{
    "model": "dBert",
    "dataset": "optimized_strict_heteronyms",
    "experiment": "train-dBert",
    "run": "test",
    "num_train_epochs": 10,
    "evaluation_strategy": "steps",
    "eval_steps": 300,
    "logging_strategy": "steps",
    "logging_steps": 300,
    "save_strategy": "steps",
    "save_steps": 300,
    "learning_rate": 2e-5,
    "per_device_train_batch_size": 128,
    "per_device_eval_batch_size": 128,
    "load_best_model_at_end": true,
    "metric_for_best_model": "loss",
    "weight_decay": 0.01,
    "save_total_limit": 5,
    "report_to": "mlflow"
}