Commits · fb35d6bef6eecc640fb865dd2fc73d8fee2a93b6 · ModelZoo / ResNet50_tensorflow

25 Feb, 2020 1 commit
- Creates modeling/performance.py to include mix prediction related stuff · fb35d6be
  Hongkun Yu authored Feb 24, 2020
```
PiperOrigin-RevId: 297002741
```
  fb35d6be
21 Feb, 2020 2 commits
- Write examples/second and steps/second summaries in TimeHistory callback. · b60dc237
  Will Cromar authored Feb 21, 2020
```
PiperOrigin-RevId: 296507807
```
  b60dc237
- Internal change · a2e95a60
  A. Unique TensorFlower authored Feb 21, 2020
```
PiperOrigin-RevId: 296469938
```
  a2e95a60
20 Feb, 2020 1 commit
- Internal change · 3f29beb0
  A. Unique TensorFlower authored Feb 19, 2020
```
PiperOrigin-RevId: 296108291
```
  3f29beb0
13 Feb, 2020 1 commit
- [Refactor] TF models: move all contents of transformer to nlp/transformer · 7f926353
  Hongkun Yu authored Feb 13, 2020
```
PiperOrigin-RevId: 294997928
```
  7f926353
29 Jan, 2020 1 commit
- remove reference_data to reduce github files. · d33a0b42
  Hongkun Yu authored Jan 29, 2020
```
PiperOrigin-RevId: 292178802
```
  d33a0b42
27 Jan, 2020 1 commit
- Remove 'num_workers' arg from get_distribution_strategy() method. · 1af7172d
  Yanhui Liang authored Jan 27, 2020
```
PiperOrigin-RevId: 291810091
```
  1af7172d
21 Jan, 2020 1 commit
- Remove explicit initialize_tpu_system call from model garden. · ac97f01b
  Peter Hawkins authored Jan 21, 2020
```
PiperOrigin-RevId: 290795650
```
  ac97f01b
17 Jan, 2020 1 commit
- Remove explicit initialize_tpu_system call from model garden. · 855ed8bc
  Ruoxin Sang authored Jan 17, 2020
```
PiperOrigin-RevId: 290354680
```
  855ed8bc
15 Dec, 2019 1 commit
- Clearly demarcate contrib symbols from standard tf symbols by importing them directly. · 722d9e57
  Hongkun Yu authored Dec 14, 2019
```
PiperOrigin-RevId: 285618209
```
  722d9e57
14 Dec, 2019 2 commits
- Internal change · 0788a23c
  Hongkun Yu authored Dec 13, 2019
```
PiperOrigin-RevId: 285533511
```
  0788a23c
- Clearly demarcate contrib symbols from standard tf symbols by importing them directly. · 357f30f4
  A. Unique TensorFlower authored Dec 13, 2019
```
PiperOrigin-RevId: 285503670
```
  357f30f4
11 Dec, 2019 1 commit
- Internal change · 26bbda73
  David Chen authored Dec 10, 2019
```
PiperOrigin-RevId: 284874717
```
  26bbda73
06 Dec, 2019 1 commit

Remove code that disables pin-to-host optimization. It's disabled by default... · 6f64dcb5

Haoyu Zhang authored Dec 05, 2019

Remove code that disables pin-to-host optimization. It's disabled by default and we already have XLA presubmit that will catch any issues when the optimization is reenabled.

PiperOrigin-RevId: 284078252

6f64dcb5

05 Dec, 2019 1 commit
- Moved set_gpu_thread_mode_and_count from vision/image_classification/common.py to · f926be0a
  A. Unique TensorFlower authored Dec 04, 2019
```
utils/misc/keras_utils

PiperOrigin-RevId: 283885102
```
  f926be0a
04 Dec, 2019 1 commit
- Internal change · 4270e416
  Yeqing Li authored Dec 04, 2019
```
PiperOrigin-RevId: 283770513
```
  4270e416
27 Nov, 2019 1 commit
- Remove 'default' in get_distribution_strategy which is complex and error-prone · 04256053
  Hongkun Yu authored Nov 26, 2019
```
PiperOrigin-RevId: 282669615
```
  04256053
25 Nov, 2019 1 commit
- Internal change · 5b2575c2
  Zongwei Zhou authored Nov 24, 2019
```
PiperOrigin-RevId: 282293668
```
  5b2575c2
21 Nov, 2019 1 commit
- Internal change · 4c872f63
  Sai Ganesh Bandiatmakuri authored Nov 21, 2019
```
PiperOrigin-RevId: 281846531
```
  4c872f63
19 Nov, 2019 1 commit
- Properly save/restore flags in model tests · 1b8c0ee8
  Jose Baiocchi authored Nov 18, 2019
```
PiperOrigin-RevId: 281192912
```
  1b8c0ee8
18 Nov, 2019 1 commit
- Internal change · 5a3b762c
  Allen Wang authored Nov 18, 2019
```
PiperOrigin-RevId: 281063737
```
  5a3b762c
11 Nov, 2019 1 commit
- [Clean up] Move TF1 only TPU utils to r1 folder. · 0c5919ef
  Hongkun Yu authored Nov 11, 2019
```
PiperOrigin-RevId: 279798675
```
  0c5919ef
28 Oct, 2019 1 commit
- Add Resnet50 benchmark suite that read training data from remote storage · 06f22a59
  Zongwei Zhou authored Oct 28, 2019
```
PiperOrigin-RevId: 277082247
```
  06f22a59
21 Oct, 2019 1 commit
- Fix typo in utils/flags/_conventions.py · c3bf4a79
  minoring authored Oct 21, 2019
```
arparse -> argparse
```
  c3bf4a79
16 Oct, 2019 1 commit

Add support for the tf.keras.mixed_precision API in NCF · cb913691

Reed Wanderman-Milne authored Oct 16, 2019

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

10 Oct, 2019 1 commit

change benchmark's log verbosity to logging.INFO. it seems to me that DEBUG... · 1f9efe10

A. Unique TensorFlower authored Oct 10, 2019

change benchmark's log verbosity to logging.INFO. it seems to me that DEBUG map to ---v=1 internally, which is way to verbose for the purpose of benchmarking.

PiperOrigin-RevId: 274040907

1f9efe10

07 Oct, 2019 1 commit
- Internal change · 41293260
  A. Unique TensorFlower authored Oct 07, 2019
```
PiperOrigin-RevId: 273371605
```
  41293260
24 Sep, 2019 1 commit
- Use experimental_connect_to_cluster API in TPU lib to support training on a slice of a TPU pod. · 497989e0
  Bruce Fontaine authored Sep 24, 2019
```
PiperOrigin-RevId: 270926016
```
  497989e0
09 Sep, 2019 1 commit

Unexpose some flags from models which do not use them. · e91c41c2

Reed Wanderman-Milne authored Sep 09, 2019

--stop_threshold, --num_gpu, --hooks, --export_dir, and --distribution_strategy have been unexposed from models which do not use them

PiperOrigin-RevId: 268032080

e91c41c2

04 Sep, 2019 1 commit

Unexpose some flags from models which do not use them. · a85c40e3

Reed Wanderman-Milne authored Sep 03, 2019

--clean, --train_epochs, and --epochs_between_evals have been unexposed from models which do not use them

PiperOrigin-RevId: 267065651

a85c40e3

30 Aug, 2019 1 commit
- Fix bug where dynamic loss scaling was broken · 765da424
  Reed Wanderman-Milne authored Aug 30, 2019
```
PiperOrigin-RevId: 266376708
```
  765da424
26 Aug, 2019 1 commit

Unexpose some flags from models which do not use them. · 560b3af4

Reed Wanderman-Milne authored Aug 26, 2019

--synthetic_data, --dtype, --all_reduce_alg, and --num_packs have been unexposed from models which do not use them

PiperOrigin-RevId: 265483564

560b3af4

23 Aug, 2019 2 commits
- Switch wide-deep movielens to use a threadpool rather than a forkpool. · 1d610ef9
  Taylor Robie authored Aug 23, 2019
```
PiperOrigin-RevId: 265114251
```
  1d610ef9
- Unexpose some flags from models which do not use them. · 882e51a4
  Reed Wanderman-Milne authored Aug 22, 2019
```
--num_parallel_calls, --inter_op_parallelism_threads, and --intra_op_parallelism_threads have been unexposed from models which do not use them

PiperOrigin-RevId: 264965788
```
  882e51a4
21 Aug, 2019 1 commit
- Mute no-name-in-module · dd03f167
  Hongkun Yu authored Aug 20, 2019
```
PiperOrigin-RevId: 264527204
```
  dd03f167
20 Aug, 2019 3 commits
- fix transformer amp · c9c05e9b
  Vinh Nguyen authored Aug 20, 2019
  
  c9c05e9b
- change default fp16_implementation to graph_rewrite · c186c85a
  Vinh Nguyen authored Aug 20, 2019
  
  c186c85a
- Move Estimator-only export utils to R1 · a53371eb
  Hongkun Yu authored Aug 19, 2019
```
PiperOrigin-RevId: 264300408
```
  a53371eb
19 Aug, 2019 2 commits

Do not expose --max_train_steps in models that do not use it. · 824ff2d6

Reed Wanderman-Milne authored Aug 19, 2019

Only the V1 resnet model uses --max_train_steps. This unexposes the flag in the keras_application_models, mnist, keras resnet, CTL resnet Models. Before this change, such models allowed the flag to be specified, but ignored it.

I also removed the "max_train" argument from the run_synthetic function, since this only had any meaning for the V1 resnet model. Instead, the V1 resnet model now directly passes --max_train_steps=1 to run_synthetic.

PiperOrigin-RevId: 264269836

824ff2d6

Extend synthetic data monkey patch to MultiWorkerMirroredStrategy. · ee584397
Ayush Dubey authored Aug 19, 2019
```
PiperOrigin-RevId: 264244022
```
ee584397