Commits

Breadcrumbs

History for

onr0.8

User selector

All users

All time

Commit History

Commits on Apr 11, 2016

• Fix RNN performance bug. + Additions to rnn benchmarks & benchmarks.py.

  Show description for eb161ec
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  eb161ec

  Copy full SHA for eb161ec

Commits on Apr 10, 2016

• Surface control_flow_ops.while_loop to public.

  Show description for 918e964
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  918e964

  Copy full SHA for 918e964

Commits on Mar 15, 2016

• Remove use of conitional pass-through from dynamic_rnn in favor of just

  Show description for b10c65c
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  b10c65c

  Copy full SHA for b10c65c

Commits on Mar 11, 2016

• Implements "bool strictness" for `tf.Tensor`.

  Show description for f6f77ac
  

  mrry

  authored and

  tensorflower-gardener

  committed
  f6f77ac

  Copy full SHA for f6f77ac

• Fix dependencies bugs

  Show description for 56f1d64
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  56f1d64

  Copy full SHA for 56f1d64

Commits on Mar 9, 2016

• - addressing github issue #1198 (reverse_sequence only accepts int64)

  Show description for 484a80c
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  484a80c

  Copy full SHA for 484a80c

• Remove the requirement for sequence_length input to dynamic_rnn.

  Show description for e638881
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  e638881

  Copy full SHA for e638881

Commits on Mar 7, 2016

• Make RNN calls require a known input depth (via shape inference).

  Show description for cc62d99
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  cc62d99

  Copy full SHA for cc62d99

Commits on Mar 4, 2016

• Allow multiple calls to 'dynamic_rnn'.

  Show description for 08624cf
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  08624cf

  Copy full SHA for 08624cf

Commits on Mar 1, 2016

• Added a benchmark for the performance of dyanmic rnn with memory swapping. For the large model, the overhead is roughly 35%. More improvements over this base-line implementation are coming. But if…

  Show description for 384480d
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  384480d

  Copy full SHA for 384480d

Commits on Feb 29, 2016

• Bugfix to dynamic_rnn: restore some shape information when possible during loop.

  Show description for 1827f66
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  1827f66

  Copy full SHA for 1827f66

Commits on Feb 22, 2016

• Add cache_device to Variable construction and variable_scope.

  Show description for e7c34a5
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  e7c34a5

  Copy full SHA for e7c34a5

Commits on Feb 10, 2016

• Add fully dynamic RNN op. Unofficial version (not yet documented).

  Show description for 45197e0
  

  ebrevdo

  authored and

  tensorflower-gardener

  committed
  45197e0

  Copy full SHA for 45197e0

Commits on Feb 4, 2016

• Merge nodes do not currently forward shapes correctly. Set statically the

  Show description for cfbd9f9
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  cfbd9f9

  Copy full SHA for cfbd9f9

Commits on Feb 3, 2016

• tf.nn.rnn(..., sequence_length) now conditionally selects states and output.

  Show description for 50a8628
  

  ebrevdo

  authored and

  keveman

  committed
  50a8628

  Copy full SHA for 50a8628

Commits on Jan 28, 2016

• Breaking change in TF RNN python api: Return the final state instead of the

  Show description for fea55e1
  

  ebrevdo

  authored and

  Vijay Vasudevan

  committed
  fea55e1

  Copy full SHA for fea55e1

Commits on Jan 25, 2016

• Running our linter on a lot of files.

  Show description for 9b70316
  

  Vijay Vasudevan

  committed
  9b70316

  Copy full SHA for 9b70316

Commits on Dec 17, 2015

• - making sequence_length for bidirectional RNN optional as sufficient if the

  Show description for 635fa56
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  635fa56

  Copy full SHA for 635fa56

Commits on Dec 7, 2015

• TensorFlow: upstrea changes from git.

  Show description for cd53f3c
  

  Vijay Vasudevan

  committed
  cd53f3c

  Copy full SHA for cd53f3c

Commits on Dec 4, 2015

• TensorFlow: upstream changes to git

  Show description for 54a644f
  

  Vijay Vasudevan

  committed
  54a644f

  Copy full SHA for 54a644f

Commits on Dec 1, 2015

• TensorFlow: upstream changes to git

  Show description for 795f35d
  

  Vijay Vasudevan

  committed
  795f35d

  Copy full SHA for 795f35d

Loading