ShantanuT01/dactyl-modernbert-base-pretrained

{
    "training_split": "training",
    "evaluation_split": "testing",
    "results_path": "bce-pretraining-modernbert.csv",
    "num_epochs": 5,
    "model_path": "answerdotai/ModernBERT-base",
    "tokenizer": "answerdotai/ModernBERT-base",
    "optimizer": "AdamW",
    "optimizer_type": "torch",
    "optimizer_args": {
        "lr": 2e-05,
        "weight_decay": 0.01
    },
    "loss_fn": "BCEWithLogitsLoss",
    "reset_classification_head": false,
    "loss_type": "torch",
    "loss_fn_args": {},
    "needs_loss_fn_as_parameter": false,
    "save_path": "ShantanuT01/dactyl-modernbert-base-pretrained",
    "training_args": {
        "batch_size": 64,
        "needs_sampler": false,
        "needs_index": false,
        "shuffle": true,
        "sampling_rate": null,
        "apply_sigmoid": false
    }
}