提交 · v2.11.0-rc0 · wux_labs / Tensorflow

18 10月, 2022 9 次提交
- R
  Added info about Windows CPU build destination · ba36eac6
  由 Rishika Sinha 提交于 10月 18, 2022
```
Added info about Windows CPU build destination
```
  ba36eac6
- M
  Merge pull request #58130 from tensorflow/fix-relnotes · 31223f49
  由 Mihai Maruseac 提交于 10月 17, 2022
```
Fix minor typos in release notes
```
  31223f49
- 8
  
  Refactor/update TensorFlow 2.11 RELEASE.md (#58132) · c60fb2af
  由 8bitmp3 提交于 10月 18, 2022
  
  c60fb2af
- M
  
  Fix minor typos in release notes · 7371c20e
  由 Mihai Maruseac 提交于 10月 17, 2022
  
  7371c20e
- L
  Merge pull request #58129 from tensorflow-jenkins/version-numbers-2.11.0rc0-5477 · 80dceb1a
  由 learning-to-play 提交于 10月 17, 2022
```
Update version numbers for TensorFlow 2.11.0-rc0
```
  80dceb1a
- T
  
  Update version numbers to 2.11.0-rc0 · 9f4f1b29
  由 TensorFlow Release Automation 提交于 10月 17, 2022
  
  9f4f1b29
- L
  Merge pull request #58128 from tensorflow-jenkins/relnotes-2.11.0rc0-9386 · 29f40f63
  由 learning-to-play 提交于 10月 17, 2022
```
Update release notes for TensorFlow 2.11.0
```
  29f40f63
- V
  Update release notes for TensorFlow 2.11.0 · 7b3cda3b
  由 Vinila S 提交于 10月 17, 2022
```
Update release notes for TensorFlow 2.11.0
```
  7b3cda3b
- T
  
  Insert release notes place-fill · ac084ab8
  由 TensorFlow Release Automation 提交于 10月 17, 2022
  
  ac084ab8
17 10月, 2022 3 次提交
- A
  [XLA] Eliminated constant folding for operations that have large number of... · 736d0db2
  由 A. Unique TensorFlower 提交于 10月 16, 2022
```
[XLA] Eliminated constant folding for operations that have large number of elements in their operands.

PiperOrigin-RevId: 481504087
```
  736d0db2
- L
  Pretty print TFRT and RunHandlerWorkQueue options. Otherwise, the options are... · 80f0668a
  由 Li Lao 提交于 10月 16, 2022
```
Pretty print TFRT and RunHandlerWorkQueue options. Otherwise, the options are shown as binary data byte-by-byte.

PiperOrigin-RevId: 481494878
```
  80f0668a
- T
  Merge pull request #58099 from elfringham:linalg_fails · 4eaf9e33
  由 TensorFlower Gardener 提交于 10月 16, 2022
```
PiperOrigin-RevId: 481492512
```
  4eaf9e33
16 10月, 2022 7 次提交
- A
  compat: Update forward compatibility horizon to 2022-10-16 · f7e3e12b
  由 A. Unique TensorFlower 提交于 10月 16, 2022
```
PiperOrigin-RevId: 481443251
```
  f7e3e12b
- A
  Update GraphDef version to 1286. · 82038f03
  由 A. Unique TensorFlower 提交于 10月 16, 2022
```
PiperOrigin-RevId: 481443061
```
  82038f03
- T
  [XNNPACK] Support Slice node in XNNPACK delegate · 4064b3d1
  由 Terry Heo 提交于 10月 15, 2022
```
PiperOrigin-RevId: 481382408
```
  4064b3d1
- T
  Merge pull request #55677 from PatriceVignola:add-get-input-tensor-from-variable-validation · 1b1ebd2b
  由 TensorFlower Gardener 提交于 10月 15, 2022
```
PiperOrigin-RevId: 481375400
```
  1b1ebd2b
- A
  Go: Update generated wrapper functions for TensorFlow ops. · ba3bd04e
  由 A. Unique TensorFlower 提交于 10月 15, 2022
```
PiperOrigin-RevId: 481372151
```
  ba3bd04e
- J
  Fix TensorArraySplitV3 markdown issues · af783344
  由 Jacques Pienaar 提交于 10月 15, 2022
```
Seems to be last case in #57926

PiperOrigin-RevId: 481366370
```
  af783344
- A
  [xla:gml_st] Relax a check that refuses to tile parallel dimensions when op... · b4587874
  由 A. Unique TensorFlower 提交于 10月 15, 2022
```
[xla:gml_st] Relax a check that refuses to tile parallel dimensions when op contains a reduction dimension

The user is responsible for making sure to not ask to generate parallel loops on reduction dimension.

PiperOrigin-RevId: 481366310
```
  b4587874
15 10月, 2022 21 次提交
- A
  
  Disable failing linalg tests on ARM_CI · 3381cb7c
  由 Andrew Goodbody 提交于 10月 15, 2022
  
  3381cb7c
- A
  Update GraphDef version to 1285. · be8c8442
  由 A. Unique TensorFlower 提交于 10月 15, 2022
```
PiperOrigin-RevId: 481322408
```
  be8c8442
- A
  compat: Update forward compatibility horizon to 2022-10-15 · 3da84eb2
  由 A. Unique TensorFlower 提交于 10月 15, 2022
```
PiperOrigin-RevId: 481322404
```
  3da84eb2
- A
  Update TFRT dependency to use revision · 7ed7254f
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
http://github.com/tensorflow/runtime/commit/4ce3e4da2e21ae4dfcee9366415e55f408c884ec.

