提交 · d40155f0ba2b6e5bbef53cd0292bb525d9768158 · weixin_51669992 / tensorflow

02 11月, 2020 15 次提交

Do not allow dilated convolution when there are dynamic dims either at first or · d40155f0

由 Jaesung Chung 提交于 11月 02, 2020

last dim.

Also fixed the shape inference code for SpaceToBatchNDOp in order to correctly
handle the given inputs, that contain dynamic dimensions.

PiperOrigin-RevId: 340226617
Change-Id: I054f4ba794b5eb206dba69a67a80eb25514b4a27

d40155f0

A
Fix forward broken test · 8520302c
由 A. Unique TensorFlower 提交于 11月 02, 2020
```
PiperOrigin-RevId: 340218738
Change-Id: Ibfe70403d8fb9f5d351ec34fdf8a85a8143abda0
```
8520302c
T
Merge pull request #44456 from mansnils:stm32f4 · 67f54bb8
由 TensorFlower Gardener 提交于 11月 02, 2020
```
PiperOrigin-RevId: 340200935
Change-Id: I7192c5e05d25e29e4d1c7f268494ddcb09d0ca0c
```
67f54bb8

PR #43950: Print the version of the cuDNN library found at runtime · c289aa97

由 Sanjoy Das 提交于 11月 02, 2020

Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/43950

This is especially useful if we build against X.Y.z and run with X.Y.w
where w != z.
Copybara import of the project:

--
e5183817a249adc6998e7dd29c70974bf2b61221 by Sanjoy Das <sanjoy@playingwithpointers.com>:

Print the version of the cuDNN library found at runtime

This is especially useful if we build against X.Y.z and run with X.Y.w
where w != z.

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/tensorflow/pull/43950 from sanjoy:print-cudnn-version e5183817a249adc6998e7dd29c70974bf2b61221
PiperOrigin-RevId: 340187371
Change-Id: If7251624ed2599ced7ac50f8f5e501bdf46e6b59

c289aa97

A
compat: Update forward compatibility horizon to 2020-11-02 · 942e7121
由 A. Unique TensorFlower 提交于 11月 02, 2020
```
PiperOrigin-RevId: 340184983
Change-Id: Ia7e047792256386b3c8ea4087855d00f66d198cb
```
942e7121
A
Update GraphDef version to 573. · dea86531
由 A. Unique TensorFlower 提交于 11月 02, 2020
```
PiperOrigin-RevId: 340184980
Change-Id: Ifb2eeee36053b6d9bbcc21a1e4bc01f504ddcf8a
```
dea86531
T
Merge pull request #43148 from Tessil:toupstream/ensure_all_16x8_operators_check_null_zero_point · 0e488497
由 TensorFlower Gardener 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340171268
Change-Id: Ifcd3ea5b57b1c97f9a020270d472bd28751ceb07
```
0e488497

Do not legalize tf.Conv2DOp when the filter type is not determined. · c33f9482

由 Jaesung Chung 提交于 11月 01, 2020

Currently, the converter generates a wrong constant when the condition is hold.

PiperOrigin-RevId: 340170033
Change-Id: Ia4af732f456254738c76840a32673ecff5f0fb6a

c33f9482

Try to fix windows build failure. · 7d4e7b86

由 Scott Zhu 提交于 11月 01, 2020

The "\" might be converted to "\\" in windows and causing copybara failure during conversion.

PiperOrigin-RevId: 340164770
Change-Id: I0def5626427735e60e8bff377e35f3998335ca4f

7d4e7b86

T
Merge pull request #41344 from wwwind:16x8_conv_activation_op_tests · c40d0d3b
由 TensorFlower Gardener 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340161487
```
c40d0d3b

Mount host's current dirrectory to /host_dir instead of /tmp · e500cde7

由 Thai Nguyen 提交于 11月 01, 2020

When mounting to /tmp, temporary files in the container is saved
to host directory, making it messy even after container exited.

