Commits

Branch selector

 master

User selector

pavithrasv

Datepicker

All time

Commit History

Commits on Jun 11, 2019

• Marking `apply` and `add_variable` on Layer as deprecated and removing docs for them as they are aliases for other functions on Layer.

  Show description for 20db62a
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  20db62a

  Copy full SHA for 20db62a

Commits on Jun 7, 2019

• Move `Node` class into a separate file.

  Show description for f4bb606
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  f4bb606

  Copy full SHA for f4bb606

Commits on Jun 6, 2019

• Using new virtual CPUs strategy combination for testing loss scaling utilities instead of relying on GPUs.

  Show description for cdb7d30
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  cdb7d30

  Copy full SHA for cdb7d30

Commits on Jun 5, 2019

• Fix: Keras Sequential training fails after adding a new layer and re-compiling

  Show description for 5c727ed
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  5c727ed

  Copy full SHA for 5c727ed

• Deserializing loss class in hdf5 format.

  Show description for 868f22a
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  868f22a

  Copy full SHA for 868f22a

Commits on Jun 1, 2019

• Create utilities for custom training loop users to be able to scale losses correctly.

  Show description for 9071255
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  9071255

  Copy full SHA for 9071255

Commits on May 29, 2019

• Small optimization in metrics that inherit from Mean. Mean takes care of updating weight dimension so it is not required in the individual metrics.

  Show description for bba4554
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  bba4554

  Copy full SHA for bba4554

• Support channels first data format in cross entropy losses with logits and support tensors with unknown ranks

  Show description for c1d56bd
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  c1d56bd

  Copy full SHA for c1d56bd

Commits on May 21, 2019

• Add error message for usage of activity regularizer in a control flow branch.

  Show description for ae6f0bf
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  ae6f0bf

  Copy full SHA for ae6f0bf

Commits on May 17, 2019

• Update error message when we check for graph consistency to not say `tf.cond` as a fix. Removing graph consistency check for callable losses.

  Show description for 70c8c28
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  70c8c28

  Copy full SHA for 70c8c28

Commits on May 16, 2019

• Scale output loss after computing the aggregated output loss.

  Show description for a113ceb
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  a113ceb

  Copy full SHA for a113ceb

Commits on May 13, 2019

• Add check to see if model contains `_sample_weight_modes` property when creating execution function as fit may be called without compile. We should raise the correct error message in this case.

  Show description for dc87891
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  dc87891

  Copy full SHA for dc87891

Commits on May 9, 2019

• Wrap loss and metric calls in a graph context if any of the input is a symbolic tensor.

  Show description for 9ea8059
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  9ea8059

  Copy full SHA for 9ea8059

Commits on May 7, 2019

• Add information about None reduction type is handled in Keras fit/eval.

  Show description for 14a3a7d
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  14a3a7d

  Copy full SHA for 14a3a7d

• Internal cleanup

  Show description for 6d9643a
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  6d9643a

  Copy full SHA for 6d9643a

Commits on May 2, 2019

• Add test case for checking multiple metric instances with default parameters.

  Show description for 34f34bf
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  34f34bf

  Copy full SHA for 34f34bf

Commits on May 1, 2019

• Switch to using V2 name scope in Keras in V2.

  Show description for 3571007
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  3571007

  Copy full SHA for 3571007

Commits on Apr 24, 2019

• Remove the `sample_weights` placeholder for models that don't have sample weights.

  Show description for 41abdd1
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  41abdd1

  Copy full SHA for 41abdd1

• Add validation for loss reduction during creation. We currently validate on call.

  Show description for a0fd161
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  a0fd161

  Copy full SHA for a0fd161

Commits on Apr 23, 2019

• Remove backend.mean() from result tensors in eager mode.

  Show description for 408cea8
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  408cea8

  Copy full SHA for 408cea8

Commits on Apr 22, 2019

• Add prefix to tf op layer name so that it is different from Keras layer names.

  Show description for 0a9f689
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  0a9f689

  Copy full SHA for 0a9f689

Commits on Apr 18, 2019

• Set default loss reduction as `AUTO` for improving reliability of loss scaling with distribution strategy and custom training loops. Raise an error in distribution strategy + custom training loop u…

  Show description for e19be1a
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  e19be1a

  Copy full SHA for e19be1a

Commits on Apr 17, 2019

• Improve loss scaling with distribution strategy.

  Show description for 677a148
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  677a148

  Copy full SHA for 677a148

Commits on Apr 4, 2019

• Automated rollback of commit cdd0dcc6f78d2443d7713094c80a4a169531b6c1

  Show description for cc14235
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  cc14235

  Copy full SHA for cc14235

Commits on Apr 3, 2019

• Move loss and weighted metrics graph compilation into its own function.

  Show description for aa8bbb0
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  aa8bbb0

  Copy full SHA for aa8bbb0

Commits on Apr 1, 2019

• Throw error if there are unexpected keys in sample_weight / class_weight dictionaries.

  Show description for 3cade69
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  3cade69

  Copy full SHA for 3cade69

Commits on Mar 26, 2019

• Wrapping Keras subclassed layer call in autograph.

  Show description for cdd0dcc
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  cdd0dcc

  Copy full SHA for cdd0dcc

• Fix - Get layer nested layer metrics.

  Show description for 08d8f30
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  08d8f30

  Copy full SHA for 08d8f30

Commits on Mar 22, 2019

• Add error msg if metrics dictionary in compile contains a name not in model output names.

  Show description for dac0725
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  dac0725

  Copy full SHA for dac0725

Commits on Mar 14, 2019

• Rename v1 initializers to compat.v1 in upgrade script. Fixes issue #26665

  Show description for a236ae2
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  a236ae2

  Copy full SHA for a236ae2

Commits on Mar 13, 2019

• Internal cleanup

  Show description for bacd851
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  bacd851

  Copy full SHA for bacd851

Commits on Mar 8, 2019

• 1. Remove make_fit_function and make_eval_function from Keras training.

  Show description for 83873bd
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  83873bd

  Copy full SHA for 83873bd

• Removing control dependency between update ops and outputs in graph execution function and replace with control dependency only on loss output.

  Show description for dd3c661
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  dd3c661

  Copy full SHA for dd3c661

Commits on Mar 7, 2019

• Replace V1 loss reduction with V2 loss reduction enum in V2 Canned Estimator head APIs.

  Show description for d2e479d
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  d2e479d

  Copy full SHA for d2e479d

Commits on Mar 5, 2019

• Fix performance issue in 2.0 for Keras model that takes dict input in function mode.

  Show description for 254e39f
  

  pavithrasv

  authored and

  tensorflower-gardener

  committed
  254e39f

  Copy full SHA for 254e39f

Pagination

Previous

Next