{ "model_name": "/scratch/nicolasal97/gec_extractor/longformer-dapt", "num_tasks": 5, "num_classes": 3, "random_seed": 42, "label_columns": [ "modernizante", "estructuralista", "neoliberal", "posmoderna", "voluntarista" ], "eos_token_id": null, "pad_token_id": null, "bos_token_id": null, "focal_loss_config": { "gamma": 2.0, "alpha_per_task": { "modernizante": { "0": 1063000000.3333334, "1": 0.6399759183222958, "2": 0.6956806284904015 }, "estructuralista": { "0": 1.1504329004329004, "1": 0.917169974115617, "2": 0.961121157323689 }, "neoliberal": { "0": 1.2024886877828054, "1": 0.9779208831646734, "2": 0.8727422003284072 }, "posmoderna": { "0": 5.7771739130434785, "1": 0.4794767704104646, "2": 1.348984771573604 }, "voluntarista": { "0": 3.851449275362319, "1": 0.7072521623419827, "2": 0.7539007092198582 } } }, "class_weights_per_task": { "modernizante": { "0": 1063000000.3333334, "1": 0.6399759183222958, "2": 0.6956806284904015 }, "estructuralista": { "0": 1.1504329004329004, "1": 0.917169974115617, "2": 0.961121157323689 }, "neoliberal": { "0": 1.2024886877828054, "1": 0.9779208831646734, "2": 0.8727422003284072 }, "posmoderna": { "0": 5.7771739130434785, "1": 0.4794767704104646, "2": 1.348984771573604 }, "voluntarista": { "0": 3.851449275362319, "1": 0.7072521623419827, "2": 0.7539007092198582 } }, "learning_rate": 9.48119302842016e-05, "head_learning_rate": 0.0002, "batch_size": 8, "num_epochs": 10, "warmup_ratio": 0.13536166086530452, "weight_decay": 0.09772201133644716, "max_grad_norm": 1.9124671456336675, "dropout_rate": 0.10017460004284406, "swap_ratio": 0.15, "deletion_prob": 0.15, "cross_validate": false, "use_pre_trained": true, "n_folds": 5, "hyperparameter_tuning": true, "num_trials": 10, "results_dir": "./results/csv/multi_label", "model_path": "/scratch/nicolasal97/gec_extractor/results/models/multi_label", "input_file": "./data/csv/multi_label/etiquetado_perspectivas.csv", "output_dir": "/scratch/nicolasal97/multi_label/checkpoints/model_20251215_024620", "eval_metric": "accuracy", "type_augmentation": null, "number_samples": 1100, "target_samples_per_class": 50, "max_augmentation_per_sample": 10, "num_beams": 7, "temperature": 1.2, "use_mixed_precision": true, "gradient_accumulation_steps": 4, "early_stopping_patience": 3, "save_model": true, "automatic_augmentation": true }