Skip to content

GitLab

Switch branch/tag
  1. 19 Nov, 2021 1 commit
  3. 17 Nov, 2021 1 commit
  5. 02 Jun, 2021 1 commit
  7. 13 Apr, 2021 1 commit
    • Reed Wanderman-Milne's avatar
      Use nonexperimental mixed precision API. · c0ac8d1c
      Reed Wanderman-Milne authored 
      This replaces symbols in tf.keras.mixed_precision.experimental with the corresponding nonexperimental symbols. In some cases, passing a Policy is replaced with passing a policy name for conciseness.

Additionally, for the Shakespeare model, the loss_scale flag is removed, since supporting it with the nonexperimental API is slightly more verbose and it is recommended users use the default loss scale.

PiperOrigin-RevId: 368123944
      c0ac8d1c
  9. 12 Apr, 2021 1 commit
    • Reed Wanderman-Milne's avatar
      Use nonexperimental mixed precision API for official models. · e6cda015
      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
      e6cda015
  11. 09 Apr, 2021 1 commit
    • Reed Wanderman-Milne's avatar
      Remove dynamic_loss_scale argument to define_performance. · 20ed8cf0
      Reed Wanderman-Milne authored 
      All models which support loss scaling support dynamic loss scaling, so the argument has no purpose. It used to be that some models scaled the loss manually instead of using a LossScaleOptimizer, and so did not support dynamic loss scaling.

PiperOrigin-RevId: 367719521
      20ed8cf0
  13. 06 Apr, 2021 1 commit
  15. 23 Mar, 2021 1 commit
  17. 03 Mar, 2021 1 commit
  19. 13 Sep, 2020 1 commit
  21. 10 Sep, 2020 1 commit
  23. 09 Sep, 2020 1 commit
  25. 26 Aug, 2020 5 commits
  27. 12 Aug, 2020 1 commit
  29. 10 Aug, 2020 1 commit
  31. 11 Jun, 2020 1 commit
  33. 19 May, 2020 1 commit
  35. 14 May, 2020 1 commit
  37. 26 Apr, 2020 1 commit
  39. 20 Apr, 2020 1 commit
  41. 17 Apr, 2020 2 commits
  43. 10 Apr, 2020 1 commit
  45. 04 Apr, 2020 1 commit
  47. 25 Mar, 2020 1 commit
  49. 17 Mar, 2020 1 commit
  51. 16 Mar, 2020 2 commits
  53. 12 Mar, 2020 1 commit
  55. 11 Mar, 2020 1 commit
  57. 05 Mar, 2020 1 commit
  59. 26 Feb, 2020 1 commit
  61. 24 Feb, 2020 1 commit
  63. 28 Jan, 2020 1 commit
  65. 29 Aug, 2019 1 commit
    • Reed Wanderman-Milne's avatar
      Use new mixed_float16 policy for resnet. · dcd0e7ad
      Reed Wanderman-Milne authored 
      The old infer_float32_policies policy will be removed from TensorFlow soon.

To test convergence, I ran the Resnet50KerasAccuracy.benchmark_8_gpu_fp16 benchmark. I got an accuracy of 0.76037 and an exp_per_second of 6908.

PiperOrigin-RevId: 266191126
      dcd0e7ad
  67. 22 Aug, 2019 1 commit