yxuansu/SimCTG

Issues

• pip install error

  #28 opened 2 years ago by AbrahamSanders
  2

• SimCTG BART training

  #27 opened 2 years ago by rahulseetharaman
  0

• Can you provide BartModel for SimCTG code?

  #26 opened 2 years ago by luoda888
  0

• Questions about Document Generation

  #22 opened 2 years ago by dongqian0206
  1

• What's the difference between SimCTG and CnNT?

  #25 opened 2 years ago by eyuansu62
  0

• About metric reproduction

  #24 opened 2 years ago by LeiyanGithub
  0

• Questions about evaluation metrics coherence and gen-ppl?

  #23 opened 2 years ago by AdamRain
  0

• Contrastive Training loss的部分疑惑

  #21 opened 2 years ago by aa327chenge
  1

• SimCTGT5和fast_contrastive_search相关问题（EncDecContrastiveDecodingOneStepFast）

  #20 opened 2 years ago by Xyr990127
  2

• about the repetition of the ground-turth

  #18 opened 2 years ago by LHRYANG
  3

• Bloom Ai

  #17 opened 2 years ago by lecrycry
  3

• license file

  #16 opened 2 years ago by shirayu
  6

• Typo in story_generation part

  #14 opened 2 years ago by DaehanKim
  1

• about contrastive search

  #13 opened 2 years ago by Saicat
  1

• fast_contrastive_search只支持batch_size=1吗

  #12 opened 2 years ago by awdrgyjilplij
  1

• Question about replicating the MAUVE scores

  #7 opened 2 years ago by XiangLi1999
  3

• Dialogue generation training with simctg library

  #11 opened 2 years ago by lecrycry
  1

• 请问contrastive loss对生成的影响程度如何，应如何定义contrastive loss和生成的cross-entropy loss的权重大小？

  #10 opened 2 years ago by yuanhuachao
  2

• 关于fast_contrastive_search实现中新生成的token的embedding的理解(hv)

  #9 opened 2 years ago by HUSTHY
  6

• similar sentence generation

  #8 opened 2 years ago by nikky4D
  2

• hi ,have you test the best sample_step when using diverse_contrastive_search?

  #6 opened 2 years ago by wuzhiye7
  1

• repeated token is generated comparing to beam-search, when using fast_contrastive_search on T5

  #5 opened 2 years ago by wuzhiye7
  4

• We have updated instructions on how to apply contrastive search on encoder-decoder models (e.g. BART and T5).

  #4 opened 2 years ago by yxuansu
  8

• Have you tested the effect of Contrastive Search method on encoder+decoder model such as BART?

  #2 opened 3 years ago by maojinyang
  6

• How to generate text without giving prefix ?

  #3 opened 3 years ago by Atharva-Phatak
  1

• Could you provide the implementations of baselines?

  #1 opened 3 years ago by Jxu-Thu
  1