PiperOrigin-RevId: 481300942
```
  7ed7254f
- E
  [xla:runtime] Use structs to define gpu runtime custom calls · 62bd5df8
  由 Eugene Zhulenev 提交于 10月 14, 2022
```
LLVM seems to have troubles with inlining custom call handlers defined by function pointers. When we use struct, then the custom call body is typically fully inlined into the CustomCallHandler template instantiation and generates better code.

PiperOrigin-RevId: 481300474
```
  62bd5df8
- E
  [xla:runtime] NFC: Move HostContextAsyncTaskRunner from XLA to TFRT · 2b124645
  由 Eugene Zhulenev 提交于 10月 14, 2022
```
Remove last HostContext dependency from XLA

PiperOrigin-RevId: 481300092
```
  2b124645
- A
  [XLA:GPU] Check that the types used in the matmul are supported by cublasLt.... · fceaaa08
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
[XLA:GPU] Check that the types used in the matmul are supported by cublasLt. If they are not supported, we fall back to legacy cublas.

This fixes an issue found by some JAX dot_general tests which depended on device (p100/v100). There are certain combinations of types which work on some devices but not on others. However, according to the official...

PiperOrigin-RevId: 481291878
```
  fceaaa08
- T
  lite: Remove '--config=monolithic' from selective build script · 645d658a
  由 Terry Heo 提交于 10月 14, 2022
```
The flag doesn't compatible with //tensorflow/python/tools:print_selective_registration_header
which is needed to generate header file via genrule().

PiperOrigin-RevId: 481287663
```
  645d658a
- A
  [mhlo/hlo] Roundtrip execution_thread for non-async ops · b9646c44
  由 Anish Tondwalkar 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481281063
```
  b9646c44
- A
  Update TFRT dependency to use revision · 62a1a9b2
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
http://github.com/tensorflow/runtime/commit/e03a192239c4f59c63db1b4d54b4c15f1d0a1ad7.

PiperOrigin-RevId: 481277525
```
  62a1a9b2
- A
  TypeDispatchTable uses FunctionType · 8d8abfb5
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481275791
```
  8d8abfb5
- A
  Integrate LLVM at llvm/llvm-project@d8415b02a519 · 2cffb704
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
Updates LLVM usage to match
[d8415b02a519](https://github.com/llvm/llvm-project/commit/d8415b02a519)

PiperOrigin-RevId: 481275732
```
  2cffb704
- F
  TypeDispatchTable uses FunctionType · ddb5de62
  由 Faizan Muhammad 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481266358
```
  ddb5de62
- T
  Merge pull request #57163 from PatriceVignola:link-tensorflow-dll-kernels-experimental-windows · e78176bf
  由 TensorFlower Gardener 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481265147
```
  e78176bf
- J
  Refactor MlirBridgePass::Run to observe pass state before running MLIR ridge. · 956c1b52
  由 Jian Cai 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481258105
```
  956c1b52
- A
  Handle a special case of lambda usage properly for tf.function freevar detection · f1e2961e
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481253023
```
  f1e2961e
- A
  [tf:jitrt] Make use of default values of ListOption in TfJitRtPipelineOptions · 4d8e3dd2
  由 A. Unique TensorFlower 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481252019
```
  4d8e3dd2
- V
  [XLA:GPU] Check that the types used in the matmul are supported by cublasLt.... · d58a8dfc
  由 Victor Stone 提交于 10月 14, 2022
```
[XLA:GPU] Check that the types used in the matmul are supported by cublasLt. If they are not supported, we fall back to legacy cublas.

This fixes an issue found by some JAX dot_general tests which depended on device (p100/v100). There are certain combinations of types which work on some devices but not on others. However, according to the official cublasLt documentation, these combinations of types are unsupported by cublasLt on all devices.

PiperOrigin-RevId: 481251576
```
  d58a8dfc
- C
  Splitting out TF specific functionality from xScopedDiagnosticHandler · 0549fd10
  由 Catherine Payne 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481251198
```
  0549fd10
- P
  Guard an aarch64-specific TF-oneDNN code block with an ifdef to make sure it... · 034f6df3
  由 Penporn Koanantakool 提交于 10月 14, 2022
```
Guard an aarch64-specific TF-oneDNN code block with an ifdef to make sure it doesn't affect the x86 backend, per discussion here:
https://github.com/tensorflow/tensorflow/pull/57987#discussion_r993731524

PiperOrigin-RevId: 481247334
```
  034f6df3
- Y
  Guard the new channel_handle feature on mlir_api_version for backwards compatibility · 8f8f589f
  由 Yash Katariya 提交于 10月 14, 2022
```
PiperOrigin-RevId: 481246613
```
  8f8f589f