dropreg/R-Drop

Issues

• Clarification on Using Concatenated Input for R-Drop Training

  #33 opened a year ago by xyb314
  1

• Training configuration for the WMT14 EnDe dataset?

  #19 opened 3 years ago by frankang
  5

• How to use the data parallel in r-drop.

  #32 opened 2 years ago by xinxinxing
  1

• Question of the proof

  #31 opened 2 years ago by SYSUykLin
  1

• Some question about reproducing GLUE

  #30 opened 2 years ago by wpwpwpyo
  0

• Can not reproduce following the hyperparameter in the paper for finefuning ViT on Cifar100

  #25 opened 2 years ago by NamlessM
  4

• How the `warmup steps` affects the performance?

  #28 opened 3 years ago by Doragd
  2

• kl loss in ViT example supposed to be divided by 2?

  #29 opened 3 years ago by sieu-n
  1

• Can I use R-Drop in Semantic Search?

  #27 opened 3 years ago by ralgond
  1

• Unable to preprocess data for summarization

  #14 opened 3 years ago by samiksome92
  2

• error: argument --task: invalid choice: 'rdrop_translation'

  #22 opened 3 years ago by tairan-w
  1

• pip install --editable .报错

  #15 opened 3 years ago by Shiwen-Ni
  1

• can not reproduce the results following the hyparameters in the paper

  #23 opened 3 years ago by leoozy
  1

• About the implementation in transformers, where the reduction in ce_loss uses the mean (by default), while KL uses the reduction is sum ?

  #24 opened 3 years ago by XiaoqingNLP
  1

• Inconsistency for KL loss and CE loss hyper-parameters and baselines results in GLUE

  #6 opened 3 years ago by zhangzhenyu13
  5

• JS divergence in the research paper?

  #17 opened 3 years ago by sieu-n
  1

• R-drop makes my model broken.

  #12 opened 3 years ago by MayDomine
  9

• Where is R-Drop code in R-Drop/huggingface_transformer_src/bert_rdrop/run_glue.py?

  #18 opened 3 years ago by zhenshiqi1996
  6

• difference between R-Drop and SimCse + Smart

  #21 opened 3 years ago by cuixuage
  1

• unable to reproduce results on GLUE

  #16 opened 3 years ago by 1024er
  2

• Can mseloss replace KL divergence？

  #20 opened 3 years ago by 18335100284
  1

• What's Wrong with my TensorFlow (1.14 or 1.15) implementation?

  #13 opened 3 years ago by guotong1988
  2

• Will KLD loss degrease very fast?

  #5 opened 3 years ago by snsun
  9

• Readme File for RoBerta example.

  #11 opened 3 years ago by ShreyPandit
  1

• CUDA error: CUBLAS_STATUS_EXECUTION_FAILED

  #9 opened 3 years ago by paul-chelarescu
  2

• Summarization task fails with 'Trying to backward through the graph a second time'

  #8 opened 3 years ago by paul-chelarescu
  2

• A simple way to double the impact of R-Drop

  #10 opened 3 years ago by guotong1988
  3

• what the dropout should be set when we predict or test?

  #7 opened 3 years ago by hitwangshuai
  2

• Fairseq tasks install work?

  #4 opened 3 years ago by kungfu-eric
  1

• Why you use (p, q_tec) and (q, p_tec) rather than (p, q) and (q, p) to compute kl-loss?

  #3 opened 3 years ago by JaheimLee
  0

• What are the core code lines of R-Drop? Thank you very much.

  #2 opened 3 years ago by guotong1988
  1

• What are the core code lines of R-Drop? Thank you very much.

  #1 opened 3 years ago by guotong1988
  0