PiperOrigin-RevId: 340155058
Change-Id: I87b1358875f0df0319d6d9572ae226c6b8aca9df

e500cde7

A
Internal change · 1d5d379f
由 A. Unique TensorFlower 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340148964
Change-Id: Ib0aa3d63285ee5a74b3ca002f4a220ce8919cf79
```
1d5d379f
A
Internal change · b5db1ca4
由 A. Unique TensorFlower 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340146798
Change-Id: I22b6f23e305c21c7c84d1ad2c0f22ee6fbef1343
```
b5db1ca4
T
Merge pull request #41344 from wwwind:16x8_conv_activation_op_tests · 3ac50899
由 TensorFlower Gardener 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340146286
```
3ac50899
T
Add back support for tracking variables from submetrics of a Metric. · b2c3e950
由 Thomas O'Malley 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340120524
Change-Id: I76d5880c95aafe7a6a9f64b5be1dc23af464e988
```
b2c3e950

01 11月, 2020 4 次提交
- A
  Update GraphDef version to 572. · 898ae24a
  由 A. Unique TensorFlower 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340089160
Change-Id: Id65ad44864afbdbac7c969ff44feb6cb2433ea2a
```
  898ae24a
- A
  compat: Update forward compatibility horizon to 2020-11-01 · 42d9d640
  由 A. Unique TensorFlower 提交于 11月 01, 2020
```
PiperOrigin-RevId: 340089158
Change-Id: I05383a31dc403068576b7feb02cef8b01bae046f
```
  42d9d640
- J
  [tf.data] Minor cleanup · 6679bc94
  由 Jiri Simsa 提交于 10月 31, 2020
```
PiperOrigin-RevId: 340071266
Change-Id: Ic21209a25a1f8efa1122c9cee4a8ab3b8043c308
```
  6679bc94
- M
  Remove keras dependency from TFLite tests and BUILD files · d3bcfc94
  由 Meghna Natraj 提交于 10月 31, 2020
```
PiperOrigin-RevId: 340033160
Change-Id: Ia7007ddb1bcd128c684712eafa284af79fedac05
```
  d3bcfc94
31 10月, 2020 21 次提交

X
Disable flaky timing out tests (MWMS host training loop). · 89b07879
由 Xinyi Wang 提交于 10月 31, 2020
```
PiperOrigin-RevId: 340026157
Change-Id: Ibf94e17f0cd2cf88738d1d08be06cbe68a848ecf
```
89b07879

tf.parallel_stack is implemented as a graph rewrite and therefore has no... · 868256f0

由 Rohan Jain 提交于 10月 31, 2020

tf.parallel_stack is implemented as a graph rewrite and therefore has no support in eager mode. Added documentation to parallel_stack and raising an error now to reflect that.

PiperOrigin-RevId: 340016749
Change-Id: I7c695dfb98549bbc56105592b1b1c6b1ea06b231

868256f0

A
Update GraphDef version to 571. · 377434ff
由 A. Unique TensorFlower 提交于 10月 31, 2020
```
PiperOrigin-RevId: 340005180
Change-Id: Id2be30ca6dc141ef6c51113e0ef915d03568ba3b
```
377434ff
A
compat: Update forward compatibility horizon to 2020-10-31 · b9598b65
由 A. Unique TensorFlower 提交于 10月 31, 2020
```
PiperOrigin-RevId: 340005178
Change-Id: Ic63905f93733940c57d64156c502b95f067bba4f
```
b9598b65
Y
[XLA:SPMD] Fix contracting dim loop einsum threshold · 5b76abd4
由 Yuanzhong Xu 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339989040
Change-Id: I46eaffb4e4b28c168b5f0182c4c0bb946d6eefe8
```
5b76abd4
S
Fork the tf function related util to keras. · 05483cd4
由 Scott Zhu 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339983364
Change-Id: I877f2394f13b899ace0bb2893e6cb5f073b03458
```
05483cd4

Update cmake support of TFLite minimal example · 0509af7e

由 Terry Heo 提交于 10月 30, 2020

Created a CMakeLists.txt under the lite/examples/minimal

Usage)

$ git clone https://github.com/tensorflow/tensorflow.git tensorflow_src
$ mkdir minimal_build
$ cd miniaml_build
$ cmake ../tensorflow_src/tensorflow/lite/examples/minimal
$ cmake --build . -j

PiperOrigin-RevId: 339977906
Change-Id: I85cada5430407e72354d0a31a17584c48407db2f

0509af7e

[XLA] Don't unnecessarily process buffers that got default mem allocations. · 92bb7896

由 Berkin Ilbeyi 提交于 10月 30, 2020

When processing buffers that are placed in tuples, we create (possibly
redundant) additional tuple and gte instructions, which we later simplify. We
don't need to create these additional instructions for buffers that were
unchanged after memory space assignment. We were facing compile time and OOM
issues without this change.

PiperOrigin-RevId: 339972236
Change-Id: Ieb82767ed52d39260141f3e3179387fbf5783920

92bb7896

R
ConvConstants generation changed to use storage type properties instead of specific storage types. · 086bd64e
由 Raman Sarokin 提交于 10月 30, 2020
```
ConvConstants extended for better mapping on thin dst tensors.

