Skip to content

Instantly share code, notes, and snippets.

@aloha12345x
Created March 23, 2021 03:31
Show Gist options
  • Save aloha12345x/b2b81c52dd5fee7b47e3a4eb537232f3 to your computer and use it in GitHub Desktop.
Save aloha12345x/b2b81c52dd5fee7b47e3a4eb537232f3 to your computer and use it in GitHub Desktop.
GAIIC Track3 Share

GAIIC Track3 “猫老板”队分享

数据处理

  • 清洗:分开连在一起的英文单词,规范标点符号
  • 增强:左右交换pair

Pretrain

  • 使用了hfl/chinese-roberta-wwm-exthfl/chinese-roberta-wwm-ext-large
  • 单句和pair都用
  • 使用各种mask

Finetune

  • 左右交换pair,增强数据
  • hfl/chinese-roberta-wwm-ext-large:lr=1e-5,weigt clip=1
  • 五折交叉验证
  • 多个checkpoint ensemble
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment