Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit history

Commits on Dec 18, 2017

• [StreamExecutor] Change "variance" to "inv_var" in BatchNormalizationBackward.

  Show description for 483e439
  

  Justin Lebar

  authored and

  tensorflower-gardener

  committed
  483e439

  Copy full SHA for 483e439

Commits on Dec 15, 2017

• Remove Stream::BlockHostUntilDoneWithStatus; all callers use BlockHostUntilDone.

  Show description for e355290
  tensorflower-gardener

  committed
  e355290

  Copy full SHA for e355290

Commits on Dec 13, 2017

• Stream::BlockHostUntilDone now returns Status rather than bool.

  Show description for 246d5b9
  tensorflower-gardener

  committed
  246d5b9

  Copy full SHA for 246d5b9

Commits on Dec 7, 2017

• Add BlockHostUntilDoneWithStatus, which returns Status rather than bool.

  Show description for 64c4e8f
  tensorflower-gardener

  committed
  64c4e8f

  Copy full SHA for 64c4e8f

Commits on Nov 3, 2017

• Support Cudnn RNN Fp16

  Show description for 509d0f2
  

  protoget

  authored and

  tensorflower-gardener

  committed
  509d0f2

  Copy full SHA for 509d0f2

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 6, 2017

• Add int8 version of fused_conv2d_bias_activation operator for the forward phase,

  Show description for 2b15bad
  tensorflower-gardener

  committed
  2b15bad

  Copy full SHA for 2b15bad

Commits on Aug 24, 2017

• Automated g4 rollback of changelist 166276461

  Show description for 91b00a1
  tensorflower-gardener

  committed
  91b00a1

  Copy full SHA for 91b00a1

• Add int8 version of fused_conv2d_bias_activation operator for the forward phase,

  Show description for 1f41602
  tensorflower-gardener

  committed
  1f41602

  Copy full SHA for 1f41602

Commits on Jul 21, 2017

• Let GetBlasGemmAlgorithms() always return true.

  Show description for 3e3306e
  

  Yangzihao Wang

  authored and

  tensorflower-gardener

  committed
  3e3306e

  Copy full SHA for 3e3306e

Commits on Jul 19, 2017

• Automated g4 rollback of changelist 162423171

  Show description for 491beb7
  tensorflower-gardener

  committed
  491beb7

  Copy full SHA for 491beb7

Commits on Jul 18, 2017

• Add autotuning code for matmul operator.

  Show description for 06accca
  

  Yangzihao Wang

  authored and

  tensorflower-gardener

  committed
  06accca

  Copy full SHA for 06accca

• Support float64 CuDNN RNN

  Show description for ed03ee3
  

  protoget

  authored and

  tensorflower-gardener

  committed
  ed03ee3

  Copy full SHA for ed03ee3

Commits on Jul 6, 2017

• Add support for int8 x int8 -> int32 matrix multiplication via cublasGemmEx to stream_executor.

  Show description for a2ee8bc
  tensorflower-gardener

  committed
  a2ee8bc

  Copy full SHA for a2ee8bc

Commits on Jun 27, 2017

• [SE] ThenConvolveWithAlgorithm vlogs algorithm configs.

  Show description for 7f0375e
  

  Jingyue Wu

  authored and

  tensorflower-gardener

  committed
  7f0375e

  Copy full SHA for 7f0375e

Commits on Jun 20, 2017

• [SE] Support alpha scale in cudnnTransformTensor

  Show description for 44cf398
  tensorflower-gardener

  committed
  44cf398

  Copy full SHA for 44cf398

Commits on Jun 13, 2017

• TransformTensor supports NCHW_VECT_C layout and int8 data type.

  Show description for de70229
  

  Jingyue Wu

  authored and

  tensorflower-gardener

  committed
  de70229

  Copy full SHA for de70229

Commits on Jun 6, 2017

• Pass int parameter by value, not by const reference

  Show description for e9a889c
  tensorflower-gardener

  committed
  e9a889c

  Copy full SHA for e9a889c

Commits on Jun 5, 2017

• [SE] Add cudnnTransformTensor to StreamExecutor.

  Show description for 9e68997
  

  Jingyue Wu

  authored and

  tensorflower-gardener

  committed
  9e68997

  Copy full SHA for 9e68997

Commits on Jun 2, 2017

• Add functional support for cudnnConvolutionBiasActivationForward().

  Show description for 69075f3
  

  Yangzihao Wang

  authored and

  tensorflower-gardener

  committed
  69075f3

  Copy full SHA for 69075f3

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 10, 2017

• Merge changes from github.

  Show description for 0386a01
  

  Dandelion Mané

  authored and

  tensorflower-gardener

  committed
  0386a01

  Copy full SHA for 0386a01

Commits on Mar 3, 2017

• [XLA] [StreamExecutor] Tune GEMMs when possible.

  Show description for 0119469
  

  Justin Lebar

  authored and

  tensorflower-gardener

  committed
  0119469

  Copy full SHA for 0119469

• [XLA:GPU] Cache GPU substreams across executions

  Show description for 112a534
  tensorflower-gardener

  committed
  112a534

  Copy full SHA for 112a534

Commits on Jan 24, 2017

• Add options argument for DNN activation

  Show description for eb65eff
  tensorflower-gardener

  committed
  eb65eff

  Copy full SHA for eb65eff

Commits on Jan 19, 2017

• Add convolve quantized ops to StreamExecutor API

  Show description for 598583c
  tensorflower-gardener

  committed
  598583c

  Copy full SHA for 598583c

Commits on Jan 17, 2017

• Add several operations to the StreamExecutor API

  Show description for b398f90
  tensorflower-gardener

  committed
  b398f90

  Copy full SHA for b398f90

Commits on Sep 16, 2016

• Add the interface in steam executor to call cuDNN batch normalization functions.

  Show description for 04df8d8
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  04df8d8

  Copy full SHA for 04df8d8

Commits on Aug 19, 2016

• Add stream-executor changes to enable Cudnn fused LSTM/RNN support.

  Show description for 859e47f
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  859e47f

  Copy full SHA for 859e47f

Commits on Jul 13, 2016

• Roll-forward of "Local Response Normalization GPU support via Stream Executor."

  Show description for 35df3ed
  

  rryan

  authored and

  tensorflower-gardener

  committed
  35df3ed

  Copy full SHA for 35df3ed

Commits on Jul 11, 2016

• Automated rollback of change 127123966

  Show description for 3f1b887
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  3f1b887

  Copy full SHA for 3f1b887

• Local Response Normalization GPU support via Stream Executor.

  Show description for 71c9fb3
  

  rryan

  authored and

  tensorflower-gardener

  committed
  71c9fb3

  Copy full SHA for 71c9fb3

Commits on Jun 21, 2016

• Improve convolution autotune process. The max batch size VGG model can handle

  Show description for 466eb29
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  466eb29

  Copy full SHA for 466eb29

Commits on Jun 7, 2016

• Enable fp16 for most of the pooling ops (MaxPool, AvgPool, associated

  Show description for 5a84537
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  5a84537

  Copy full SHA for 5a84537

Commits on Jun 3, 2016

• Enable fp16 for most of the pooling ops (MaxPool, AvgPool, associated

  Show description for 8abe7db
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  8abe7db

  Copy full SHA for 8abe7db