Commits · 5eb294f84bd3f415b548980e69fee63db1f6f1df · ModelZoo / ResNet50_tensorflow

31 Jul, 2020 2 commits
- Internal change · 5eb294f8
  Abdullah Rashwan authored Jul 30, 2020
```
PiperOrigin-RevId: 324140487
```
  5eb294f8
- Internal change · a62c2bfc
  Hongkun Yu authored Jul 30, 2020
```
PiperOrigin-RevId: 324137297
```
  a62c2bfc
30 Jul, 2020 7 commits

Internal change · 2e785497
A. Unique TensorFlower authored Jul 30, 2020
```
PiperOrigin-RevId: 324084377
```
2e785497
func name change · 673231ff
xinliupitt authored Jul 30, 2020

673231ff
transformer test case · 72efa854
xinliupitt authored Jul 30, 2020

72efa854
get_config and doc · 1c5dca7f
xinliupitt authored Jul 30, 2020

1c5dca7f
Internal change · 5366f605
A. Unique TensorFlower authored Jul 29, 2020
```
PiperOrigin-RevId: 323948101
```
5366f605

Pre-emptively disable the KerasTensors refactoring for the detection models in... · ccc60760

Tomer Kaftan authored Jul 29, 2020

Pre-emptively disable the KerasTensors refactoring for the detection models in tensorflow models/official/vision/detection, because they rely on several unsupported things that will stop working entirely when the refactoring goes live.

Specifically:
* The custom layers implement `__call__` instead of `call` and rely on manually enter the keras backend graph
* The vision models try to use `tf.while_loop` as Keras op layers during functional API construction, which is unsupported.

Updating the models to avoid this would subtly change the variable names and break the pre-existing tf1-style name-based checkpoints, so for now we will just disable the KerasTensors refactoring for these models.

PiperOrigin-RevId: 323937426

ccc60760

Remove backward compatibility hack. · 52504780
Hongkun Yu authored Jul 29, 2020
```
PiperOrigin-RevId: 323927843
```
52504780

29 Jul, 2020 7 commits
- whitespace · 3e0fa932
  xinliupitt authored Jul 29, 2020
  
  3e0fa932
- docstrings · d39321b1
  xinliupitt authored Jul 29, 2020
  
  d39321b1
- transformer, attention layers · 23804bc5
  xinliupitt authored Jul 29, 2020
  
