Run from the project root, passing the filename of a main configuration file (as found directly under conf/):
python train.py --conf multimodal_MT_freeze.ymlAdd --kfold N to train/evaluate on a specific fold (requires the fold's CSVs to already exist, see 3.2). Training automatically resumes from the latest matching checkpoint in checkpoint_folder, if any, and runs the test phase (on the best checkpoint by validation loss) right after training completes.
Previous: 4. Configuration files · Next: 6. Test