ML Tooling provides different backends for storing trained models. Currently, we support local file storage, as well as Artifactory based storage.
If you want to use Artifactory as a backend, first install the optional dependencies by running pip install ml_tooling['artifactory'].
>>> from ml_tooling.storage import ArtifactoryStorage
>>> from ml_tooling import Model
...
>>> artifactory_url = "http://artifactory.com/artifactory"
>>> artifactory_repo = "advanced-analytics/dev/myfolder"
>>>
>>> storage = ArtifactoryStorage(artifactory_url, artifactory_repo)
...
>>> estimators_in_myfolder = storage.get_list()
>>> estimators_in_myfolder
[ArtifactoryPath('http://artifactory.com/artifactory/advanced-analytics/dev/LinearRegression_2019-10-16_15:10:34.290209.pkl'), ArtifactoryPath('http://artifactory.com/artifactory/advanced-analytics/dev/LinearRegression_2019-10-16_15:14:02.114818.pkl')]
>>> my_estimator = storage.load(estimators_in_myfolder[0])
>>> my_estimator
LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False)
...
>>> from ml_tooling import Model
>>> model = Model(my_estimator)
>>> model
<Model: LinearRegression>
>>> # ...
>>> # make some adjustments to the model
...
>>> # now save the estimator, ideally using the Model instance
>>> model.save_estimator(storage)
[14:34:08] - Saved estimator to http://artifactory-singlep.com/artifactory/advanced-analytics/dev/LinearRegression_2019-10-30_14:34:08.116648.pkl
ArtifactoryPath('http://artifactory.com/artifactory/advanced-analytics/dev/LinearRegression_2019-10-30_14:34:08.116648.pkl')
>>> # if you want more control over folders/name destinations use the Storage directly
>>> new_estimator_path = "my_new_folder/LinearRegression.pkl"
>>> storage.save(model.estimator, new_estimator_path)
ArtifactoryPath('http://artifactory.com/artifactory/advanced-analytics/dev/my_new_folder/LinearRegression.pkl')For more information see ~ml_tooling.storage.ArtifactoryStorage.
>>> from ml_tooling.storage import FileStorage
>>> from ml_tooling import Model
...
>>> my_folder = "./estimators"
...
>>> storage = FileStorage(my_folder)
...
>>> estimators_in_myfolder = storage.get_list()
>>> my_estimator = storage.load(estimators_in_myfolder[0])
...
>>> new_model_name = "my_new_folder/LinearRegression.pkl"
>>> storage.save(model.estimator)
PosixPath('estimators/my_new_folder/LinearRegression.pkl')See ~ml_tooling.storage.FileStorage for more information
Continue to plotting