lucidrains/h-transformer-1d

Issues

• error in test

  #25 opened 10 months ago by jizhang02
  1

• Billion Word Benchmark - Reproducibility

  #24 opened a year ago by DavidHerel
  0

• Use for relatively short sequences and small datasets

  #23 opened 2 years ago by Vedasheersh
  0

• Number of processed inputs in causal version is off

  #22 opened 2 years ago by vladyorsh
  2

• Approximated values are off

  #20 opened 3 years ago by jglaser
  2

• Masking not working in training, thanks

  #18 opened 3 years ago by junyongyou
  6

• Sequence classfication, thanks a lot

  #17 opened 3 years ago by junyongyou
  0

• Application to sequence classification?

  #12 opened 3 years ago by trpstra
  4

• Multiple applications of positional embeddings?

  #14 opened 3 years ago by Clemens123
  1

• Add Norm Missing

  #16 opened 3 years ago by wwx13
  2

• Mask not working

  #15 opened 3 years ago by wwx13
  2

• Algorithm Mismatch

  #13 opened 3 years ago by jinmang2
  3

• Does its model include relative position embedding?

  #11 opened 3 years ago by hadaev8
  1

• eos token does not work in batch mode generation

  #9 opened 3 years ago by tmphex
  4

• One simple question

  #10 opened 3 years ago by CiaoHe
  2

• Does h-transformer-1d have a subquadratic cost?

  #7 opened 3 years ago by jinmang2
  1

• RuntimeError: Tensor type unknown to einops <class 'torch.return_types.max'>

  #6 opened 3 years ago by wajihullahbaig
  4

• ModuleNotFoundError: No module named 'rotary_embedding_torch'

  #5 opened 3 years ago by wajihullahbaig
  1

• Sequence length issue when `causal = True`

  #8 opened 3 years ago by jaak-s
  1

• I have some questions about implementation details

  #1 opened 3 years ago by jinmang2
  3

• Mini-batching (b > 1) does not work with masking

  #4 opened 3 years ago by jaak-s
  2

• Example in README does not work

  #3 opened 3 years ago by jaak-s
  2

• H-Transformer for Cross-Attention?

  #2 opened 3 years ago by Vbansal21
  4