PiperOrigin-RevId: 339971199
Change-Id: Ia14be12361ec81c70c0522baa9516c8cb28ba7ff
```
086bd64e

ConvolutionTransposed3x3 generation changed to use storage type properties... · 1f2e88a9

由 Raman Sarokin 提交于 10月 30, 2020

ConvolutionTransposed3x3 generation changed to use storage type properties instead of specific storage types.

PiperOrigin-RevId: 339967056
Change-Id: I5f6e192e21cbabf33bd152f91f4ab664b139fd14

1f2e88a9

Multi-worker tutorial: Add the workflow of MWMS+CTL example that is going to... · 64edb2fb

由 Rick Chao 提交于 10月 30, 2020

Multi-worker tutorial: Add the workflow of MWMS+CTL example that is going to be added in the tutorial in multi_worker_tutorial_test.

Fix the flakiness of the test and re-enable in TAP.

PiperOrigin-RevId: 339966024
Change-Id: Icb866f8a7054fa88f2e474c02960982a57c542b3

64edb2fb

Fix tape recording of attributes of vectorized ops · 3d1f1b06

由 Allen Lavoie 提交于 10月 30, 2020

nest.flatten was mangling some attribute lists

Fixes #40895.

PiperOrigin-RevId: 339965951
Change-Id: Ibdbbb5a12b2d06a077b62902764ca4849a2b8d2a

3d1f1b06

DepthwiseConvolution generation changed to use storage type properties instead... · b37b0d00

由 Raman Sarokin 提交于 10月 30, 2020

DepthwiseConvolution generation changed to use storage type properties instead of specific storage types.

PiperOrigin-RevId: 339964553
Change-Id: I2cded9c306a40b136002c08e610daca2d75e1758

b37b0d00

R
environment renamed to device_info and changed to cc target (instead of objc). · 50ec85cd
由 Raman Sarokin 提交于 10月 30, 2020
```
Added A14 GPU to enum.

PiperOrigin-RevId: 339964411
Change-Id: I13439c83c6f3d0f75a166b950a5f088843dd2023
```
50ec85cd
M
Fix failing tests in model_coverage_lib_test.py · 41dc507d
由 Meghna Natraj 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339959753
Change-Id: I492d1f750e7a6fd97d4fdccbb838492e699c56b6
```
41dc507d
A
Fix and re-enable cwise ops benchmark. · 3ad80ca3
由 Anna R 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339956677
Change-Id: Id4cea5f63b3f257170c121d74e9cb2995a10b629
```
3ad80ca3
M
Refactor keras dependency code · 0b752d92
由 Meghna Natraj 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339954904
Change-Id: Id9f6717da5c32bff185a10a37e9682be64cc6501
```
0b752d92
N
Add int16->int32 quantization to TFLM quantize op. · 431b12d9
由 Nat Jeffries 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339953940
Change-Id: I568589fc09fa6e5aae137e7e8533e1dab86190d3
```
431b12d9
M
Update MobileNetV2 models for model_coverage tests · 8e229bb6
由 Meghna Natraj 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339951415
Change-Id: Ic3bbb2e474a4fb287fb48f24509beb44625ee2e4
```
8e229bb6
A
Add stream executor C API symbols to exported symbols for _pywrap_tensorflow_internal.so. · f296fd6f
由 Anna R 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339951343
Change-Id: I02b02143721f7c656af8c1d19176dc44f3bffa20
```
f296fd6f
S
Fork the object_identity related class to Keras. · f4280211
由 Scott Zhu 提交于 10月 30, 2020
```
PiperOrigin-RevId: 339948109
Change-Id: Ia2612efa66b48ac5784d2a53cdeb0ce9001f4025
```
f4280211

weixin_51669992 / tensorflow 与 Fork 源项目一致

weixin_51669992 / tensorflow
与 Fork 源项目一致