Skip to main content

Dataset Management

Training data is the data used to train a model or algorithm. This training data can be uploaded to and managed in navio as datasets. The training dataset can either be uploaded together with the model (packaged in the model archive) or uploaded separately. Datasets are shared across an entire workspace.

A user can upload datasets separately to navio in two ways via the user interface on the Datasets page:

  1. Upload a dataset as a CSV file
  2. Import data from a relational database via a JDBC connection

upload datasets

Dataset Format​

Each column must contain exactly one feature. The target column is identified by its name, as defined in the model's request schema.

Upload as CSV File​

The training dataset CSV file can be at most 1GB in size. This limit is a configuration property of the navio instance and can be increased if required.

Import data via a Database​

Tables from a relational database can be imported as datasets to navio using a JDBC connection. Simply input your connection string and credentials and select the table you want to import. Keep in mind that the table needs to be in the correct format.

database import data