提交 · ad76a4b373d72997584b46abc85af11840cf14db · qq_38905368 / tensorflow

26 2月, 2019 2 次提交
- B
  [XLA:TF] Enable argminmax test on CPU · ad76a4b3
  由 Benjamin Kramer 提交于 2月 25, 2019
```
It's not clear what this comment refers to, but it's from 2017 and the test
just works.

PiperOrigin-RevId: 235539638
```
  ad76a4b3
- T
  Rename ReplicaLocalVariable to SyncOnReadVariable. · 72b4dc9d
  由 Tom Hennigan 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235536186
```
  72b4dc9d
25 2月, 2019 16 次提交
- A
  Replace usage of tf.saved_model.load() with the public symbol on saved model · ebcc4f68
  由 Andr? Susano Pinto 提交于 2月 25, 2019
```
integration tests.

PiperOrigin-RevId: 235526039
```
  ebcc4f68
- C
  Add simple Mnist example on TPU. · 6fac0472
  由 Chris Jones 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235519570
```
  6fac0472
- T
  Make V1 and V2 VariableAggregation enums interchangeable. · 8b5a12a1
  由 Tom Hennigan 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235518682
```
  8b5a12a1
- A
  Catching memory corruption when initializing ConvParameters. · 2e1ea487
  由 A. Unique TensorFlower 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235511159
```
  2e1ea487
- A
  Return Status from CudaLaunchKernel. · 8a653bd9
  由 A. Unique TensorFlower 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235507181
```
  8a653bd9
- B
  [XLA] Fix and re-enable clamp test. · 0da7f0fd
  由 Benjamin Kramer 提交于 2月 25, 2019
```
This was fixed in 0bd78003

PiperOrigin-RevId: 235501348
```
  0da7f0fd
- B
  [XLA] Implement elemental LLVM codegen for replica-id · 64e7e114
  由 Benjamin Kramer 提交于 2月 25, 2019
```
This is a stub implementation that always returns zero as replication is not
supported on CPU or GPU currently.

PiperOrigin-RevId: 235497567
```
  64e7e114
- A
  Automated rollback of commit 9d8bc6b2 · ad675e01
  由 A. Unique TensorFlower 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235488332
```
  ad675e01
- A
  Multi-threading reduce_mean quantized using gemmlowp::tasks & optimized using neon. · 9d8bc6b2
  由 A. Unique TensorFlower 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235485786
```
  9d8bc6b2
- A
  compat: Update forward compatibility horizon to 2019-02-25 · e9a09aaf
  由 A. Unique TensorFlower 提交于 2月 25, 2019
```
PiperOrigin-RevId: 235484830
```
  e9a09aaf
- A
  Convert called bfloat16 computations in HandleMultipleOutputs(). · af82e5f0
  由 Adrian Kuegel 提交于 2月 25, 2019
```
This probably didn't matter before, but now Sort has a comparison computation.
Also add an optimization to not use a convert to BF16 if the user is a convert to F32.

