| {"overwrite_output_dir": true, "predict_with_generate": true, "debug_mode": false, "append_verb_form": true, "use_bilateral_predicate_marker": true, "fp16": true, "load_best_model_at_end": true, "do_eval_on": "validation", "per_device_train_batch_size": 12, "per_device_eval_batch_size": 12, "save_strategy": "steps", "logging_strategy": "steps", "evaluation_strategy": "steps", "logging_steps": 500, "eval_steps": 500, "save_steps": 500, "metric_for_best_model": "eval_loss", "predicate_marker_type": "generic", "description": "qanom baseline with output_linearization=permutate_sample_num_of_qas, with best hyperparamters for 'by_answer_order' (30ep, accum=8) ", "model_type": "t5", "train_dataset": "qanom", "train_epochs": 30, "qanom_joint_factor": 1, "dir_switch": "qanom/linearization/permutate_sample_num_of_qas", "gradient_accumulation_steps": 8, "learning_rate": 0.001, "dropout_rate": 0.15, "seed": 44, "source_prefix": "parse: ", "preprocess_input_func": "input_predicate_marker", "preprocess_output_func": "permutate_sample_num_of_qas", "num_beams": 3} |