  23804bc5
- Create output directories when converting bert checkpoint. · bda18166
  André Susano Pinto authored Jul 29, 2020
```
PiperOrigin-RevId: 323827690
```
  bda18166
- Internal change · a565d720
  A. Unique TensorFlower authored Jul 29, 2020
```
PiperOrigin-RevId: 323732686
```
  a565d720
- Remove TokenClassication network which is a single dense layer. · 250701c6
  Hongkun Yu authored Jul 28, 2020
```
PiperOrigin-RevId: 323711016
```
  250701c6
- Use absl app. · 5a1b5af3
  Hongkun Yu authored Jul 28, 2020
```
PiperOrigin-RevId: 323685656
```
  5a1b5af3
28 Jul, 2020 5 commits
- Internal change · ab2c8ba9
  Hongkun Yu authored Jul 28, 2020
```
PiperOrigin-RevId: 323653596
```
  ab2c8ba9
- Remove oauth2client version · 902e05aa
  Hongkun Yu authored Jul 28, 2020
```
PiperOrigin-RevId: 323644473
```
  902e05aa
- Internal change · 322d2eb7
  A. Unique TensorFlower authored Jul 28, 2020
```
PiperOrigin-RevId: 323627192
```
  322d2eb7
- Internal change · a540c3d7
  A. Unique TensorFlower authored Jul 27, 2020
```
PiperOrigin-RevId: 323515364
```
  a540c3d7
- Migrate to tf gelu. Will remove activations/gelu.py after dependencies are cleaned up. · f8c2a917
  Hongkun Yu authored Jul 27, 2020
```
PiperOrigin-RevId: 323499265
```
  f8c2a917
27 Jul, 2020 1 commit
- Internal change · e9b70e67
  A. Unique TensorFlower authored Jul 27, 2020
```
PiperOrigin-RevId: 323430879
```
  e9b70e67
24 Jul, 2020 5 commits
- Internal change · 2e96fbee
  Allen Wang authored Jul 24, 2020
```
PiperOrigin-RevId: 323016002
```
  2e96fbee
- Change the dataset_transform_fn argument in InputReader's constructor to transform_and_batch_fn. · 2659ca30
  Chen Chen authored Jul 24, 2020
```
PiperOrigin-RevId: 323013252
```
  2659ca30
- Fixed typos · b3f63b00
  Srihari Humbarwadi authored Jul 24, 2020
```
 - Fixed `intermediate_scale` in  `Anchor`
 - "ratio" and "divisible" in doc string
```
  b3f63b00
- Fix typo. · bdaa525b
  Hongkun Yu authored Jul 24, 2020
```
PiperOrigin-RevId: 322960082
```
  bdaa525b
- Internal change · ce93cb75
  A. Unique TensorFlower authored Jul 23, 2020
```
PiperOrigin-RevId: 322923416
```
  ce93cb75
23 Jul, 2020 2 commits
- Internal change · b91a9481
  Allen Wang authored Jul 23, 2020
```
PiperOrigin-RevId: 322884473
```
  b91a9481
- Internal change · a829e648
  A. Unique TensorFlower authored Jul 23, 2020
```
PiperOrigin-RevId: 322844988
```
  a829e648
22 Jul, 2020 3 commits
- Internal change · 7fcd7cba
  Allen Wang authored Jul 22, 2020
```
PiperOrigin-RevId: 322657835
```
  7fcd7cba
- Allow export only path takes the hub module. · acaef198
  Hongkun Yu authored Jul 22, 2020
```
PiperOrigin-RevId: 322594580
```
  acaef198
- Explicit set the dtype in embedding layer. Otherwise, using the layer in TF1.x... · 567bd18d
  A. Unique TensorFlower authored Jul 21, 2020
```
Explicit set the dtype in embedding layer. Otherwise, using the layer in TF1.x will create an int32 tensor.

PiperOrigin-RevId: 322473236
```
  567bd18d
21 Jul, 2020 4 commits
- [MultiheadAttention] Apply suggestions from RFC: · 36101ab4
  Hongkun Yu authored Jul 21, 2020
```
(1) call() consume kwargs and implement _build_from_signature (layers added inside init_scope)
(2) make build/call_attention as public.

PiperOrigin-RevId: 322454619
```
  36101ab4
- Internal change · 391c9dd6
  A. Unique TensorFlower authored Jul 21, 2020
```
PiperOrigin-RevId: 322428923
```
  391c9dd6
- Fix NCF input pipeline to avoid reading the same file multiple times in one epoch. · 6a2de9bb
  Bruce Fontaine authored Jul 21, 2020
```
PiperOrigin-RevId: 322415899
```
  6a2de9bb
- Remove some comments out-dated. · f97e0231
  Hongkun Yu authored Jul 21, 2020
```
PiperOrigin-RevId: 322387911
```
  f97e0231
20 Jul, 2020 4 commits

Add in RetinaNet MLIR benchmarks. · f05df686
Allen Wang authored Jul 20, 2020
```
PiperOrigin-RevId: 322244056
```
f05df686
Internal change · 7e3fbeee
Allen Wang authored Jul 20, 2020
```
PiperOrigin-RevId: 322234532
```
7e3fbeee

Make hack in official/vision/detection models that enters the backend keras... · 363a36cd

Tomer Kaftan authored Jul 20, 2020

Make hack in official/vision/detection models that enters the backend keras graph stop happening once we enable the Functional API KerasTensors refactoring:

As a workaround hack for the tf op layer conversion being fragile, the detection models have to explicitly enter the Keras backend graph.

When we enable the KerasTensors refactoring of the Functional API internals, the op layer conversion will be much more reliable and this hack will not be necessary. In addition, the hack actually causes the models to break when we enable the refactoring (because it causes tensors to leak out of a graph).

So, this CL changes the existing hack to stop applying once we've enabled the KerasTensors refactoring.

PiperOrigin-RevId: 322229802

363a36cd

Internal change · cdda0906
Allen Wang authored Jul 20, 2020
```
PiperOrigin-RevId: 322216928
```
cdda0906