PiperOrigin-RevId: 235481597
```
  af82e5f0
- A
  Automated rollback of commit ced4557e · fb7f8cc2
  由 A. Unique TensorFlower 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235476472
```
  fb7f8cc2
- S
  [XLA:CPU] Do not output fuse bias-add to complex GEMV · 9f9efff9
  由 Sanjoy Das 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235440605
```
  9f9efff9
- A
  Automated rollback of commit 9ca8321d · d10b88cd
  由 A. Unique TensorFlower 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235438039
```
  d10b88cd
- A
  Unbreak Windows GPU Build · 9ca8321d
  由 A. Unique TensorFlower 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235435925
```
  9ca8321d
- G
  Reenable building with TRT in docker. · 3c203367
  由 Guangda Lai 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235418348
```
  3c203367
24 2月, 2019 4 次提交
- D
  Automated rollback of commit e42a66b2 · 1721fa8b
  由 Dan Moldovan 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235411715
```
  1721fa8b
- T
  Merge pull request #25188 from yongtang:01242019-link · 1ad34b4f
  由 TensorFlower Gardener 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235409829
```
  1ad34b4f
- A
  compat: Update forward compatibility horizon to 2019-02-24 · 781fb015
  由 A. Unique TensorFlower 提交于 2月 24, 2019
```
PiperOrigin-RevId: 235395188
```
  781fb015
- J
  [SE] Fix function signature of stream_executor::CalculateDimensionality. · 6aa83398
  由 Justin Lebar 提交于 2月 23, 2019
```
PiperOrigin-RevId: 235369844
```
  6aa83398
23 2月, 2019 18 次提交
- L
  Updated documentation, reference to deleted TensorQueue. · 14523a55
  由 Lasse Espeholt 提交于 2月 23, 2019
```
PiperOrigin-RevId: 235340188
```
  14523a55
- T
  Add `batch_size` to allowed kwargs for backwards compatibility. · 7ef26685
  由 Thomas O'Malley 提交于 2月 23, 2019
```
PiperOrigin-RevId: 235323179
```
  7ef26685
- M
  Add TensorFlow license to c/jni tarballs. Fixes #21913. · 20dd116e
  由 Martin Wicke 提交于 2月 23, 2019
```
PiperOrigin-RevId: 235320635
```
  20dd116e
- A
  compat: Update forward compatibility horizon to 2019-02-23 · 13910e7b
  由 A. Unique TensorFlower 提交于 2月 23, 2019
```
PiperOrigin-RevId: 235320364
```
  13910e7b
- L
  Implement Reverse_Sequence. · 2ae06ca4
  由 Lu Wang 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235309643
```
  2ae06ca4
- T
  Merge pull request #25796 from duncanriach-nvidia:cudnn_deterministic_testing · caf3f171
  由 TensorFlower Gardener 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235309152
```
  caf3f171
- S
  Remove redundant namespace specifications from unqualified names in tf2tensorrt · d0494a89
  由 Smit Hinsu 提交于 2月 22, 2019
```
No functional changes.

PiperOrigin-RevId: 235303926
```
  d0494a89
- G
  Use TrtNodeValidator to determine whether an op type is supported, and register · 16d954d8
  由 Guangda Lai 提交于 2月 22, 2019
```
all node converters as validators as well.

This also fix a bug where it says Floor is supported (since it's in the op list
in IsTensorRTCandidate(), and validator_.ValidateNode() returns OK since no
validator was registered for it), but actually it's not supported until TRT
5.1.

PiperOrigin-RevId: 235295283
```
  16d954d8
- A
  Configuration for TPU embedding hot ID optimization. · 8af58fde
  由 A. Unique TensorFlower 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235294450
```
  8af58fde
- K
  Don't create stripped binary for debug builds · 3157d811
  由 Karim Nosir 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235294387
```
  3157d811
- T
  Merge pull request #25079 from lhendre:master · f45257b6
  由 TensorFlower Gardener 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235288226
```
  f45257b6
- A
  Upgrade gRPC version used in OSS Tensorflow · ced4557e
  由 A. Unique TensorFlower 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235285807
```
  ced4557e
- J
  [XLA] Add exhaustive tests for log1p and expm1. · 8c1887bb
  由 Justin Lebar 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235284387
```
  8c1887bb
- H
  Pass config proto through session options in eager graph execution. · dfb34a1a
  由 Haoyu Zhang 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235282812
```
  dfb34a1a
- J
  [XLA] Remove math_exhaustive_test. · fd53853d
  由 Justin Lebar 提交于 2月 22, 2019
```
Now mostly covered by exhaustive_op_test.  The big missing gap is the trig
functions.  But to fix these, we need to carefully define how they're supposed
to work over complex inputs, and that's a bigger task than I'm ready to take on
at the moment.

PiperOrigin-RevId: 235281417
```
  fd53853d
- T
  Merge pull request #24092 from schaber:patch-1 · 150f75e4
  由 TensorFlower Gardener 提交于 2月 22, 2019
```
PiperOrigin-RevId: 235281210
```
  150f75e4
- M
  Make submodules the preferred names, except for modules, they should prefer the root. · c4c1a239
  由 Mark Daoust 提交于 2月 22, 2019
```
Prefer `tf.submodule.Symbol` to `tf.Symbol`.
When the object is a module, then prefer
`tf.submodule` over `tf.package.submodule` so that `tf.submodule.Symbol` finds `tf.submodule`.

PiperOrigin-RevId: 235278774
```
  c4c1a239
- J
  [XLA] Improve accuracy of digamma function. · 9e8c24df
  由 Justin Lebar 提交于 2月 22, 2019
```
This function suffered from the same accuracy problems when computing
cot(pi * x) as lgamma.

PiperOrigin-RevId: 235278472
```
  9e8c24df

qq_38905368 / tensorflow 与 Fork 源项目一致

qq_38905368 / tensorflow
与 Fork 源项目一致