Skip to content

Vectice

Vectice package.

The Vectice package is a library allowing data-scientists to record their progress in the Vectice app.

This package exposes essential Vectice classes and methods:

  • the autolog method
  • the connect method
  • the Workspace class
  • the Project class
  • the Phase class
  • the Iteration class
  • the Dataset class
  • the Model class

  • the Table class

    IMPORTANT INFORMATION

    Vectice calculates dataframe statistics only if the dataframe contains more than a hundred rows for privacy reasons. Those statistics are calculate on a sample of a million rows and on the first 400 columns by default. All of these values can be changed on the Organization Settings page by the organization admin.

code_capture module-attribute 

code_capture = True

Global code capture flag, enabled by default.

Code capture is triggered when registering a dataset or a model, and only works when a valid Git repository is found. Otherwise a warning is logged, telling what might be misconfigured in the repository.

Captured information include the repository name, URL, branch name, commit hash, and whether the repository is dirty (has uncommitted changes).

Examples:

To disable code capture globally:

>>> import vectice
>>> vectice.code_capture = False

To re-enable code capture globally:

>>> import vectice
>>> vectice.code_capture = True

code_file_capture module-attribute 

code_file_capture = False

Global code file capture flag, disabled by default.

Controls whether code file is captured when logging asset into Vectice as part of the lineage. When enabled (set to True), logging an asset will also add the executed file to its lineage. Enabling this feature may increase API runtime due to file transfer. Databricks is currently not supported.

Examples:

To enable code_file_capture globally:

>>> import vectice
>>> vectice.code_file_capture = True

To re-enable code file capture globally:

>>> import vectice
>>> vectice.code_file_capture = False

auto_extract module-attribute 

auto_extract = True

Global auto extraction flag, enabled by default.

Extraction is automatically performed when registering a dataset or a model with an attachment. Currently, it exclusively operates on Excel files and extracts sheets (as CSV files) and images to be referenced inside Vectice.

Examples:

To disable auto extraction of files globally globally:

>>> import vectice
>>> vectice.auto_extract = False

To re-enable auto extraction globally:

>>> import vectice
>>> vectice.auto_extract = True

pickle_capture module-attribute 

pickle_capture = True

Global pickle capture flag, enabled by default.

Pickle capture is triggered when registering a Vectice Model and only works when a valid predictor parameter is passed or when Autolog detects an estimator, for example a scikit-learn regressor. The predictor is pickled and attached to the Model Version in Vectice.

Examples:

To disable pickle capture globally:

>>> import vectice
>>> vectice.pickle_capture = False

To re-enable pickle capture globally:

>>> import vectice
>>> vectice.pickle_capture = True

Standard API

>>> from vectice import Model
...
>>> my_estimator = LinearRegression()
>>> model = Model(predictor=my_estimator)

Autolog

>>> from vectice import autolog
...
>>> my_estimator = LinearRegression()
>>> autolog.cell()