Commits · bdf8299a7d14c972bd90329b7a997931e0797132 · ModelZoo / ResNet50_tensorflow

15 Oct, 2020 1 commit
- Remove repeating prefetch. · bdf8299a
  Hongkun Yu authored Oct 14, 2020
```
PiperOrigin-RevId: 337242735
```
  bdf8299a
14 Oct, 2020 2 commits
- Implement XLNet QA model. · 44fc54a1
  Allen Wang authored Oct 14, 2020
```
PiperOrigin-RevId: 337134894
```
  44fc54a1
- Internal change · 41bcd7d0
  Hongkun Yu authored Oct 13, 2020
```
PiperOrigin-RevId: 336991213
```
  41bcd7d0
13 Oct, 2020 10 commits
- Add sample_fn to input_reader. · d32be917
  A. Unique TensorFlower authored Oct 13, 2020
```
PiperOrigin-RevId: 336973004
```
  d32be917
- Internal change · 2b2d4820
  A. Unique TensorFlower authored Oct 13, 2020
```
PiperOrigin-RevId: 336967249
```
  2b2d4820
- Internal change · f041db19
  Hongkun Yu authored Oct 13, 2020
```
PiperOrigin-RevId: 336960641
```
  f041db19
- Internal change · a027c8a6
  Abdullah Rashwan authored Oct 13, 2020
```
PiperOrigin-RevId: 336951637
```
  a027c8a6
- Internal change · 0ef802a5
  A. Unique TensorFlower authored Oct 13, 2020
```
PiperOrigin-RevId: 336942308
```
  0ef802a5
- Internal change · 9376c567
  Abdullah Rashwan authored Oct 13, 2020
```
PiperOrigin-RevId: 336939703
```
  9376c567
- Slice Transformer outputs with a TF Op layer instead of a Lambda layer. · 5879ca04
  A. Unique TensorFlower authored Oct 13, 2020
```
Lambda layers are fundamentally non-portable since they serialize
Python bytecode.

PiperOrigin-RevId: 336917172
```
  5879ca04
- Internal change · 575c7458
  A. Unique TensorFlower authored Oct 13, 2020
```
PiperOrigin-RevId: 336914823
```
  575c7458
- Slice BERT output with a TF Op layer instead of a Lambda layer. · eb4f8124
  A. Unique TensorFlower authored Oct 13, 2020
```
Lambda layers are fundamentally non-portable since they serialize
Python bytecode.

PiperOrigin-RevId: 336910013
```
  eb4f8124
- Move core configs to core/ folder. Leave the configs for legacy models. · d6fe2516
  Hongkun Yu authored Oct 12, 2020
```
PiperOrigin-RevId: 336795303
```
  d6fe2516
12 Oct, 2020 4 commits
- Merge pull request #9317 from tensorflow:mobilenet-beta · caa61e1b
  A. Unique TensorFlower authored Oct 12, 2020
```
PiperOrigin-RevId: 336771016
```
  caa61e1b
- Internal change · cd4b8c23
  A. Unique TensorFlower authored Oct 12, 2020
```
PiperOrigin-RevId: 336717785
```
  cd4b8c23
- Reduce batch size of resnet fp16 eager benchmark to prevent occasional OOM · 64755c7f
  Zongwei Zhou authored Oct 12, 2020
```
PiperOrigin-RevId: 336715941
```
  64755c7f
- Implement SpanLabeler for XLNet. · 3e74d7e1
  Allen Wang authored Oct 12, 2020
```
PiperOrigin-RevId: 336709640
```
  3e74d7e1
09 Oct, 2020 4 commits
- Internal change · 98c20111
  Zhenyu Tan authored Oct 09, 2020
```
PiperOrigin-RevId: 336353171
```
  98c20111
- Fix doc string after the code is updated. · 5bc5d4e6
  Hongkun Yu authored Oct 09, 2020
```
PiperOrigin-RevId: 336339472
```
  5bc5d4e6
- Internal change · 52c4ffe9
  Abdullah Rashwan authored Oct 08, 2020
```
PiperOrigin-RevId: 336229644
```
  52c4ffe9
- Internal change · b2f501e3
  Ran Chen authored Oct 08, 2020
```
PiperOrigin-RevId: 336215069
```
  b2f501e3
08 Oct, 2020 6 commits
- Internal change · e53fd183
  Abdullah Rashwan authored Oct 08, 2020
```
PiperOrigin-RevId: 336132820
```
  e53fd183
- Factor out optimizer creation logic from the trainer, so as to avoid access... · 76fffd2a
