run_config.json

{
    "run_name": "docket-duration",
    "base_model": "answerdotai/ModernBERT-large",
    "push_to_hub": "nadahlberg/docket-duration",
    "run_args": {
        "max_length": 768,
        "labels": [
            "DURATION",
            "IMPRISONMENT_DURATION",
            "SUPERVISED_RELEASE_DURATION"
        ],
        "return_offsets_mapping": true
    },
    "training_args": {
        "num_train_epochs": 2,
        "per_device_train_batch_size": 4,
        "per_device_eval_batch_size": 4,
        "gradient_accumulation_steps": 4,
        "learning_rate": 5e-05,
        "warmup_ratio": 0.05,
        "weight_decay": 0.1,
        "max_grad_norm": 1,
        "save_total_limit": 2,
        "load_best_model_at_end": true,
        "eval_steps": 0.1,
        "save_steps": 0.1,
        "eval_strategy": "steps",
        "report_to": "tensorboard"
    },
    "model_args": {},
    "routine": "multi-task"
}