Skip to content

GitLab

"src/include/vscode:/vscode.git/clone" did not exist on "60a0f2860f3152e1cb325fb70c5cba8435020c46"
Switch branch/tag
  1. 13 Feb, 2020 2 commits
  3. 06 Feb, 2020 1 commit
  5. 03 Feb, 2020 1 commit
  7. 01 Feb, 2020 1 commit
  9. 31 Jan, 2020 1 commit
  11. 30 Jan, 2020 1 commit
  13. 29 Jan, 2020 2 commits
  15. 28 Jan, 2020 3 commits
  17. 23 Jan, 2020 1 commit
  19. 17 Jan, 2020 1 commit
  21. 19 Dec, 2019 1 commit
  23. 18 Dec, 2019 1 commit
  25. 17 Dec, 2019 3 commits
  27. 14 Dec, 2019 1 commit
  29. 13 Dec, 2019 1 commit
  31. 12 Dec, 2019 1 commit
  33. 06 Dec, 2019 1 commit
  35. 05 Dec, 2019 1 commit
    • Haoyu Zhang's avatar
      Remove tweaked benchmark for no_dist_strat in function mode. · ef8aed79
      Haoyu Zhang authored 
      No need to test explicit_gpu_placement + no_dist_strat + function case, since the performance issue is already fixed by unifying the Keras codepaths. However, there is still performance gap using the explicit_gpu_placement flag when run_eagerly.

PiperOrigin-RevId: 283890008
      ef8aed79
  37. 04 Dec, 2019 1 commit
  39. 03 Dec, 2019 1 commit
  41. 27 Nov, 2019 1 commit
  43. 25 Nov, 2019 2 commits
  45. 22 Nov, 2019 2 commits
  47. 20 Nov, 2019 1 commit
  49. 12 Nov, 2019 1 commit
  51. 11 Nov, 2019 2 commits
  53. 08 Nov, 2019 1 commit
  55. 06 Nov, 2019 2 commits
  57. 02 Nov, 2019 1 commit
  59. 01 Nov, 2019 1 commit
    • Reed Wanderman-Milne's avatar
      Support fp16 using tf.keras.mixed_precision in CTL resnet. · ffa522ea
      Reed Wanderman-Milne authored 
      To test, I ran the following command:

python resnet_ctl_imagenet_main.py --batch_size=2048 --data_dir ~/imagenet --datasets_num_private_threads=14 --epochs_between_evals=10 --model_dir ~/tmp_model_dir --clean --num_gpus=8 --train_epochs=90 --dtype=fp16

I got 76.15% final evaluation accuracy.

PiperOrigin-RevId: 278010061
      ffa522ea