Commits

Branch selector

 master

User selector

hawkinsp

Datepicker

All time

Commit History

Commits on Mar 3, 2017

• [TF:XLA] Clarify order of operations for ReduceWindow and SelectAndScatter.

  Show description for 1dc89c1
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  1dc89c1

  Copy full SHA for 1dc89c1

• [TF:XLA] Implement SparseSoftmaxCrossEntropyWithLogits.

  Show description for eaa668e
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  eaa668e

  Copy full SHA for eaa668e

Commits on Mar 1, 2017

• [TF:XLA] Add implementation of tf.one_hot to the XLA bridge.

  Show description for f424ca3
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  f424ca3

  Copy full SHA for f424ca3

Commits on Feb 27, 2017

• [TF:XLA] Improvements to resource variables:

  Show description for b436f41
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  b436f41

  Copy full SHA for b436f41

• [TF:XLA] Refactor build targets in //tensorflow/compiler/jit, split ops into more traditional ops/ and kernels/ directories.

  Show description for 746dc3b
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  746dc3b

  Copy full SHA for 746dc3b

• [TF:XLA] Silence a number of compiler warnings, in particular warnings repeatedly issued by code in headers.

  Show description for b4d091d
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  b4d091d

  Copy full SHA for b4d091d

Commits on Feb 24, 2017

• [XLA] Add an IsFinite operation that tests elementwise whether values are finite (i.e., not NaN or Inf).

  Show description for 8120e2a
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  8120e2a

  Copy full SHA for 8120e2a

Commits on Feb 22, 2017

• [TF:XLA] Add support for resource variables to the Tensorflow/XLA bridge.

  Show description for 542c3cb
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  542c3cb

  Copy full SHA for 542c3cb

• Update LLVM release to @295711 to fix build breakage.

  Show description for 1968de7
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  1968de7

  Copy full SHA for 1968de7

Commits on Feb 16, 2017

• [TF:XLA] Refactor XlaOpRegistry, moving metadata about how to compile operators on a device into a struct.

  Show description for 6640d3f
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  6640d3f

  Copy full SHA for 6640d3f

Commits on Feb 15, 2017

• [TF:XLA] Split XlaOpRegistry out of xla_compilation_device.{cc,h} into a separate xla_op_registry.{cc,h}.

  Show description for a8c325e
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  a8c325e

  Copy full SHA for a8c325e

Commits on Feb 14, 2017

• Fix code that ignores tensorflow::Status.

  Show description for bc225bf
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  bc225bf

  Copy full SHA for bc225bf

Commits on Feb 13, 2017

• [TF:XLA] Implement Placeholder on XLA_ devices using the common Placeholder kernel, rather than via a dummy op. Implement PlaceholderV2 as well.

  Show description for 02b8cbf
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  02b8cbf

  Copy full SHA for 02b8cbf

Commits on Feb 11, 2017

• [TF:XLA] Refactor XlaContext, moving some of its reponsibilities to XlaCompiler and XlaOpKernelContext.

  Show description for 4e24bec
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  4e24bec

  Copy full SHA for 4e24bec

• Fix double evaluation of macro argument that was causing duplicate CUDA batched GEMM calls.

  Show description for 2d364af
  

  hawkinsp

  authored and

  yifeif

  committed
  2d364af

  Copy full SHA for 2d364af

Commits on Feb 10, 2017

• Remove problematic SE_RETURN_STATUS_AS_BOOL macro

  Show description for 52fd3b9
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  52fd3b9

  Copy full SHA for 52fd3b9

Commits on Feb 9, 2017

• [TF:XLA] Remove locking from XlaContext, which is only ever used from a single thread.

  Show description for e65fff6
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  e65fff6

  Copy full SHA for e65fff6

• Fix double evaluation of macro argument that was causing duplicate CUDA batched GEMM calls.

  Show description for f439016
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  f439016

  Copy full SHA for f439016

Commits on Feb 8, 2017

• [TF:XLA] Disable the XLA CPU jit by default when the JIT is requested via the OptimizerOptions.

