Dataset description file

In order to add datasets to the database, you can use the command ‘qanat dataset new’. If no arguments is provided, a prompt will be shown asking the informations about your dataset. Alternatively, you can provide a YAML file containing the description of your dataset:

qanat dataset new -f <yaml_file>

The YAML file must contain the following fields:

• name: string

  The name of the dataset. This name will be used to identify the dataset in the database.

• description: string

  A short description of the dataset.

• path: string

  The path to the dataset (from the Qanat projet root). This path must be accessible from the machine running the a Qanat experiment.

• tags: list of strings

  A list of tags associated to the dataset. Tags are used to filter the datasets in the database.

Example

name: mnist
description: MNIST dataset
path: data/mnist
tags: [mnist, classification, images]