Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

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

Commits on May 11, 2017

• Fixes markdown formatting.

  Show description for a9c50ea
  tensorflower-gardener

  committed
  a9c50ea

  Copy full SHA for a9c50ea

• Merge changes from github.

  Show description for ee112cf
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  ee112cf

  Copy full SHA for ee112cf

Commits on May 4, 2017

• TensorFlow documentation fixes

  Show description for 4a4a1eb
  

  nealwu

  authored and

  tensorflower-gardener

  committed
  4a4a1eb

  Copy full SHA for 4a4a1eb

Commits on May 3, 2017

• [tf layers] Delay marking a layer as built until the end of its first apply().

  Show description for 485a24e
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  485a24e

  Copy full SHA for 485a24e

Commits on Apr 27, 2017

• Refactor Keras layers to rely on core TF layers.

  Show description for 1e3e5d4
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  1e3e5d4

  Copy full SHA for 1e3e5d4

Commits on Apr 18, 2017

• Merge changes from github.

  Show description for 69a4cf8
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  69a4cf8

  Copy full SHA for 69a4cf8

Commits on Apr 12, 2017

• Ensure that moving_mean and moving_variance are grabbed before being updated. Note that this should not be necessary since they are only updated during training and used during testing (unlike reno…

  Show description for 348b3b1
  tensorflower-gardener

  committed
  348b3b1

  Copy full SHA for 348b3b1

Commits on Apr 9, 2017

• Prevent the renorm_weight from being updated too early.

  Show description for a61cff6
  tensorflower-gardener

  committed
  a61cff6

  Copy full SHA for a61cff6

Commits on Apr 7, 2017

• Adds batch renormalization.

  Show description for 55c809b
  tensorflower-gardener

  committed
  55c809b

  Copy full SHA for 55c809b

Commits on Apr 5, 2017

• Merge changes from github.

  Show description for ccbc899
  tensorflower-gardener

  committed
  ccbc899

  Copy full SHA for ccbc899

Commits on Mar 2, 2017

• Fix typos and inconsistencies in layers docstrings.

  Show description for 37b4005
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  37b4005

  Copy full SHA for 37b4005

Commits on Jan 19, 2017

• Default reuse should be None, never False. Correct it in layers and add tests.

  Show description for 1ed4e69
  

  Lukasz Kaiser

  authored and

  tensorflower-gardener

  committed
  1ed4e69

  Copy full SHA for 1ed4e69

Commits on Jan 13, 2017

• Merge changes from github.

  Show description for 769496b
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  769496b

  Copy full SHA for 769496b

Commits on Jan 11, 2017

• Now that moments are numerically stable by default, not need to use moving_mean.

  Show description for 37af1b8
  

  sguada

  authored and

  tensorflower-gardener

  committed
  37af1b8

  Copy full SHA for 37af1b8

Commits on Dec 13, 2016

• Avoid both conditional branches being computed in Dropout and BatchNormalization.

  Show description for 5d43d55
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  5d43d55

  Copy full SHA for 5d43d55

Commits on Dec 12, 2016

• Changes to zeros_initializer which seem to have been missed in the migration.

  Show description for 2e47277
  tensorflower-gardener

  committed
  2e47277

  Copy full SHA for 2e47277

Commits on Dec 3, 2016

• Update base Layer to be consistent with the rest of TF.

  Show description for e0e2c69
  

  sguada

  authored and

  tensorflower-gardener

  committed
  e0e2c69

  Copy full SHA for e0e2c69

Commits on Dec 2, 2016

• Add BatchNormalization layer class and its functional interface.

  Show description for f56c0ab
  

  fchollet

  authored and

  tensorflower-gardener

  committed
  f56c0ab

  Copy full SHA for f56c0ab

Loading