  Show description for b9deaac
  

  hawkinsp

  authored and

  gunan

  committed
  b9deaac

  Copy full SHA for b9deaac

Commits on Feb 7, 2017

• Propagate TF_BUILD_ENABLE_XLA environment variable into the CI build docker container from ci_parameterized_build.sh.

  Show description for 2934985
  

  hawkinsp

  authored and

  Todd Wang

  committed
  2934985

  Copy full SHA for 2934985

• Update LLVM to r293266.

  Show description for 3f4fa88
  

  hawkinsp

  authored and

  gunan

  committed
  3f4fa88

  Copy full SHA for 3f4fa88

• [TF:XLA] Remove duplicate main() from randomized_tests.

  Show description for f873800
  

  hawkinsp

  authored and

  gunan

  committed
  f873800

  Copy full SHA for f873800

• Propagate devices when instantiating function bodies.

  Show description for ef2c658
  

  hawkinsp

  authored and

  gunan

  committed
  ef2c658

  Copy full SHA for ef2c658

• [TF:XLA] Add a placeholder implementation of Log1p (via log(1+x), which is not numerically accurate for x near 0).

  Show description for eefb945
  

  hawkinsp

  authored and

  gunan

  committed
  eefb945

  Copy full SHA for eefb945

• Decrement initial reference to XlaDeviceContext object to fix memory leak.

  Show description for 0ed5741
  

  hawkinsp

  authored and

  gunan

  committed
  0ed5741

  Copy full SHA for 0ed5741

• [TF:XLA] Remove two concat gradient test cases that took an unreasonably long time to run.

  Show description for be4ec67
  

  hawkinsp

  authored and

  gunan

  committed
  be4ec67

  Copy full SHA for be4ec67

• [XLA] Update comment on RNG_UNIFORM in xla_data.proto

  Show description for 2525297
  

  hawkinsp

  authored and

  gunan

  committed
  2525297

  Copy full SHA for 2525297

• [XLA] Change XLA operation RngUniform semantics to return values in a semi-open interval rather than a closed interval to match Tensorflow tf.random_uniform semantics.

  Show description for 882b196
  

  hawkinsp

  authored and

  gunan

  committed
  882b196

  Copy full SHA for 882b196

• [TF:XLA] Add an option to disable special handling of compile-time constants, instead leaving them as run-time outputs of the compiled computation.

  Show description for 7984588
  

  hawkinsp

  authored and

  gunan

  committed
  7984588

  Copy full SHA for 7984588

• [XLA] Add test for ReduceWindow with an uncommon reduce function.

  Show description for 50253e6
  

  hawkinsp

  authored and

  gunan

  committed
  50253e6

  Copy full SHA for 50253e6

• [TF:XLA] Make //tensorflow/compiler/tests:concat_ops_test medium size to fix Jenkins test timeout.

  Show description for 0348cff
  

  hawkinsp

  authored and

  gunan

  committed
  0348cff

  Copy full SHA for 0348cff

• [TF:XLA] Add support for compiling computations with no return values.

  Show description for 45fd545
  

  hawkinsp

  authored and

  gunan

  committed
  45fd545

  Copy full SHA for 45fd545

• [XLA] Remove superfluous ConsumeValueOrDie() call in LocalService::CompileAheadOfTime().

  Show description for 1a05c79
  

  hawkinsp

  authored and

  gunan

  committed
  1a05c79

  Copy full SHA for 1a05c79

• [TF:XLA] Mark _XlaLaunch as a stateful op since XLA's random-number generators are stateful.

  Show description for bd371a3
  

  hawkinsp

  authored and

  gunan

  committed
  bd371a3

  Copy full SHA for bd371a3

• [TF:XLA] Disable the XLA CPU jit by default when the JIT is requested via the OptimizerOptions.

  Show description for 9600720
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  9600720

  Copy full SHA for 9600720

Pagination

Previous

Next