Commits

Branch selector

 master

User selector

ebrevdo

Datepicker

All time

Commit History

Commits on Oct 13, 2017

• Temporarily make tf.where emit an old-style NodeDef (cast non-bool inputs to bool)

  Show description for d100f65
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  d100f65

  Copy full SHA for d100f65

Commits on Oct 12, 2017

• Wrap grad_ys tensors passed to tf.gradients in the tf.gradients name scope.

  Show description for d6b6169
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  d6b6169

  Copy full SHA for d6b6169

Commits on Oct 11, 2017

• More Variant cross-device support:

  Show description for 4a13939
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  4a13939

  Copy full SHA for 4a13939

Commits on Oct 10, 2017

• Move LinearOperator to tf.linalg (with backwards compatibility support in contrib.linalg.)

  Show description for 23418e4
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  23418e4

  Copy full SHA for 23418e4

• Automated g4 rollback of changelist 170772848

  Show description for 1bd776c
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  1bd776c

  Copy full SHA for 1bd776c

• Fix gradient behavior of fully dynamic tensor arrays + stop_gradients on tf.scan.

  Show description for 091504a
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  091504a

  Copy full SHA for 091504a

Commits on Oct 4, 2017

• Improve gradient shape validation errors.

  Show description for 32dc203
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  32dc203

  Copy full SHA for 32dc203

• Bugfix to LSTMBlockCell and friends: clipping is off by default.

  Show description for f9f037c
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  f9f037c

  Copy full SHA for f9f037c

Commits on Oct 3, 2017

• Internal Variant API allowing registering Variants to be copied from/to GPU.

  Show description for 0a11eaf
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  0a11eaf

  Copy full SHA for 0a11eaf

Commits on Oct 2, 2017

• Automated g4 rollback of changelist 170758184

  Show description for a2b23b0
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  a2b23b0

  Copy full SHA for a2b23b0

• TF WhereOp now acts more like np.where: extend input types to any numeric type.

  Show description for a470779
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  a470779

  Copy full SHA for a470779

Commits on Sep 30, 2017

• [tf.contrib.seq2seq] Better docstrings for AttentionWrapper and BeamSearchDecoder.

  Show description for aae34fa
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  aae34fa

  Copy full SHA for aae34fa

Commits on Sep 29, 2017

• Add complex kernel registrations for GatherNd and ScatterNd.

  Show description for 3ab081d
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  3ab081d

  Copy full SHA for 3ab081d

Commits on Sep 28, 2017

• Add CudaAtomicAdd for complex64, complex128 for SM30 and below.

  Show description for 1811923
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  1811923

  Copy full SHA for 1811923

Commits on Sep 27, 2017

• SymbolicGradients: create the underlying runtime with the correct step container.

  Show description for 2a5fb08
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  2a5fb08

  Copy full SHA for 2a5fb08

• Add new ReffedStatusCallback util class. This class allows multiple threads to

  Show description for ee37da0
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  ee37da0

  Copy full SHA for ee37da0

• Add new ReffedStatusCallback util class. This class allows multiple threads to

  Show description for 2ce49b2
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  2ce49b2

  Copy full SHA for 2ce49b2

Commits on Sep 26, 2017

• TF: GatherNd and ScatterNd updates.

  Show description for f5ceb90
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  f5ceb90

  Copy full SHA for f5ceb90

Commits on Sep 20, 2017

• Add tf.conj support for Variants, using a new variant op registration.

  Show description for 6e2f629
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  6e2f629

  Copy full SHA for 6e2f629

Commits on Sep 16, 2017

• Internal change.

  Show description for bd3ba66
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  bd3ba66

  Copy full SHA for bd3ba66

Commits on Sep 15, 2017

• Bugfix to tf.contrib.seq2seq beam_search_ops: GPU edge case of seq_len == 0.

  Show description for efbffdc
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  efbffdc

  Copy full SHA for efbffdc

Commits on Sep 9, 2017

• Reduce cmake log mess.

  Show description for 92bed17
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  92bed17

  Copy full SHA for 92bed17

Commits on Sep 8, 2017

• Modify variant registry to have UnaryOp and BinaryOp registrations. Speed up registry lookup.

  Show description for 477a221
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  477a221

  Copy full SHA for 477a221

• TF OpDefBuilder: support compound types when using tuple restrictions.

  Show description for 7565452
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  7565452

  Copy full SHA for 7565452

Commits on Sep 5, 2017

• Extended ScratchSpace to expose its underlying scratch tensor object.

  Show description for 6078160
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  6078160

  Copy full SHA for 6078160

Commits on Sep 1, 2017

• Added registry for variants for op ZerosLike.

  Show description for 0acf5bb
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  0acf5bb

  Copy full SHA for 0acf5bb

Commits on Aug 30, 2017

• Delete tf.contrib.training.python_input.

  Show description for 48e3b62
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  48e3b62

  Copy full SHA for 48e3b62

Commits on Aug 29, 2017

• Add int32 gpu registrations for SetZero & SetOne functors.

  Show description for 2bca5a1
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  2bca5a1

  Copy full SHA for 2bca5a1

Commits on Aug 24, 2017

• Implement Variant changes suggested by Manjunath Kudlur; simplify API.

  Show description for 105dd2a
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  105dd2a

  Copy full SHA for 105dd2a

Commits on Aug 22, 2017

• Add shape op registry for unary Variant tensors.

  Show description for 81ae1c6
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  81ae1c6

  Copy full SHA for 81ae1c6

• Bugfix to DropoutWrapper: never drop out the "c" (memory) of an LSTMStateTuple.

  Show description for cb33141
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  cb33141

  Copy full SHA for cb33141

• Add variant dtype op registry, first registration mechanism: shape ops.

  Show description for 2f6f329
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  2f6f329

  Copy full SHA for 2f6f329

Commits on Aug 21, 2017

• Fix link time bug caused by missing header include.

  Show description for 2ba7ec1
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  2ba7ec1

  Copy full SHA for 2ba7ec1

• Add ability for constant_op to accept encoded Variant tensors.

  Show description for 77b5f6a
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  77b5f6a

  Copy full SHA for 77b5f6a

Commits on Aug 16, 2017

• Variant objects can now be used inside the TensorFlow execution graph.

  Show description for 3292c1a
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  3292c1a

  Copy full SHA for 3292c1a

Pagination

Previous

Next