hparam: !!python/object:argparse.Namespace accumulate_grad_batches: 1 adam_epsilon: 1.0e-08 callback: !!python/object:argparse.Namespace early_stop: true lr_scheduler: true model_checkpoint: true push_to_hub: true dataset: !!python/object:argparse.Namespace config: english_v12 name: conll2012_ontonotesv5 eval_batch_size: 8 huggingface_hub: !!python/object:argparse.Namespace repo_name: t5_conll_ontonotes_en14 user_name: NanaEilish log_name: t5_pretrain lr: 0.0003 lr_decay_min_lr: 1.0e-05 lr_decay_rate: 0.5 lr_decay_steps: 20 lr_scheduler: linear_wmup max_epochs: 5 max_grad_norm: 1 max_len: 512 max_output_len: 100 model_name: t5-base monitor_max_or_min: min monitor_metrics: val_loss n_gpu: 4 num_workers: 8 output_dir: output/t5_pretrain seed: 42 train_batch_size: 8 train_size: 30 valid_size: 1 warmup_ratio: 0.2 weight_decay: 1.0e-05