Configuration

The process of assigning the initial property values used to train a model, including:

the model's composing layers

the location of the data

In machine learning projects, configuration can be done through a special configuration file or using configuration libraries such as the following:

Created for this library

1.
An ML platform team stores model configuration in version-controlled YAML files so any past training run can be reproduced exactly.
2.
A computer vision team encodes preprocessing, model, and optimizer configuration in a single config file so experiments are reproducible across team members.
3.
A startup's research team uses a hierarchical configuration system so a base configuration can be overridden per experiment without duplicating common settings.

The process of assigning the initial property values used to train a model, including:

the model's composing layers

the location of the data

In machine learning projects, configuration can be done through a special configuration file or using configuration libraries such as the following:

Created for this library

1.
An ML platform team stores model configuration in version-controlled YAML files so any past training run can be reproduced exactly.
2.
A computer vision team encodes preprocessing, model, and optimizer configuration in a single config file so experiments are reproducible across team members.
3.
A startup's research team uses a hierarchical configuration system so a base configuration can be overridden per experiment without duplicating common settings.

Loading…