提交 · 27fb855b027ead16d2616dcb59c67409a2176b7f · 曾经的那一瞬间 / Models

19 1月, 2022 2 次提交
- L
  Internal change · 27fb855b
  由 Le Hou 提交于 1月 18, 2022
```
PiperOrigin-RevId: 422638136
```
  27fb855b
- A
  Internal change · 871c4e0a
  由 A. Unique TensorFlower 提交于 1月 18, 2022
```
PiperOrigin-RevId: 422612138
```
  871c4e0a
18 11月, 2021 1 次提交
- H
  Deprecate the graph rewrite path for fp16. This is no longer a TF2 api and there is no usage. · 51cb03b0
  由 Hongkun Yu 提交于 11月 17, 2021
```
PiperOrigin-RevId: 410629444
```
  51cb03b0
10 4月, 2021 1 次提交

Remove dynamic_loss_scale argument to define_performance. · e353e4e5

由 Reed Wanderman-Milne 提交于 4月 09, 2021

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

e353e4e5

10 3月, 2021 1 次提交
- F
  Internal change · 5a898973
  由 Frederick Liu 提交于 3月 09, 2021
```
PiperOrigin-RevId: 361957289
```
  5a898973
29 8月, 2020 1 次提交
- Z
  Internal change · 3f8132e9
  由 Zongwei Zhou 提交于 8月 28, 2020
```
PiperOrigin-RevId: 329042049
```
  3f8132e9
28 8月, 2020 2 次提交
- H
  [Minor Cleanup] Move clip_by_global_norm_callback to model_training_utils · 88effbc3
  由 Hongkun Yu 提交于 8月 27, 2020
```
PiperOrigin-RevId: 328888268
```
  88effbc3
- Z
  Internal change · 3fc70674
  由 Zongwei Zhou 提交于 8月 27, 2020
```
PiperOrigin-RevId: 328842362
```
  3fc70674
13 8月, 2020 2 次提交
- H
  Internal change · 999fae62
  由 Hongkun Yu 提交于 8月 12, 2020
```
PiperOrigin-RevId: 326286926
```
  999fae62
- H
  Internal change · 88253ce5
  由 Hongkun Yu 提交于 8月 12, 2020
```
PiperOrigin-RevId: 326286926
```
  88253ce5
26 5月, 2020 1 次提交

Add --sub_model_export_name to run_classifier and run_squad. · 38800db3

由 André Susano Pinto 提交于 5月 26, 2020

This allows one to finetune a BERT model into a task before using it for
another task. E.g. SQuAD before finetune another QA type of tasks.

PiperOrigin-RevId: 313145768

38800db3

19 5月, 2020 1 次提交

When not specified by user, make steps_per_loop be dependent on the devices available. · 866176ea

由 André Susano Pinto 提交于 5月 19, 2020

A default of 1 all the times is bad for TPU users which end up not using the device effectively.
A larger default all the times is bad for GPU users.

So compromise and make this dependent on the devices available.

