CLUEbenchmark/CLUEPretrainedModels

Issues

• transformers中使用clue/roberta_chinese_pair_tiny的疑问

  #23 opened 3 years ago by zy614582280
  1

• 继续 Pretraining 的问题

  #22 opened 3 years ago by Jhangsy
  0

• 你好 作者，论文里面提出来一种新的注意力机制，能否给下代码，写的太抽象了，无法理解

  #21 opened 4 years ago by zhaolulul
  1

• 预训练时的max_Seq_length

  #20 opened 4 years ago by gsxf997
  1

• MLM能否正常inference

  #18 opened 4 years ago by YuxianMeng
  0

• CLUE发布的roberta模型，预训练时是否使用了wwm呢？

  #17 opened 4 years ago by waywaywayw
  0

• roberta_tiny_clue在IFLYTEK’的训练参数

  #16 opened 4 years ago by selephantjy
  0

• 关于RoBERTa pair的预训数据构造形式

  #14 opened 4 years ago by songt96
  1

• XLnet

  #15 opened 4 years ago by onlyonewater
  0

• bert_config.json 里的参数问题

  #12 opened 4 years ago by Jhangsy
  2

• 请问能开放基于clue数据的base模型，large太大训不动，tiny又不如bert base。

  #11 opened 4 years ago by InsaneLife
  2

• 能否考虑将预训练模型放到讯飞云？

  #10 opened 4 years ago by Fan9
  0

• 大佬好，请问是否有发布base版BERT/RoBERTa的计划呢？

  #9 opened 4 years ago by justzhanghong
  2

• 句子对任务的RoBERTa-tiny-pair的ckpt文件的问题

  #4 opened 5 years ago by drzqb
  8

• RoBERTa-tiny-clue 和RoBERTa-tiny-pair的异同

  #2 opened 5 years ago by chros425
  2

• 新版transformers中已经没有WarmupLinearSchedule了

  #3 opened 5 years ago by DrDavidS
  1

• 如果用大模型（比如roberta_large）的话，推荐使用多大的学习率

  #1 opened 5 years ago by zzy99
  3