Commits

Breadcrumbs

History for

onroadmap-update

User selector

All users

All time

Commit History

Commits on Jan 24, 2017

• Remove concat_v2 from r1.0 branch

  

  annarev

  authored and

  gunan

  committed
  ba45e40

  Copy full SHA for ba45e40

Commits on Jan 20, 2017

• Change weight broadcasting for losses.sparse_softmax_cross_entropy to be the same as sparse_softmax_cross_entropy_with_logits. This means allowing greater than rank 2 inputs.

  Show description for 8a01c64
  

  tensorflower-gardener

  authored and

  gunan

  committed
  8a01c64

  Copy full SHA for 8a01c64

• Make dtype consistent with docs for TP/TN/FP/FN ops.

  Show description for 5572e35
  

  tensorflower-gardener

  authored and

  gunan

  committed
  5572e35

  Copy full SHA for 5572e35

• Restrict weights rank to be the same as the broadcast target, to avoid ambiguity on broadcast rules.

  Show description for b731e37
  

  tensorflower-gardener

  authored and

  gunan

  committed
  b731e37

  Copy full SHA for b731e37

Commits on Jan 17, 2017

• Use integer division to compute number of updates.

  Show description for 0b56eab
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  0b56eab

  Copy full SHA for 0b56eab

Commits on Jan 15, 2017

• Default naming of scopes should continue, not reset, after jumps to reused scopes.

  Show description for e720678
  

  Lukasz Kaiser

  authored and

  tensorflower-gardener

  committed
  e720678

  Copy full SHA for e720678

Commits on Jan 13, 2017

• Makes the ops used by ParallelConcat internal.

  Show description for b24a0b2
  tensorflower-gardener

  committed
  b24a0b2

  Copy full SHA for b24a0b2

• Make weight broadcasting more strict (in particular, catch the case where a 1-dim in values shape could be broadcast across a non-1 dimension in weights).

  Show description for 16cc887
  tensorflower-gardener

  committed
  16cc887

  Copy full SHA for 16cc887

• Merge changes from github.

  Show description for 769496b
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  769496b

  Copy full SHA for 769496b

• More new ops for speeding up input pipline: stage and unstage.

  Show description for 85acf52
  tensorflower-gardener

  committed
  85acf52

  Copy full SHA for 85acf52

• Remove tf.mul, tf.neg and tf.sub from the public API

  Show description for 78b11ce
  

  aselle

  authored and

  tensorflower-gardener

  committed
  78b11ce

  Copy full SHA for 78b11ce

Commits on Jan 12, 2017

• Losses should support scalar logits/labels.

  Show description for 5ccc994
  tensorflower-gardener

  committed
  5ccc994

  Copy full SHA for 5ccc994

• Transpose confusion matrix output to make it consistent with sklearn.

  Show description for 4d9a240
  tensorflower-gardener

  committed
  4d9a240

  Copy full SHA for 4d9a240

• Fix bug in shape inference for set operations. It errantly had been setting first dim of output indices and values to first dim of inputs.

  Show description for f7240d8
  tensorflower-gardener

  committed
  f7240d8

  Copy full SHA for f7240d8

• Fix documentation of rows vs columns of confusion matrix.

  Show description for 91685f8
  tensorflower-gardener

  committed
  91685f8

  Copy full SHA for 91685f8

Commits on Jan 11, 2017

• Add optimized reverse kernel for reversing rows of a 3D tensor. This is used in

  Show description for 51e6e84
  tensorflower-gardener

  committed
  51e6e84

  Copy full SHA for 51e6e84

• Deprecate tf.neg, tf.mul, tf.sub (and remove math_ops.{neg,mul,sub} usages

  Show description for be270ec
  

  aselle

  authored and

  tensorflower-gardener

  committed
  be270ec

  Copy full SHA for be270ec

• Switch tf.concat_v2 references in third_party/tensorflow to tf.concat.

  Show description for 0e226af
  tensorflower-gardener

  committed
  0e226af

  Copy full SHA for 0e226af

• Automated rollback of change 141622306

  Show description for df52532
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  df52532

  Copy full SHA for df52532

• Add tf.concat with the new argument order: tf.concat(values, axis, name). Also,

  Show description for 063801b
  tensorflower-gardener

  committed
  063801b

  Copy full SHA for 063801b

Commits on Jan 10, 2017

• Remove all remaining tf.pack,tf.unpack references and remove tf.pack/tf.unpack

  Show description for 7ff06bc
  tensorflower-gardener

  committed
  7ff06bc

  Copy full SHA for 7ff06bc

• Set up the generated gradient while loops properly so that `None` is propagated when using tf.stop_gradient.

  Show description for 28fd303
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  28fd303

  Copy full SHA for 28fd303

• Refactor intializer functions as initializer classes. This is purely a refactoring change and does not affect any API.

  Show description for 905fcf5
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  905fcf5

  Copy full SHA for 905fcf5

Commits on Jan 9, 2017

• Remove tf.concat(concat_dim, values, name). I will follow up with another cl that will re-add tf.concat, but with the new argument order (values, axis, name).

  Show description for f1b755a
  tensorflower-gardener

  committed
  f1b755a

  Copy full SHA for f1b755a

• Change reduce_join to treat reduction_indices in the same way as

  Show description for a910b52
  tensorflower-gardener

  committed
  a910b52

  Copy full SHA for a910b52

• Added TensorArray.identity which returns a TensorArray with the same content and properties but that inherits the control dependencies from the context.

  Show description for 5eebf7d
  tensorflower-gardener

  committed
  5eebf7d

  Copy full SHA for 5eebf7d

Commits on Jan 6, 2017

• Benchmark.run_op_benchmark now also stores (memory) allocator maximum usage

  Show description for 99d4517
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  99d4517

  Copy full SHA for 99d4517

• Add support for gradient calculation for concat_v2 with negative axis to fix github

  Show description for f3d5990
  tensorflower-gardener

  committed
  f3d5990

  Copy full SHA for f3d5990

• Allow empty tf.Features in the FeatureLists for sparse sequence examples.

  Show description for a245f73
  tensorflower-gardener

  committed
  a245f73

  Copy full SHA for a245f73

• Small fix for named arguments in SparseFeature parsing.

  Show description for 088a5df
  tensorflower-gardener

  committed
  088a5df

  Copy full SHA for 088a5df

• Removes certain ops from docs that should never have been added:

  Show description for 5942d9d
  tensorflower-gardener

  committed
  5942d9d

  Copy full SHA for 5942d9d

• Setting losses_test to medium since it times out on Jenkins at times.

  Show description for 788705c
  tensorflower-gardener

  committed
  788705c

  Copy full SHA for 788705c

• Removing the constant checking for tf.parallel_stack since it converts

  Show description for 4fe340b
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  4fe340b

  Copy full SHA for 4fe340b

Commits on Jan 5, 2017

• Change arg order for {softmax,sparse_softmax,sigmoid}_cross_entropy_with_logits to be (labels, predictions), and force use of named args to avoid accidents.

  Show description for 333dc32
  

  martinwicke

  authored and

  tensorflower-gardener

  committed
  333dc32

  Copy full SHA for 333dc32

Commits on Jan 4, 2017

• Make Empty Op Stateful.

  Show description for 2eb1604
  tensorflower-gardener

  committed
  2eb1604

  Copy full SHA for 2eb1604