turintech-evoml-client

Python client library to support script and notebook interactions with the platform

Installation

There are two options to install the client library:

Default installation

Installs basic requirements and versions of libraries like pandas, numpy.

pip install turintech_evoml_client

Full installation

If you want to be able to generate models locally make sure you use the full extra when installing evoml_client.

pip install turintech_evoml_client[full]

Note that if you are using zsh, you might need to use one the following command instead:

pip install turintech_evoml_client\[full\]
pip install "turintech_evoml_client[full]"
noglob pip install turintech_evoml_client[full]

Description

Currently supports the following concepts and actions:

Datasets:

• Creation - EvoML client supports: Pandas DataFrames, numpy arrays and csv files. With Additional support to come.
• Uploading - Send the locally defined Dataset to the platform in preparation for future trials.
• Statistical analysis - Using the Analyser class retrieve statistical info on the uploaded dataset.

Trials:

• Configuration - Multiple levels of abstraction provided by the TrialConfig class to allow specification of as much or as little technical details as desired.
• Uploading - Send the locally defined trial to the platform to await execution.
• Execution - Trigger the optimization of the given trial.
• Result retrieval - Get either all models or the best model for a trial.
• local execution - Once retrieved and built, models can be run locally against new data.

AutoML:

• Given a Pandas Dataframe, Numpy array or EvoML dataset and a target column returns the best model trained to predict the given target column.

You can find further information and examples in our documentation page: https://docs.evoml.ai/evoml-client/introduction