Configure MTL Experiment

Modified

August 16, 2025

This section will guide you to configure custom multi-task learning experiment (MTL).

Prepare Configs

The multi-task learning experiments works the same as continual learning experiments. You need to create YAML file in the experiment/ folder as well. Please refer to Configure CL Main Experiment section.

Usage of clarena train mtl

The command clarena train mtl locates the config folder configs/, parse the configuration of the specified multi-task learning experiment, and run the experiment:

clarena train mtl experiment=<experiment-name>

Please make sure the configs/ folder meeting the requirements above exists in the directory where you run the commands. The <experiment-name> is the name of the YAML file in the experiment/ subfolder. For example, if the YAML file mtl_scifar100_jointlearning.yaml is in experiment/mtl_train/ subfolder, the <experiment-name> is mtl_train/mtl_scifar100_jointlearning.

The MTL experiment configs follow the same logic of hierarchy and overriding as CL experiment configs. Please refer to Configure CL Main Experiment section. We jump straight to the required config fields, starting from the experiment index config. Please go to Experiment Index Config section.

Back to top