提交 · 5228739673b0da092d0c54d47694c0d716ada649 · weixin_51669992 / tensorflow

24 8月, 2020 15 次提交

R
PSv2: Remove a TODO that has been addressed. · 52287396
由 Rick Chao 提交于 8月 24, 2020
```
PiperOrigin-RevId: 328129947
Change-Id: If3fc9bd3ccb4a271c60e48e052969c07174326e4
```
52287396
T
[XLA:GPU] Allow fusion of producers with expensive HLOs that have just one consumer. · 3746d5e4
由 Thomas Joerg 提交于 8月 24, 2020
```
PiperOrigin-RevId: 328128307
Change-Id: Ia9a5e9f4c20d78deee32b738ae1002d80eb935c1
```
3746d5e4
M
Remove dependency on Dialect global registration from //tensorflow/compiler/mlir/lite/... · 051ed1cb
由 Mehdi Amini 提交于 8月 24, 2020
```
PiperOrigin-RevId: 328109152
Change-Id: Ia460e89f785e9a2aaf21538083733e7e13730299
```
051ed1cb
A
Update GraphDef version to 503. · 227b3487
由 A. Unique TensorFlower 提交于 8月 24, 2020
```
PiperOrigin-RevId: 328100708
Change-Id: Ib79b4c72b07ce8c03bdaaac246f4bc492a116914
```
227b3487
A
compat: Update forward compatibility horizon to 2020-08-24 · c31ebc9f
由 A. Unique TensorFlower 提交于 8月 24, 2020
```
PiperOrigin-RevId: 328100698
Change-Id: Ia54de53b1a7abf269de44d5aa1817c14b9da1053
```
c31ebc9f

Integrate LLVM at llvm/llvm-project@b999400a4fb6 · 8995157a

由 Mehdi Amini 提交于 8月 24, 2020

Updates LLVM usage to match
[b999400a4fb6](https://github.com/llvm/llvm-project/commit/b999400a4fb6)

PiperOrigin-RevId: 328096769
Change-Id: I42c48819078ea634ec7a7bf4d73579c7ddda9c47

8995157a

Enable passing TFRT test, `testUniformIntsDegenerate`. · 09f5609f

由 Kibeom Kim 提交于 8月 23, 2020

TFE_TensorHandleToNumpy seems to be working with TFRT.

PiperOrigin-RevId: 328077896
Change-Id: I0fc569e67440c00327009d87e50960734ed2bce1

09f5609f

Use BuiltinOpResolverWithoutDefaultDelegates instead of BuiltinOpResolver for... · d4dcba13

由 A. Unique TensorFlower 提交于 8月 23, 2020

Use BuiltinOpResolverWithoutDefaultDelegates instead of BuiltinOpResolver for unit tests of xnnpack delegate itself to prepare for enabling xnnpack delegate by default across all platforms in the next 2.4.0 release.

PiperOrigin-RevId: 328076934
Change-Id: I69e21a6fbbe1b0e7146669ccd6481b774dcd9d2e

d4dcba13

Integrate LLVM at llvm/llvm-project@f6decfa36d89 · 536d5658

由 Mehdi Amini 提交于 8月 23, 2020

Updates LLVM usage to match
[f6decfa36d89](https://github.com/llvm/llvm-project/commit/f6decfa36d89)

PiperOrigin-RevId: 328073633
Change-Id: I5cd74bcf36c453cf073766f910a0f8442b66cb93

536d5658

Use BuiltinOpResolverWithoutDefaultDelegates instead of BuiltinOpResolver for... · b57f2238

由 Chao Mei 提交于 8月 23, 2020

PiperOrigin-RevId: 328068258
Change-Id: I3459bc3e7f25d2925da65fba3e19ac2bad57fff1

b57f2238

A
PFor inputs should be ndarrays. · 10332bb8
由 Akshay Modi 提交于 8月 23, 2020
```
PiperOrigin-RevId: 328068227
Change-Id: Ia084d946f3a0e5d071d7e8fec4263d1da26d9671
```
10332bb8

Integrate LLVM at llvm/llvm-project@f164534ca8e0 · 01b030b7

由 Mehdi Amini 提交于 8月 23, 2020

Updates LLVM usage to match
[f164534ca8e0](https://github.com/llvm/llvm-project/commit/f164534ca8e0)

PiperOrigin-RevId: 328046788
Change-Id: I714164211a50e0d273ec49046c66f7e484989428

01b030b7

[XLA:SPMD] Avoid unnecessary collective permutes · 2e8dec07

由 Yuanzhong Xu 提交于 8月 23, 2020

1. Try to reuse the original target tiled sharding when finding compatible target from partial sharding.
2. If the HLO is a broadcast, check if data is already the same between source/target pairs.

