Skip to content

GitLab

Switch branch/tag
  1. 13 Jul, 2021 1 commit
  3. 09 Jul, 2021 1 commit
    • Reed Wanderman-Milne's avatar
      With float16, always use LossScaleOptimizer. · be3575f5
      Reed Wanderman-Milne authored 
      Before, it was too easy to accidentally forget to set runtime.loss_scale, which had to always be done if mixed precision is used, otherwise the model would converge to worse accuracy. Now, all that needs to be done to use mixed precision is to set runtime.mixed_precision_dtype=float16.

PiperOrigin-RevId: 383767033
      be3575f5
  5. 24 Jun, 2021 1 commit
  7. 23 Jun, 2021 1 commit
    • Reed Wanderman-Milne's avatar
      Improve error message when certain flags are not specified. · 8b47c484
      Reed Wanderman-Milne authored 
      In nlp/train.py and vision/beta/train.py, certain flags are marked as required. Additionally, in certain functions, error messages are improved if a necessary flag is not specified, which is a fallback in case a file calling define_flags() does not mark the necessary flags are required. Previously if any of these flags were not specified, it would crash with a cryptic error message, making it hard to tell what went wrong.

In a subsequent change, I will mark flags as required in more files which call define_flags().

PiperOrigin-RevId: 381066985
      8b47c484
  9. 22 Jun, 2021 1 commit
  11. 20 Jun, 2021 1 commit
  13. 16 Jun, 2021 2 commits
  15. 11 Jun, 2021 2 commits
  17. 01 Jun, 2021 2 commits
  19. 28 May, 2021 2 commits
  21. 17 May, 2021 2 commits
  23. 14 May, 2021 2 commits
  25. 13 May, 2021 2 commits
  27. 06 May, 2021 2 commits
  29. 16 Apr, 2021 2 commits
  31. 13 Apr, 2021 4 commits
  33. 12 Apr, 2021 2 commits
    • Reed Wanderman-Milne's avatar
      Use nonexperimental mixed precision API for official models. · ba8ad4f5
      Reed Wanderman-Milne authored 
      For all modified calls to set_mixed_precision_policy(), the loss_scale argument was removed, as it cannot be passed if the nonexperimental API is used. For all such callers, the loss_scale is later used to explicitly create a LossScaleOptimizer, so removing the argument has no impact.

Switching to the non-experimental LossScaleOptimizer has no effect, as it has near identical behavior and all isinstance checks within the official models check for the non-experimental version.

PiperOrigin-RevId: 368101975
      ba8ad4f5
    • Reed Wanderman-Milne's avatar
      Use nonexperimental mixed precision API for official models. · 0d8f9807
      Reed Wanderman-Milne authored 
      For all modified calls to set_mixed_precision_policy(), the loss_scale argument was removed, as it cannot be passed if the nonexperimental API is used. For all such callers, the loss_scale is later used to explicitly create a LossScaleOptimizer, so removing the argument has no impact.

Switching to the non-experimental LossScaleOptimizer has no effect, as it has near identical behavior and all isinstance checks within the official models check for the non-experimental version.

PiperOrigin-RevId: 368101975
      0d8f9807
  35. 05 Apr, 2021 6 commits
  37. 02 Apr, 2021 2 commits
  39. 01 Apr, 2021 2 commits