Configuration files

In order to make the framework easy to understand and allow the users to tune the parameters, the software uses YAML file to store customizable variables to be used during training or even for other operations which may require user-defined parameters. In this page, we list all the parameters that can be used:

Training files

There are 4 main top-level entries that are required in the configuration file when doing training:

• trainer : This specifies the type of training to be used (at the moment only generic type is available).

• output_dir : Where to store the results of the training.

• data_config : Information regarding dataset to be used.

• model_config : Information regarding the neural network model to be used.

• train_config : Details of the training.