  Hongkun Yu authored Oct 08, 2020
```
Factor out optimizer creation logic from the trainer, so as to avoid access runtime config inside Trainer.
Add validation method on the ExperimentConfig as clients may use other containers instead of the model garden config dataclass.

PiperOrigin-RevId: 336132223
```
  76fffd2a
- Internal change · cc1f6f6e
  A. Unique TensorFlower authored Oct 08, 2020
```
PiperOrigin-RevId: 336124157
```
  cc1f6f6e
- Change the default validation steps in the config to -1. · 23ed4dd3
  Hongkun Yu authored Oct 08, 2020
```
PiperOrigin-RevId: 336112806
```
  23ed4dd3
- internal change · 2d62cf36
  Chen Chen authored Oct 08, 2020
```
PiperOrigin-RevId: 336043643
```
  2d62cf36
- internal change · e8e987a6
  Chen Chen authored Oct 08, 2020
```
PiperOrigin-RevId: 336042908
```
  e8e987a6
07 Oct, 2020 4 commits
- Fix Typo in XLNet readme · bd73276e
  Allen Wang authored Oct 07, 2020
```
PiperOrigin-RevId: 335973329
```
  bd73276e
- Internal change · 635b089a
  A. Unique TensorFlower authored Oct 07, 2020
```
PiperOrigin-RevId: 335952206
```
  635b089a
- Minor bug fixes · fba8ea2f
  Le Hou authored Oct 07, 2020
```
PiperOrigin-RevId: 335944731
```
  fba8ea2f
- Fix ground truth annotation overflow. · bc1cc8bb
  A. Unique TensorFlower authored Oct 06, 2020
```
PiperOrigin-RevId: 335759679
```
  bc1cc8bb
06 Oct, 2020 6 commits
- Internal change · afb20994
  Abdullah Rashwan authored Oct 06, 2020
```
PiperOrigin-RevId: 335692122
```
  afb20994
- Fix https://github.com/tensorflow/models/issues/9336 · 846a2fff
  Chen Chen authored Oct 06, 2020
```
to support both model and encoder as key in ckpt when exporting to tfhub

PiperOrigin-RevId: 335668823
```
  846a2fff
- Update the default arguments for orbit.Controller train_and_evaluate() and... · 28e3f3c3
  A. Unique TensorFlower authored Oct 06, 2020
```
Update the default arguments for orbit.Controller train_and_evaluate() and evaluate_continuously() to match the default for evaluate().

PiperOrigin-RevId: 335659533
```
  28e3f3c3
- Clean up internal seq2seqTransformer usage. · 00e598f0
  Hongkun Yu authored Oct 05, 2020
```
Add it to export. TODO: move the encoder and decoder layer out to layers/ or decoders/.

PiperOrigin-RevId: 335560860
```
  00e598f0
- add default config for mobilenet v2 that replicate the expected performance · d68c7b96
  Shixin Luo authored Oct 06, 2020
  
  d68c7b96
- Small changes in video classification task. · 1407e274
  Yeqing Li authored Oct 05, 2020
```
PiperOrigin-RevId: 335548133
```
  1407e274
05 Oct, 2020 2 commits

PR #42625: Add training call argument for MultiHeadAttention · 75c6e3bc

Tzu-Wei Sung authored Oct 05, 2020

Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/42625

Miss `training` call argument for dropout layer. /cc @tanzhenyu for visibility.
Copybara import of the project:

--
ba2198fc735a2deca08c68a3c6266d02e01dfe3b by Tzu-Wei Sung <windqaq@gmail.com>:

Add training call argument for MultiHeadAttention

Remove trailing space

PiperOrigin-RevId: 335507127

75c6e3bc

Internal change · ab5d4180
Dan Holtmann-Rice authored Oct 05, 2020
```
PiperOrigin-RevId: 335446217
```
ab5d4180

03 Oct, 2020 1 commit
- Support to convert albert tf1 ckpt to tf2's BertPretrainerV2 · 6cf1e5a1
  Chen Chen authored Oct 03, 2020
```
PiperOrigin-RevId: 335226408
```
  6cf1e5a1