bohanli/BERT-flow

Issues

• BERT-flow是什么原理，还是不懂。

  #20 opened a year ago by guotong1988
  3

• 各位大佬，有监督 和 无监督 的 训练目标 各是什么啊？

  #21 opened a year ago by guotong1988
  0

• Will it be better to replace BERT with ELECTRA?

  #9 opened a year ago by qiuwenbogdut
  0

• 无监督训练为什么会有label呢？

  #4 opened 4 years ago by levishen
  5

• do you have pytorch implement?

  #15 opened 2 years ago by lijiaoyang
  1

• 训练损失为负数，

  #19 opened 2 years ago by chenmingwei00
  0

• How can I get a single sentence embedding？

  #12 opened 4 years ago by Gpwner
  4

• Fine-tuning bert-flow embedding with supervision

  #10 opened 3 years ago by JaheimLee
  0

• about flow acc

  #8 opened 4 years ago by LLLLLLoki
  5

• Correction for a mistake in 'last2avg'

  #16 opened 4 years ago by alexlimh
  0

• loss为负

  #7 opened 4 years ago by wwwjs
  15

• Which code show that flow network is optimized while the BERT parameters remain unchanged？

  #13 opened 4 years ago by qiuwenbogdut
  0

• May I get access to BERT-base-NLI-flow ?

  #14 opened 4 years ago by yaoxingcheng
  0

• avg-last-2 取的是[-0]和[-1]层的embedding

  #11 opened 4 years ago by sc1054
  2

• predict 的问题

  #6 opened 4 years ago by WJSAI
  3

• some questions about this paper

  #5 opened 4 years ago by HUSTHY
  0

• Is there a pytorch version coeds in the future?

  #2 opened 4 years ago by HUSTHY
  1

• 请问, 校正过程所用语料覆盖面是否要足够大

  #1 opened 4 years ago by MrRace
  7

• The paper link at the beginning of README is incorrect

  #3 opened 4 years ago by jxhe
  1