Evaluation: {'eval_loss': 0.2520483732223511,
'eval_accuracy': 0.9354838709677419,
'eval_precision': 0.9397849462365592,
'eval_recall': 0.9354838709677419,
'eval_f1': 0.9204671857619577,
'eval_runtime': 0.1265,
'eval_samples_per_second': 490.111,
'eval_steps_per_second': 7.905}
Batch size: 128
lr = 5e-5
Model = 'roberta-base'
