Skip to content

GitLab

"git@developer.sourcefind.cn:lacacy/qwen_lmdeploy.git" did not exist on "902a3e16f937d4f142968dea265c9e03c8559bb8"
Switch branch/tag
  1. 02 Mar, 2020 1 commit
  3. 28 Feb, 2020 1 commit
  5. 25 Feb, 2020 1 commit
  7. 21 Feb, 2020 2 commits
  9. 20 Feb, 2020 1 commit
  11. 13 Feb, 2020 1 commit
  13. 29 Jan, 2020 1 commit
  15. 27 Jan, 2020 1 commit
  17. 21 Jan, 2020 1 commit
  19. 17 Jan, 2020 1 commit
  21. 15 Dec, 2019 1 commit
  23. 14 Dec, 2019 2 commits
  25. 11 Dec, 2019 1 commit
  27. 06 Dec, 2019 1 commit
  29. 05 Dec, 2019 1 commit
  31. 04 Dec, 2019 1 commit
  33. 27 Nov, 2019 1 commit
  35. 25 Nov, 2019 1 commit
  37. 21 Nov, 2019 1 commit
  39. 19 Nov, 2019 1 commit
  41. 18 Nov, 2019 1 commit
  43. 11 Nov, 2019 1 commit
  45. 28 Oct, 2019 1 commit
  47. 21 Oct, 2019 1 commit
  49. 16 Oct, 2019 1 commit
    • Reed Wanderman-Milne's avatar
      Add support for the tf.keras.mixed_precision API in NCF · cb913691
      Reed Wanderman-Milne authored 
      To test, I did 50 fp32 runs and 50 fp16 runs. I used the following command:

python ncf_keras_main.py --dataset=ml-20m --num_gpus=1 --train_epochs=10 --clean --batch_size=99000 --learning_rate=0.00382059 --beta1=0.783529 --beta2=0.909003 --epsilon=1.45439e-7 --layers=256,256,128,64 --num_factors=64 --hr_threshold=0.635 --ml_perf --nouse_synthetic_data --data_dir ~/ncf_data_dir_python3 --model_dir ~/tmp_model_dir --keras_use_ctl

For the fp16 runs, I added --dtype=fp16. The average hit-rate for both fp16 and fp32 was 0.6365. I also did 50 runs with the mixed precision graph rewrite, and the average hit-rate was 0.6363. The difference is likely due to noise.

PiperOrigin-RevId: 275059871
      cb913691
  51. 10 Oct, 2019 1 commit
  53. 07 Oct, 2019 1 commit
  55. 24 Sep, 2019 1 commit
  57. 09 Sep, 2019 1 commit
  59. 04 Sep, 2019 1 commit
  61. 30 Aug, 2019 1 commit
  63. 26 Aug, 2019 1 commit
  65. 23 Aug, 2019 2 commits
  67. 21 Aug, 2019 1 commit
  69. 20 Aug, 2019 3 commits