Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

Commits on Dec 4, 2017

• Enable bfloat16 use from Python:

  Show description for d87a76d
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  d87a76d

  Copy full SHA for d87a76d

Commits on Nov 22, 2017

• Merge changes from github.

  Show description for b1d8c59
  

  yifeif

  authored and

  tensorflower-gardener

  committed
  b1d8c59

  Copy full SHA for b1d8c59

• Automated g4 rollback of changelist 176615107

  Show description for e70c009
  

  yifeif

  authored and

  tensorflower-gardener

  committed
  e70c009

  Copy full SHA for e70c009

• Automated g4 rollback of changelist 176615737

  Show description for ad7eeec
  

  yifeif

  authored and

  tensorflower-gardener

  committed
  ad7eeec

  Copy full SHA for ad7eeec

• Merged commit includes the following changes:

  Show description for d0a3b2d
  tensorflower-gardener

  committed
  d0a3b2d

  Copy full SHA for d0a3b2d

• Merge changes from github.

  Show description for c6d603f
  

  yifeif

  authored and

  tensorflower-gardener

  committed
  c6d603f

  Copy full SHA for c6d603f

Commits on Nov 18, 2017

• make deep copy of axis parameter for batch norm

  Show description for c980a04
  

  chrisying

  authored and

  tensorflower-gardener

  committed
  c980a04

  Copy full SHA for c980a04

Commits on Nov 3, 2017

• Add regularizer support for fused batch norm.

  Show description for 3015655
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  3015655

  Copy full SHA for 3015655

Commits on Oct 30, 2017

• EagerVariableStore, for compatibility with functional layers.

  Show description for 8958267
  

  alextp

  authored and

  tensorflower-gardener

  committed
  8958267

  Copy full SHA for 8958267

Commits on Oct 26, 2017

• Replace the fake updates with no updates when not training. This is possible now that the tf.cond bug has been fixed, and is needed to remove rare data races.

  Show description for f13b76a
  tensorflower-gardener

  committed
  f13b76a

  Copy full SHA for f13b76a

• TFE: Add compatibility doc strings to functional layers

  Show description for 68b00b0
  

  caisq

  authored and

  tensorflower-gardener

  committed
  68b00b0

  Copy full SHA for 68b00b0

Commits on Oct 25, 2017

• Disables storing variables in the default variable store for eager.

  Show description for 56ceca4
  

  alextp

  authored and

  tensorflower-gardener

  committed
  56ceca4

  Copy full SHA for 56ceca4

Commits on Oct 23, 2017

• Adds the ability to adjust the normalized value in batchnorm. By using random adjustments, it is possible to improve the model generalization.

  Show description for 4f35cce
  tensorflower-gardener

  committed
  4f35cce

  Copy full SHA for 4f35cce

Commits on Oct 14, 2017

• Fix case where broadcasting is not necessary.

  Show description for 860f8c5
  

  chrisying

  authored and

  tensorflower-gardener

  committed
  860f8c5

  Copy full SHA for 860f8c5

Commits on Oct 13, 2017

• Remove unnecessary reshape and get_shape from fused_batch_norm.

  Show description for 45b1151
  

  chrisying

  authored and

  tensorflower-gardener

  committed
  45b1151

  Copy full SHA for 45b1151

Commits on Oct 12, 2017

• Multi-axis batch normalization support and changing from num_virtual_batches to virtual_batch_size

  Show description for 4351f88
  

  chrisying

  authored and

  tensorflower-gardener

  committed
  4351f88

  Copy full SHA for 4351f88

Commits on Oct 4, 2017

• Internal cleanup

  Show description for e7c5369
  tensorflower-gardener

  committed
  e7c5369

  Copy full SHA for e7c5369

• Added get variable utils to tf.estimator.Estimator.

  Show description for d66e77f
  

  ispirmustafa

  authored and

  tensorflower-gardener

  committed
  d66e77f

  Copy full SHA for d66e77f

• Internal cleanup

  Show description for 2fe6cf2
  tensorflower-gardener

  committed
  2fe6cf2

  Copy full SHA for 2fe6cf2

Commits on Oct 3, 2017

• Internal cleanup.

  Show description for 5123f29
  tensorflower-gardener

  committed
  5123f29

  Copy full SHA for 5123f29

Commits on Sep 28, 2017

• Implementing ghost batch norm as defined in https://arxiv.org/pdf/1705.08741.

  Show description for 125f7af
  

  chrisying

  authored and

  tensorflower-gardener

  committed
  125f7af

  Copy full SHA for 125f7af

Commits on Sep 27, 2017

• Add float16 support to tf.nn.fused_batch_norm on the GPU.

  Show description for 759690f
  

  reedwm

  authored and

  tensorflower-gardener

  committed
  759690f

  Copy full SHA for 759690f

Commits on Sep 16, 2017

• Merge changes from github.

  Show description for a373b1f
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  a373b1f

  Copy full SHA for a373b1f

Commits on Sep 11, 2017

• Enable fused batch norm, which is 15-20% faster for training and inference.

  Show description for 3491881
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  3491881

  Copy full SHA for 3491881

Commits on Sep 5, 2017

• Internal cleanup

  Show description for a8b7a32
  tensorflower-gardener

  committed
  a8b7a32

  Copy full SHA for a8b7a32

Commits on Aug 30, 2017

• Make convolutional, pooling and normalization layers to work with EAGER.

  Show description for db4afe5
  tensorflower-gardener

  committed
  db4afe5

  Copy full SHA for db4afe5

Commits on Aug 26, 2017

• Minor refactor of BatchNormalization's build, where we create gamma before beta.

  Show description for 17f26f8
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  17f26f8

  Copy full SHA for 17f26f8

Commits on Aug 24, 2017

• Increase tf.contrib.keras test coverage to 90%.

  Show description for fa83a27
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  fa83a27

  Copy full SHA for fa83a27

Commits on Aug 23, 2017

• Add an environment variable to test fused batch norm before we enable it by default.

  Show description for d2cf393
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  d2cf393

  Copy full SHA for d2cf393

Commits on Aug 9, 2017

• Add support for weight constraints in tf.layers API.

  Show description for b5c5985
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  b5c5985

  Copy full SHA for b5c5985

• Add a test for checkpoint compatibility between fused and non-fused batch norm.

  Show description for 594f320
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  594f320

  Copy full SHA for 594f320

Commits on Jun 21, 2017

• Make changes to prepare for the fused option of batch norm to be set to None (None means using fused batch norm if possible).

  Show description for b3f33ad
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  b3f33ad

  Copy full SHA for b3f33ad

Commits on Jun 3, 2017

• Add fused batch norm to tf.layers.

  Show description for 675d36b
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  675d36b

  Copy full SHA for 675d36b

Commits on May 19, 2017

• Fixed up the documentation for tf.contrib.layers.batch_norm and tf.layers.batch_normalization explaining how to include the update ops in the train_op. Also simplified the existing tf.layers.batch_…

  Show description for 675f0f9
  

  nealwu

  authored and

  tensorflower-gardener

  committed
  675f0f9

  Copy full SHA for 675f0f9

Commits on May 17, 2017

• Refactor Keras layers to rely on core TF layers, specifically:

  Show description for 35253fa
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  35253fa

  Copy full SHA for 35253fa