facebookresearch/mega

Issues

• Does the training script retrain the entire model from scratch?

  #14 opened a year ago by Tylersuard
  1

• RuntimeError: shape '[32, 1, 1000, 64]' is invalid for input of size 2273280

  #16 opened a year ago by Tylersuard
  2

• Instructions for generating tokens from LM?

  #18 opened a year ago by Tylersuard
  0

• Script hangs forever on validation step

  #17 opened a year ago by Tylersuard
  0

• New Pathfinder-X2 Dataset!

  #15 opened a year ago by Tylersuard
  0

• Maximum context length?

  #13 opened a year ago by Tylersuard
  0

• Any plans to merge with `fairseq`?

  #12 opened 2 years ago by mahnerak
  0

• How to get test accuracy

  #9 opened 2 years ago by aleksandar-terzic
  2

• Attention masking in MovingAverageGatedAttention

  #11 opened 2 years ago by jambo6
  3

• Tokenization for downstream tasks

  #10 opened 2 years ago by danigoju
  1

• Non-commercial license

  #7 opened 2 years ago by mnaylor5
  4

• ONNX support

  #6 opened 2 years ago by leonid-pishchulin
  1

• Regarding the damping factor δ

  #4 opened 2 years ago by ciaua
  1

• Fail for transformer_lra_pf32

  #3 opened 2 years ago by nachewigkeit
  3

• The lack of warmup-updates

  #2 opened 2 years ago by nachewigkeit
  1

• Lack of src-bin/dict.txt

  #1 opened 2 years ago by nachewigkeit
  0