PiperOrigin-RevId: 312230371

866176ea

05 5月, 2020 1 次提交
- H
  Move gin flags to hyperparams_flags are they are in the same category and we... · d93a1206
  由 Hongkun Yu 提交于 5月 04, 2020
```
Move gin flags to hyperparams_flags are they are in the same category and we will use them more widely.

PiperOrigin-RevId: 309779408
```
  d93a1206
17 4月, 2020 1 次提交
- H
  Remove hooks. · 2ddd716e
  由 Hongkun Yu 提交于 4月 16, 2020
```
PiperOrigin-RevId: 306994199
```
  2ddd716e
11 4月, 2020 1 次提交
- A
  Add end learning rate (end_lr) FLAG for linear learning rate decay. · 22de957a
  由 A. Unique TensorFlower 提交于 4月 10, 2020
```
PiperOrigin-RevId: 305948522
```
  22de957a
28 3月, 2020 1 次提交
- A
  Support Lamb optimizer within BERT. · da5860f2
  由 A. Unique TensorFlower 提交于 3月 27, 2020
```
PiperOrigin-RevId: 303356961
```
  da5860f2
26 3月, 2020 1 次提交
- H
  Internal change · 38486725
  由 Hongkun Yu 提交于 3月 25, 2020
```
PiperOrigin-RevId: 302977474
```
  38486725
12 3月, 2020 1 次提交
- W
  Add TimeHistory callback to BERT. · b39958cf
  由 Will Cromar 提交于 3月 11, 2020
```
PiperOrigin-RevId: 300433601
```
  b39958cf
08 3月, 2020 1 次提交
- H
  Add TimeHistory callback to BERT. · 7d86c317
  由 Hongkun Yu 提交于 3月 07, 2020
```
PiperOrigin-RevId: 299594839
```
  7d86c317
06 3月, 2020 1 次提交
- H
  Internal change · 2565629c
  由 Hongkun Yu 提交于 3月 05, 2020
```
PiperOrigin-RevId: 299169021
```
  2565629c
03 3月, 2020 1 次提交
- W
  Add TimeHistory callback to BERT. · 533d1e6b
  由 Will Cromar 提交于 3月 02, 2020
```
PiperOrigin-RevId: 298466825
```
  533d1e6b
27 2月, 2020 1 次提交
- H
  Internal change · ce83a9db
  由 Hongkun Yu 提交于 2月 26, 2020
```
PiperOrigin-RevId: 297383836
```
  ce83a9db
26 2月, 2020 1 次提交
- Z
  Internal change · 2781377d
  由 Zongwei Zhou 提交于 2月 25, 2020
```
PiperOrigin-RevId: 297222995
```
  2781377d
25 2月, 2020 2 次提交
- H
  Creates modeling/performance.py to include mix prediction related stuff · fb35d6be
  由 Hongkun Yu 提交于 2月 24, 2020
```
PiperOrigin-RevId: 297002741
```
  fb35d6be
- C
  Branch a run_squad.py for albert. · 02d78796
  由 Chen Chen 提交于 2月 24, 2020
```
PiperOrigin-RevId: 296933982
```
  02d78796
19 2月, 2020 1 次提交
- A
  Internal change · 2926ba69
  由 A. Unique TensorFlower 提交于 2月 18, 2020
```
PiperOrigin-RevId: 295757618
```
  2926ba69
31 1月, 2020 1 次提交
- A
  BERT enable gpu thread mode as "gpu_private" for 8-GPU runs. · 2741cc5f
  由 A. Unique TensorFlower 提交于 1月 30, 2020
```
PiperOrigin-RevId: 292369407
```
  2741cc5f
29 1月, 2020 1 次提交
- Z
  Enable all_reduce_alg flag · 4f3acbda
  由 Zongwei Zhou 提交于 1月 28, 2020
```
PiperOrigin-RevId: 292029030
```
  4f3acbda
28 1月, 2020 1 次提交
- Y
  Add multi-worker dist-strat benchmark/accuracy for BERT SQUAD model. · d4ef56b1
  由 Yanhui Liang 提交于 1月 27, 2020
```
PiperOrigin-RevId: 291851815
```
  d4ef56b1
11 12月, 2019 1 次提交
- C
  Internal change · f7fd59b8
  由 Chen Chen 提交于 12月 10, 2019
```
PiperOrigin-RevId: 284792715
```
  f7fd59b8
20 11月, 2019 1 次提交
- H
  Use flags utils and distribution_utils · 252e6384
  由 Hongkun Yu 提交于 11月 19, 2019
```
PiperOrigin-RevId: 281337671
```
  252e6384
06 11月, 2019 1 次提交
- H
  Fix a comment. · 0203278a
  由 Hongkun Yu 提交于 11月 05, 2019
```
PiperOrigin-RevId: 278672795
```
  0203278a
11 10月, 2019 1 次提交
- H
  Bert Hub usage for squad · 223f5148
  由 Hongkun Yu 提交于 10月 10, 2019
```
PiperOrigin-RevId: 274090672
```
  223f5148
09 10月, 2019 1 次提交
- A
  Internal change · 7207422d
  由 A. Unique TensorFlower 提交于 10月 08, 2019
```
PiperOrigin-RevId: 273653001
```
  7207422d
18 9月, 2019 1 次提交
- H
  Move Bert to NLP. Tasks are moved to nlp/bert/ · 1862b9c3
  由 Hongkun Yu 提交于 9月 17, 2019
```
Refactor basic utils to modeling/

PiperOrigin-RevId: 269600561
```
  1862b9c3
04 9月, 2019 1 次提交
- H
  move collection trainable variables outside loop. · a009f4fb
  由 Hongkun Yu 提交于 9月 03, 2019
```
add a flag to control loss scaling.

PiperOrigin-RevId: 267091566
```
  a009f4fb
03 9月, 2019 1 次提交
- V
  
  adding fp16 implementation for BERT · 0f6ff657
  由 Vinh Nguyen 提交于 9月 03, 2019
  
  0f6ff657
27 8月, 2019 1 次提交
- H
  Internal change · f93229b9
  由 Hongkun Yu 提交于 8月 26, 2019
```
PiperOrigin-RevId: 265510206
```
  f93229b9
23 8月, 2019 1 次提交
- A
  Add flags to run BERT with MultiWorkerMirroredStrategy. · 39aea35f
  由 Ayush Dubey 提交于 8月 22, 2019
```
PiperOrigin-RevId: 264935345
```
  39aea35f
08 8月, 2019 1 次提交

Merged commit includes the following changes: (#7404) · e38d570e

由 Hongkun Yu 提交于 8月 07, 2019

262178259  by hongkuny<hongkuny@google.com>:

    We should call training=True in CTL train step.

--
262081759  by akuegel<akuegel@google.com>:

    Internal change

PiperOrigin-RevId: 262178259

e38d570e

曾经的那一瞬间 / Models 大约 1 年 前同步成功

曾经的那一瞬间 / Models
大约 1 年前同步成功