Logging

mlflow_extend.logging.log_params_flatten(params, parent_key='', sep='.')

Log a batch of params after flattening.

Parameters

• params (dict) – Dictionary of parameters to log.

• parent_key (str, default "") – Parent key.

• sep (str, default ".") – Key separator.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     params = {"a": {"b": 0}}
...     mlflow.log_params_flatten(params)
...     mlflow.log_params_flatten(params, parent_key="d")
...     mlflow.log_params_flatten(params, sep="_")
>>> r = mlflow.get_run(run.info.run_id)
>>> sorted(r.data.params.items())
[('a.b', '0'), ('a_b', '0'), ('d.a.b', '0')]

mlflow_extend.logging.log_metrics_flatten(metrics, step=None, parent_key='', sep='.')

Log a batch of metrics after flattening.

Parameters

• metrics (dict) – Dictionary of metrics to log.

• step (int, default None) – Metric step. Defaults to zero if unspecified.

• parent_key (str, default "") – Parent key.

• sep (str, default ".") – Key separator.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     metrics = {"a": {"b": 0.0}}
...     mlflow.log_metrics_flatten(metrics)
...     mlflow.log_metrics_flatten(metrics, parent_key="d")
...     mlflow.log_metrics_flatten(metrics, sep="_")
>>> r = mlflow.get_run(run.info.run_id)
>>> sorted(r.data.metrics.items())
[('a.b', 0.0), ('a_b', 0.0), ('d.a.b', 0.0)]

mlflow_extend.logging.log_plt_figure(fig, path)

Log a matplotlib figure as an artifact.

Parameters

• fig (matplotlib.pyplot.Figure) – Figure to log.

• path (str) – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     fig, ax = plt.subplots()
...     _ = ax.plot([0, 1], [0, 1])
...     mlflow.log_figure(fig, 'plt_figure.png')
>>> list_artifacts(run.info.run_id)
['plt_figure.png']

mlflow_extend.logging.log_plotly_figure(fig, path)

Log a plotly figure as an artifact.

Parameters

• fig (go.Figure) – Figure to log.

• path (str) – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     fig = go.Figure(data=[go.Bar(x=[1, 2, 3], y=[1, 3, 2])])
...     mlflow.log_figure(fig, 'plotly_figure.html')  # Must be an HTML file.
>>> list_artifacts(run.info.run_id)
['plotly_figure.html']

mlflow_extend.logging.log_figure(fig, path)

Log a matplotlib figure as an artifact.

Parameters

• fig (matplotlib.pyplot.Figure or plotly.graph_objects.Figure) – Figure to log.

• path (str) – Path in the artifact store.

Returns

None

Return type

None

Examples

Matplotlib

>>> with mlflow.start_run() as run:
...     fig, ax = plt.subplots()
...     _ = ax.plot([0, 1], [0, 1])
...     mlflow.log_figure(fig, 'plt_figure.png')
>>> list_artifacts(run.info.run_id)
['plt_figure.png']

Plotly

>>> with mlflow.start_run() as run:
...     fig = go.Figure(data=[go.Bar(x=[1, 2, 3], y=[1, 3, 2])])
...     mlflow.log_figure(fig, 'plotly_figure.html')  # Must be an HTML file.
>>> list_artifacts(run.info.run_id)
['plotly_figure.html']

mlflow_extend.logging.log_dict(dct, path, fmt=None)

Log a dictionary as an artifact.

Parameters

• dct (dict) – Dictionary to log.

• path (str) – Path in the artifact store.

• fmt (str, default None) – File format to save dict in. If None, file format is inferred from path.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     d = {'a': 0}
...     mlflow.log_dict(d, 'dict.json')
...     mlflow.log_dict(d, 'dict.yaml')
...     mlflow.log_dict(d, 'dict.yml')
>>> list_artifacts(run.info.run_id)
['dict.json', 'dict.yaml', 'dict.yml']

mlflow_extend.logging.log_df(df, path, fmt='csv')

Log a dataframe as an artifact.

Parameters

• df (pandas.DataFrame) – Dataframe to log.

• path (str) – Path in the artifact store.

• fmt (str, default "csv") – File format to save the dataframe in.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     mlflow.log_df(pd.DataFrame({'a': [0]}), 'df.csv')
>>> list_artifacts(run.info.run_id)
['df.csv']

mlflow_extend.logging.log_text(text, path)

Log a text as an artifact.

Parameters

• text (str) – Text to log.

• path (str) – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     mlflow.log_text('text', 'text.txt')
>>> list_artifacts(run.info.run_id)
['text.txt']

mlflow_extend.logging.log_numpy(arr, path)

Log a numpy array as an artifact.

Parameters

• arr (numpy.ndarray) – Numpy array to log.

• path (str) – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     mlflow.log_numpy(np.array([0]), 'array.npy')
>>> list_artifacts(run.info.run_id)
['array.npy']

mlflow_extend.logging.log_confusion_matrix(cm, path='confusion_matrix.png')

Log a confusion matrix as an artifact.

Parameters

• cm (array-like) – Confusion matrix to log.

• path (str, default "confusion_matrix.png") – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     mlflow.log_confusion_matrix([[1, 2], [3, 4]])
>>> list_artifacts(run.info.run_id)
['confusion_matrix.png']

mlflow_extend.logging.log_feature_importance(features, importances, importance_type, limit=None, normalize=False, path='feature_importance.png')

Log feature importance as an artifact.

Parameters

• features (array-like) – Feature names.

• importances (array-like) – Importance of each feature.

• importance_type (str) – Importance type (e.g. “gain”).

• path (str, default "feature_importance.png") – Path in the artifact store.

• **kwargs (dict) – Keyword arguments passed to mlflow.plotting.feature_importance.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     features = ['a', 'b', 'c']
...     importances = [1, 2, 3]
...     mlflow.log_feature_importance(features, importances, 'gain')
>>> list_artifacts(run.info.run_id)
['feature_importance.png']

mlflow_extend.logging.log_roc_curve(fpr, tpr, auc=None, path='roc_curve.png')

Log ROC curve as an artifact.

Parameters

• fpr (array-like) – False positive rate.

• tpr (array-like) – True positive rate.

• auc (float, default None) – Area under the curve.

• path (str, default "roc_curve.png") – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     mlflow.log_roc_curve([0, 1], [0, 1])
>>> list_artifacts(run.info.run_id)
['roc_curve.png']

mlflow_extend.logging.log_pr_curve(pre, rec, auc=None, path='pr_curve.png')

Log precision-recall curve as an artifact.

Parameters

• pre (array-like) – Precision.

• rec (array-like) – Recall.

• auc (float, default None) – Area under the curve.

• path (str, default "pr_curve.png") – Path in the artifact store.

Returns

None

Return type

None

Examples

>>> with mlflow.start_run() as run:
...     mlflow.log_pr_curve([1, 0], [1, 0])
>>> list_artifacts(run.info.run_id)
['pr_curve.png']