优化训练配置
=================
设备: cpu
训练集大小: 3000
测试集大小: 500
批次大小: 4
梯度累积步数: 4
最大序列长度: 128
训练轮次: 2
学习率: 3e-05
评估结果: {'eval_loss': 0.04865079000592232, 'eval_accuracy': 0.982, 'eval_f1': 0.9820013722478238, 'eval_precision': 0.9820710955710956, 'eval_recall': 0.982, 'eval_runtime': 4.7257, 'eval_samples_per_second': 105.804, 'eval_steps_per_second': 13.331, 'epoch': 2.0}