1 CBLUE_QQR
baseline1:使用官方推荐的预训练模型:chinese-bert-wwm-ext 结果:训练最终精度0.8386,略低官方baseline的结果84.4
Training Stop! The best step 2000: 0.8386491557223265
Saving models checkpoint to data/output/qqr/chinese-bert-wwm-ext
baseline2:使用官方推荐的预训练模型:chinese-roberta-wwm-ext-large 结果:训练最终精度0.8393,略低官