1. 01 Aug, 2022 1 commit
  2. 29 Jul, 2022 1 commit
  3. 28 Jul, 2022 1 commit
  4. 26 Jul, 2022 1 commit
  5. 25 Jul, 2022 1 commit
  6. 24 Jul, 2022 1 commit
  7. 21 Jul, 2022 1 commit
    • A. Unique TensorFlower's avatar
      Fix dropout rate bug. · b8234e65
      A. Unique TensorFlower authored
      Fix attention dropout rate bug. The output_dropout rate was used for attention dropout by mistake.
      
      PiperOrigin-RevId: 462470287
      b8234e65
  8. 14 Jul, 2022 1 commit
  9. 13 Jul, 2022 2 commits
  10. 30 Jun, 2022 1 commit
  11. 29 Jun, 2022 1 commit
  12. 27 Jun, 2022 1 commit
  13. 21 Jun, 2022 1 commit
  14. 20 Jun, 2022 1 commit
  15. 09 Jun, 2022 1 commit
  16. 08 Jun, 2022 1 commit
  17. 07 Jun, 2022 1 commit
  18. 21 May, 2022 1 commit
  19. 14 May, 2022 1 commit
  20. 04 May, 2022 2 commits
  21. 03 May, 2022 1 commit
  22. 20 Apr, 2022 1 commit
  23. 13 Apr, 2022 1 commit
  24. 28 Mar, 2022 1 commit
  25. 23 Mar, 2022 1 commit
  26. 11 Mar, 2022 3 commits
  27. 04 Mar, 2022 1 commit
  28. 03 Mar, 2022 1 commit
  29. 25 Jan, 2022 1 commit
  30. 19 Jan, 2022 1 commit
  31. 18 Jan, 2022 1 commit
  32. 04 Jan, 2022 1 commit
  33. 29 Dec, 2021 1 commit
    • Xin Wang's avatar
      Added block diagonal feedforward layer. · a7e60974
      Xin Wang authored
      This layer replaces the weight matrix of the output_dense layer with a block diagonal matrix to save layer parameters and FLOPs. A linear mixing layer can be added optionally to improve layer expressibility.
      
      PiperOrigin-RevId: 418828099
      a7e60974
  34. 21 Dec, 2021 1 commit
  35. 16 Dec, 2021 1 commit
  36. 03 Dec, 2021 1 commit