PiperOrigin-RevId: 328043490
Change-Id: I69dec53c50cb6cedf586afafc5181cd1ee29cdc6

2e8dec07

PR #42508: [MLIR] Erase dead lmhlo.constant ops · aaed01bd

由 Uday Bondhugula 提交于 8月 23, 2020

Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/42508

An lmhlo.constant op on an memref that is locally allocated and with
no users other than dealloc's can be deleted. Add a canonicalization
pattern for this.
Copybara import of the project:

--
8758c409a15f567e7cb8e1077faa020f5705c85a by Uday Bondhugula <uday@polymagelabs.com>:

[MLIR] Erase dead lmhlo.constant ops

An lmhlo.constant op on an memref that is locally allocated and with
no other users (other than dealloc's) can be deleted. Add a
canonicalization patter for this.

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/tensorflow/pull/42508 from polymage-labs:lhlo_constant_erase 8758c409a15f567e7cb8e1077faa020f5705c85a
PiperOrigin-RevId: 328042416
Change-Id: I27f9b5b5297bbf6fe81aff589f009197b75f49eb

aaed01bd

Explicitly load standard dialect in HLO importer (NFC) · 88d4492d

由 Eugene Burmako 提交于 8月 23, 2020

MLIR is moving to require explicitly loading of Dialect before creating entities
in a Dialect.

PiperOrigin-RevId: 328037037
Change-Id: Ib46275b26e8f77aab0fbd0f70cd2a48844dc360c

88d4492d

23 8月, 2020 6 次提交
- A
  compat: Update forward compatibility horizon to 2020-08-23 · cd00fde2
  由 A. Unique TensorFlower 提交于 8月 23, 2020
```
PiperOrigin-RevId: 328009565
Change-Id: If7bb6f781e9a5381e35e98408272a5cdb42f64c8
```
  cd00fde2
- A
  Update GraphDef version to 502. · 1b6ff795
  由 A. Unique TensorFlower 提交于 8月 23, 2020
```
PiperOrigin-RevId: 328009562
Change-Id: I462f5a36c28b5e71ce562e67c3226feea5d2fb7d
```
  1b6ff795
- P
  Add MultiWorkerMirroredStratgy to custom training loop tests. · cb4cb2dd
  由 Priya Gupta 提交于 8月 22, 2020
```
PiperOrigin-RevId: 328002300
Change-Id: I5713bc15bb0d7a8647b1097fe81570ace30cb1c5
```
  cb4cb2dd
- Y
  Internal change for Keras benchmarks. · 0c242192
  由 Yanhui Liang 提交于 8月 22, 2020
```
PiperOrigin-RevId: 327999616
Change-Id: Ie46cc103cad75561bd863c8d477eaa5c9f319452
```
  0c242192
- B
  [XLA] Fix "lambda-expression in unevaluated context" error. · f3cd4e4a
  由 Berkin Ilbeyi 提交于 8月 22, 2020
```
PiperOrigin-RevId: 327999499
Change-Id: Iccd368a8784550a0a14f146924ee8845132e0b38
```
  f3cd4e4a
- M
  Explicitly load Mhlo dialect in HLO importer (NFC) · f6cb841c
  由 Mehdi Amini 提交于 8月 22, 2020
```
MLIR is moving to require explicitly loading of Dialect before creating entities
in a Dialect.

PiperOrigin-RevId: 327996308
Change-Id: Iba1de332fbd2c7d4d6a336b54ef999decc520ed3
```
  f6cb841c
22 8月, 2020 19 次提交

F
(lite) Change layout of hide_symbols_with_allowlist.sh to conform to the Google Shell Style guide · 885a34ac
由 Fergus Henderson 提交于 8月 22, 2020
```
<https://google.github.io/styleguide/shellguide.html#s5.4-loops>

PiperOrigin-RevId: 327935590
Change-Id: I9046a7fbf51fdbcd510633513d982e27e471ceff
```
885a34ac
A
Update GraphDef version to 501. · c352bafb
由 A. Unique TensorFlower 提交于 8月 22, 2020
```
PiperOrigin-RevId: 327934747
Change-Id: I80eca5e53d3c3d1f3bc9996418d3713812cebd92
```
c352bafb
A
compat: Update forward compatibility horizon to 2020-08-22 · 1f62e910
由 A. Unique TensorFlower 提交于 8月 22, 2020
```
PiperOrigin-RevId: 327934745
Change-Id: I5b03985ca1fe0858e9a47becb5dd6615a600da90
```
1f62e910

Enhance lowering reshape op to Linalg. · fb1ed49e

由 Hanhan Wang 提交于 8月 21, 2020

Handle non-expansion and non-collapsion cases by rewriting it to two reshape
ops.

