the step.

 required
Source code in sklego/pipeline.py 
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
def default_log_callback(output, execution_time, **kwargs):
    """The default log callback which logs the step name, shape of the output and the execution time of the step.

    !!! info

        If you write your custom callback function the input is:

        | Parameter        | Type             | Description                    |
        | ---------------- | ---------------- | ------------------------------ |
        | `func`           | Callable[..., T] | The function to be wrapped     |
        | `input_args`     | tuple            | The input arguments            |
        | `input_kwargs`   | dict             | The input key-word arguments   |
        | `output`         | T                | The output of the function     |
        | `execution_time` | float            | The execution time of the step |

    Parameters
    ----------
    output : tuple[np.ndarray | pd.DataFrame, estimator | transformer]
        The output of the step and a step in the pipeline.
    execution_time : float
        The execution time of the step.
    """
    logger = logging.getLogger(__name__)
    step_result, step = output
    logger.info(f"[{step}] shape={step_result.shape} " f"time={int(execution_time)}s")