ShantanuT01/dactyl-deberta-v3-large-pretrained

{
        "training_split": "training",
        "evaluation_split": "testing",
        "results_path":"bce-debertav3-large.csv",
        "num_epochs": 1,
        "model_path": "best-debertav3-model",
        "tokenizer": "microsoft/deberta-v3-large",
        "optimizer": "AdamW",
        "optimizer_type": "torch",
        "optimizer_args": {
            "lr": 2e-5,
            "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-deberta-v3-large-pretrained",
        "training_args": {
            "batch_size": 16,
            "needs_sampler": false,
            "needs_index": false,
            "shuffle": true,
            "sampling_rate": null,
            "apply_sigmoid": false
        },
        "best_model_path": "best-debertav3-model",
        "eval_batch_size": 8
    
}