PiperOrigin-RevId: 327926863
Change-Id: I2b9f406d505ab69d9e25e892f75f38aa03467e1e

fb1ed49e

T
Merge pull request #42413 from bioothod:export_golang_functions_master · fcdb58d1
由 TensorFlower Gardener 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327913260
Change-Id: I53c662d45c68eb56e0b5002afd5a48119b240bd6
```
fcdb58d1

Temporarily disable xla_compiler_test. · 1cc66e06

由 Reed Wanderman-Milne 提交于 8月 21, 2020

Currently it fails to build

PiperOrigin-RevId: 327910344
Change-Id: I1622cfe2a5196b37b6033913d37b49d0c1379a09

1cc66e06

A
[XLA:SPMD] Fix bug in partial replicate resharding. · 41f429e7
由 A. Unique TensorFlower 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327906446
Change-Id: Ifb39e840f9a8b66a52160f971007f6672f59f679
```
41f429e7
Y
Forward compat change to LSTM. · 83c96f6b
由 Yanhui Liang 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327904183
Change-Id: Ib9b46931f3049db5ec9b72fe0548b81c2562ea1b
```
83c96f6b
A
Add support for `batch_dims < 0` to Gather pfor converter. · 83615c0a
由 A. Unique TensorFlower 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327902700
Change-Id: I5c4d0d2f007bc74f3165f701595f0ba0e3c5a5ba
```
83615c0a
R
Add build rules for SSIM, PSNR. · ccc05be4
由 Russell Power 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327902010
Change-Id: Ic3eb19ae45fb3ff22d8d4fa9241c92104fd01b7f
```
ccc05be4
R
Attempt to fx revive_test timeout on macOS · f9e6e541
由 Reed Wanderman-Milne 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327900094
Change-Id: I213eb1a62d247445b7b9ea83cec6a718337711f6
```
f9e6e541
F
Prevent initialization of TPU platform when retrieving it in TransferOps · 36d01719
由 Frank Chen 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327899732
Change-Id: I3888f3bcac86268a18dc957df0e4b197b9867351
```
36d01719
R
Temporarily disable tests that fail on Windows with Python 3.8. · 6c39bd56
由 Reed Wanderman-Milne 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327899317
Change-Id: I329b02411bbaf66ce32d0880edadccb440f3d4f6
```
6c39bd56
B
[XLA] Free up cross-program-prefetched buffers after the last use. · 63593f36
由 Berkin Ilbeyi 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327899057
Change-Id: I602aa480c35b8734b50395d1c7e0fb621ad2d0fb
```
63593f36

PSv2: API change: Use strategy.scope where variables are created, and move... · 58f4a3e0

由 Rick Chao 提交于 8月 21, 2020

PSv2: API change: Use strategy.scope where variables are created, and move failure handling context manager to module level before we settle on a module for it to live in.

PiperOrigin-RevId: 327896478
Change-Id: I7c2b3583a6031282b5c549eb8b936f0dc91b10c0

58f4a3e0

H
Adding TpuCompilationRpcLookup class for distributed cache support. · 9b276a85
由 Henry Tan 提交于 8月 21, 2020
```
PiperOrigin-RevId: 327894710
Change-Id: Iac493500325f49b6a0cc5158b3b3962d022935dd
```
9b276a85

Integrate LLVM at llvm/llvm-project@02bf5632a94d · b42ffc24

由 A. Unique TensorFlower 提交于 8月 21, 2020

Updates LLVM usage to match
[02bf5632a94d](https://github.com/llvm/llvm-project/commit/02bf5632a94d)

PiperOrigin-RevId: 327894378
Change-Id: Id377dccdd14a889fa8c565c0a952972677a8709d

b42ffc24

[TF2XLA] Support dynamic slice size in strided slice op. · 1d654624

由 Yunxing Dai 提交于 8月 21, 2020

- Add two side outputs in ValidateStridedSliceOp to help analyze dynamic dimensions.
- Correctly set strided slice op's dynamic size if the slice size (slice end) is dynamic

PiperOrigin-RevId: 327893466
Change-Id: I7a5061ed1b92006c6f1f15c71661265b5bdfec25

1d654624

Add tf.core end-to-end KPI benchmarks. · f8066f9d

由 Kibeom Kim 提交于 8月 21, 2020

Add dedicated tf.core end-to-end KPI benchmarks for the following reasons:
- Most key tf.core API's execution time depends on other factors like input size, etc,...
- End-to-end time is important as there are overheads that's not caught by internal timing measuring e.g. b/158246276

PiperOrigin-RevId: 327893393
Change-Id: Ic01f98d98a8edc9e19f3fad64804abe916d4aee0

f8066f9d

weixin_51669992 / tensorflow 与 Fork 源项目一致

weixin_51669992 / tensorflow
与 Fork 源项目一致