checkpoints/epoch_11/training_state.json

{"epoch": 11, "average_batch_loss": 12.599024105072022}