Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

Commits on Jul 29, 2016

• Make sure that kernels that are meant to run on GPU are scheduled inline. This

  Show description for f1acb3b
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  f1acb3b

  Copy full SHA for f1acb3b

Commits on Jul 18, 2016

• Reduce overhead of running kernels:

  Show description for b8ccc5a
  tensorflower-gardener

  committed
  b8ccc5a

  Copy full SHA for b8ccc5a

Commits on Jun 30, 2016

• Fix initialization issues with Variables whose shape contains a zero.

  Show description for 379df09
  tensorflower-gardener

  committed
  379df09

  Copy full SHA for 379df09

• Merge changes from github.

  Show description for 533d891
  tensorflower-gardener

  committed
  533d891

  Copy full SHA for 533d891

Commits on Jun 2, 2016

• Update copyright for 3p/tf/core.

  Show description for c8b59c0
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  c8b59c0

  Copy full SHA for c8b59c0

Commits on Jun 1, 2016

• Change function library to use a runner passed in from the Run call, instead of

  Show description for caed96e
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  caed96e

  Copy full SHA for caed96e

Commits on May 17, 2016

• Avoid calling LogMemory::IsEnabled in allocate_tensor. This saves a couple

  Show description for 20c7eca
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  20c7eca

  Copy full SHA for 20c7eca

Commits on May 4, 2016

• Only construct UniqueTensorReferences if they are needed.

  Show description for fc99eea
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  fc99eea

  Copy full SHA for fc99eea

Commits on Apr 21, 2016

• Check for op deprecation in C++ API and graph validation

  Show description for 18429eb
  

  girving

  authored and

  tensorflower-gardener

  committed
  18429eb

  Copy full SHA for 18429eb

Commits on Apr 20, 2016

• Add a .Deprecated method to REGISTER_OP

  Show description for 30334d2
  

  girving

  authored and

  tensorflower-gardener

  committed
  30334d2

  Copy full SHA for 30334d2

Commits on Apr 11, 2016

• Make the TF_CALL_*() macros drop the semicolon, instead putting that burden

  Show description for 6e8e7ca
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  6e8e7ca

  Copy full SHA for 6e8e7ca

Commits on Apr 10, 2016

• This is another step to make TensorFlow more interactive and flexible to users. It allows a tensor produced by a run call to stay "in-place" so that a future run call can use it in-place. To achiev…

  Show description for 098f930
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  098f930

  Copy full SHA for 098f930

Commits on Mar 28, 2016

• Change selective registration of op kernels to work off of the class name

  Show description for e30d80d
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  e30d80d

  Copy full SHA for e30d80d

Commits on Mar 25, 2016

• Mention OpKernel thread-safety requirements.

  Show description for c9b94f0
  

  rryan

  authored and

  tensorflower-gardener

  committed
  c9b94f0

  Copy full SHA for c9b94f0

Commits on Mar 24, 2016

• Several changes to reduce TensorFlow code size (especially important

  Show description for 821920c
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  821920c

  Copy full SHA for 821920c

Commits on Mar 22, 2016

• Apply selective registration to op registration, to remove the OpDefBuilder

  Show description for 03993ff
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  03993ff

  Copy full SHA for 03993ff

Commits on Mar 12, 2016

• Add optional cc file whitelisting for operator kernel registration.

  Show description for eb56025
  

  andrewharp

  authored and

  tensorflower-gardener

  committed
  eb56025

  Copy full SHA for eb56025

Commits on Mar 9, 2016

• Assume errors are rare in OP_REQUIRES*() macros.

  Show description for 01b60af
  

  josh11b

  authored and

  tensorflower-gardener

  committed
  01b60af

  Copy full SHA for 01b60af

Commits on Mar 2, 2016

• Changes to reduce code size (especially useful for mobile):

  Show description for b76812c
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  b76812c

  Copy full SHA for b76812c

Commits on Feb 26, 2016

• Add a step_id field to the OpKernelContext indicating which step

  Show description for e8260f3
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  e8260f3

  Copy full SHA for e8260f3

Commits on Feb 25, 2016

• Avoid some over-inlined routines. Reduces code size of TensorFlow binaries

  Show description for 9ccc4b6
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  9ccc4b6

  Copy full SHA for 9ccc4b6

Commits on Feb 22, 2016

• Added the ability to track allocation sizes in the tracking allocator itself if the underlying

  Show description for 8f8f7f1
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  8f8f7f1

  Copy full SHA for 8f8f7f1

Commits on Feb 17, 2016

• Merge changes from github.

  Show description for fe056f0
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  fe056f0

  Copy full SHA for fe056f0

Commits on Feb 16, 2016

• Several changes to reduce the overhead and the cache footprint of dispatching

  Show description for cf66101
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  cf66101

  Copy full SHA for cf66101

Commits on Feb 11, 2016

• Moves MemoryType inference code out of OpKernel so that it can reused.

  Show description for 7415dfa
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  7415dfa

  Copy full SHA for 7415dfa

• Fixed the forward redeclaration of Eigen::ThreadPoolDevice and Eigen::GpuDevice

  Show description for d77ef35
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  d77ef35

  Copy full SHA for d77ef35

Commits on Feb 5, 2016

• Allows int32 on host memory work when a function passes int32 in/out.

  Show description for 0c6c4b9
  

  A. Unique TensorFlower

  authored and

  keveman

  committed
  0c6c4b9

  Copy full SHA for 0c6c4b9

Commits on Feb 3, 2016

• Add RemoveNewDefaultAttrsFromGraphDef() for preparing a GraphDef for

  Show description for 7b4a463
  

  josh11b

  authored and

  Vijay Vasudevan

  committed
  7b4a463

  Copy full SHA for 7b4a463

Commits on Jan 26, 2016

• Global search & replace to move to the new location for

  Show description for 64aa098
  

  josh11b

  authored and

  Vijay Vasudevan

  committed
  64aa098

  Copy full SHA for 64aa098

Commits on Jan 25, 2016

• Running our linter on a lot of files.

  Show description for 9b70316
  

  Vijay Vasudevan

  committed
  9b70316

  Copy full SHA for 9b70316

• Eliminate per-op allocation of gpu device wrapper. The PerOpGpuDevice is allocated once in the OpKernelContext::Params struct, then re-used every time a new OpKernelContext uses the Params. Thus in…

  Show description for ff8522d
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  ff8522d

  Copy full SHA for ff8522d

Commits on Jan 22, 2016

• Move #include <vector> out of port.h to users of std::vector<>.

  Show description for b481783
  

  josh11b

  authored and

  Vijay Vasudevan

  committed
  b481783

  Copy full SHA for b481783

Commits on Jan 18, 2016

• Adds enough auditing to make it possible to track tensor buffers throughout an execution, and build a cost model of memory usage.

  Show description for 6f62e43
  

  A. Unique TensorFlower

  authored and

  keveman

  committed
  6f62e43

  Copy full SHA for 6f62e43

Commits on Jan 14, 2016

• Avoid allocation of std::function<> object per node executed, as

  Show description for 6dbfb95
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  6dbfb95

  Copy full SHA for 6dbfb95

Commits on Jan 13, 2016

• Added OpKernel::is_internal and used it to avoid a memcmp against the

  Show description for 94538a9
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  94538a9

  Copy full SHA for 94538a9