- 23 Oct, 2019 2 commits
-
-
Yeqing Li authored
PiperOrigin-RevId: 276317091
-
Hongkun Yu authored
PiperOrigin-RevId: 276210315
-
- 22 Oct, 2019 6 commits
-
-
Rohan Jain authored
Getting rid of experimental_allow_stateful now that we have experimental_external_state_policy instead. that option. We've replaced experimental_allow_stateful with experimental_external_state_policy with a default of WARN in which we'll only print a warning message and not fail. As a result, we can now safely remove all client code that explicitly set the allow_stateful_ops flag to True. PiperOrigin-RevId: 276115313
-
Ruoxin Sang authored
PiperOrigin-RevId: 276103495
-
Dmitry Murygin authored
-
Hongkun Yu authored
- Support first/last summary type - Support bert format input processing PiperOrigin-RevId: 276005310
-
Hongkun Yu authored
PiperOrigin-RevId: 275994710
-
Hongkun Yu authored
PiperOrigin-RevId: 275986990
-
- 21 Oct, 2019 6 commits
-
-
Ruoxin Sang authored
Pass the file patterns instead of flattened file paths into tf.data.Dataset.list_file. This helps Cloud DF4x4 Bert Startup from 11min34s to 6min53s. PiperOrigin-RevId: 275953359
-
Yeqing Li authored
PiperOrigin-RevId: 275940140
-
Frank Chen authored
PiperOrigin-RevId: 275930249
-
minoring authored
arparse -> argparse
-
Hongkun Yu authored
PiperOrigin-RevId: 275879424
-
Yeqing Li authored
PiperOrigin-RevId: 275867562
-
- 20 Oct, 2019 1 commit
-
-
Hongkun Yu authored
PiperOrigin-RevId: 275739207
-
- 19 Oct, 2019 4 commits
-
-
Jing Li authored
PiperOrigin-RevId: 275644913
-
Hongkun Yu authored
PiperOrigin-RevId: 275610155
-
Hongkun Yu authored
PiperOrigin-RevId: 275603936
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 275578662
-
- 18 Oct, 2019 5 commits
-
-
Hongkun Yu authored
Remove train_data_size flag PiperOrigin-RevId: 275545035
-
Hongkun Yu authored
PiperOrigin-RevId: 275428252
-
Hongkun Yu authored
PiperOrigin-RevId: 275417626
-
Jing Li authored
PiperOrigin-RevId: 275408074
-
Hongkun Yu authored
PiperOrigin-RevId: 275393975
-
- 17 Oct, 2019 6 commits
-
-
minoring authored
-
marcussorealheis authored
-
Yeqing Li authored
PiperOrigin-RevId: 275306010
-
Hongkun Yu authored
PiperOrigin-RevId: 275288636
-
Hongkun Yu authored
PiperOrigin-RevId: 275192365
-
minoring authored
-
- 16 Oct, 2019 5 commits
-
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 275142626
-
David Chen authored
PiperOrigin-RevId: 275103426
-
Yeqing Li authored
PiperOrigin-RevId: 275080469
-
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
-
Yeqing Li authored
PiperOrigin-RevId: 274921478
-
- 15 Oct, 2019 5 commits
-
-
Yeqing Li authored
PiperOrigin-RevId: 274918820
-
Hongkun Yu authored
PiperOrigin-RevId: 274917111
-
Jing Li authored
Add to option to init checkpoint from transformer-xl model. PiperOrigin-RevId: 274875006
-
Hongkun Yu authored
PiperOrigin-RevId: 274844449
-
Yeqing Li authored
PiperOrigin-RevId: 274807747
-