| {"dataset_config_name": ["en"], "delta_type": "compacter", "do_eval": true, "do_test": true, "do_train": true, "eval_dataset_config_name": ["en"], "eval_dataset_name": "mnli", "eval_steps": 200, "evaluation_strategy": "steps", "factorized_phm": true, "factorized_phm_rule": false, "gradient_clip": false, "greater_is_better": true, "hypercomplex_adapters": true, "hypercomplex_division": 4, "hypercomplex_nonlinearity": "glorot-uniform", "learn_phm": true, "learning_rate": 0.0003, "load_best_model_at_end": true, "max_source_length": 128, "metric_for_best_model": "average_metrics", "model_name_or_path": "../../../../plm_cache/t5-base", "non_linearity": "gelu_new", "normalize_phm_weight": false, "num_train_epochs": 3, "output_dir": "outputs/lora/t5-base/mnli", "overwrite_output_dir": true, "per_device_eval_batch_size": 32, "per_device_train_batch_size": 32, "phm_c_init": "normal", "phm_clamp": false, "phm_init_range": 0.0001, "predict_with_generate": true, "push_to_hub": true, "save_steps": 200, "save_strategy": "steps", "save_total_limit": 1, "seed": 42, "shared_phm_rule": false, "split_validation_test": true, "task_name": "mnli", "test_dataset_config_name": ["en"], "test_dataset_name": "mnli", "tokenizer_name": "../../../../plm_cache/t5-base", "unfrozen_modules": ["deltas", "layer_norm", "final_layer_norm"], "use_bias_down_sampler": true, "use_bias_up_sampler": true, "warmup_steps": 0} |