Commits

Breadcrumbs

History for

onv

User selector

All users

All time

Commit History

Commits on Feb 8, 2017

• Adding support for non-fused Winograd algorithm from Cudnn 5.1.

  Show description for 768a41a
  

  zheng-xq

  authored and

  gunan

  committed
  768a41a

  Copy full SHA for 768a41a

Commits on Dec 20, 2016

• Fix initialization order problem for mutexes on Mac GPU opensource build.

  Show description for 1466d8a
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  1466d8a

  Copy full SHA for 1466d8a

Commits on Dec 15, 2016

• Merge changes from github.

  Show description for 2e4869a
  

  martinwicke

  authored and

  tensorflower-gardener

  committed
  2e4869a

  Copy full SHA for 2e4869a

Commits on Nov 30, 2016

• StreamExecutor: Optimize kernel argument packing

  Show description for bada4a5
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  bada4a5

  Copy full SHA for bada4a5

Commits on Nov 4, 2016

• Change FileExists to return tensorflow::Status.

  Show description for 879e0ac
  

  Jonathan Hseu

  authored and

  tensorflower-gardener

  committed
  879e0ac

  Copy full SHA for 879e0ac

• Merge changes from github.

  Show description for 818993c
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  818993c

  Copy full SHA for 818993c

Commits on Oct 28, 2016

• Merge changes from github.

  Show description for e2d51a8
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  e2d51a8

  Copy full SHA for e2d51a8

Commits on Oct 20, 2016

• Merge changes from github.

  Show description for c5ab3dd
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  c5ab3dd

  Copy full SHA for c5ab3dd

Commits on Sep 22, 2016

• This fixes a crash when CUDATimer::Init() fails for any reason.

  Show description for dffe0fa
  tensorflower-gardener

  committed
  dffe0fa

  Copy full SHA for dffe0fa

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 Sep 10, 2016

• Merge changes from github.

  Show description for 54a7178
  

  Dan Mané

  authored and

  tensorflower-gardener

  committed
  54a7178

  Copy full SHA for 54a7178

Commits on Aug 25, 2016

• Merge changes from github.

  Show description for 2c598e8
  tensorflower-gardener

  committed
  2c598e8

  Copy full SHA for 2c598e8

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 25, 2016

• Merge changes from github.

  Show description for 21716d8
  

  martinwicke

  authored and

  tensorflower-gardener

  committed
  21716d8

  Copy full SHA for 21716d8

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 11, 2016

• Merge changes from github.

  Show description for 5a65d43
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  5a65d43

  Copy full SHA for 5a65d43

Commits on Jun 9, 2016

• Remove unused MultiOpActivation from stream executor.

  Show description for 404b6fa
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  404b6fa

  Copy full SHA for 404b6fa

Commits on Jun 8, 2016

• Change a couple of error messages related to getting the NUMA affinity

  Show description for 8f10920
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  8f10920

  Copy full SHA for 8f10920

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 6, 2016

• Merge changes from github.

  Show description for b1b2dc8
  

  martinwicke

  authored and

  tensorflower-gardener

  committed
  b1b2dc8

  Copy full SHA for b1b2dc8

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

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

  Show description for 9bedadc
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  9bedadc

  Copy full SHA for 9bedadc

Commits on Jun 2, 2016

• Update copyright for 3p/tf.

  Show description for 122cdce
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  122cdce

  Copy full SHA for 122cdce

Commits on May 25, 2016

• StreamExecutor add CUDA support for cudnnConvolutionBackwardBias

  Show description for a9f3979
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  a9f3979

  Copy full SHA for a9f3979

Commits on May 16, 2016

• Add Autotune support for Winograd on Cudnn R5.

  Show description for 75b1f2f
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  75b1f2f

  Copy full SHA for 75b1f2f

Commits on May 13, 2016

• In the StreamExecutor, make lack of CUDA 7.5 a non-fatal error for SGEMM

  Show description for cfa7ffb
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  cfa7ffb

  Copy full SHA for cfa7ffb

• Maintain current cuda context in CUDADriver by checking against a threadlocal

  Show description for 069547b
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  069547b

  Copy full SHA for 069547b

Commits on May 11, 2016

• Add fp16 cuDNN convolution support to StreamExecutor. (TensorFlow ops will

  Show description for 61b12f5
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  61b12f5

  Copy full SHA for 61b12f5

• Add fp16 matrix multiplication (GEMM) support to StreamExecutor, gated on

  Show description for 5230554
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  5230554

  Copy full SHA for 5230554

• Add the autotune version for the backward passes. It is currently

  Show description for 939ede0
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  939ede0

  Copy full SHA for 939ede0

Commits on May 10, 2016

• fp16 support for BiasAdd. Includes support for atomic adds for Eigen::half,

  Show description for aec09b6
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  aec09b6

  Copy full SHA for aec09b6

Commits on May 9, 2016

• Adding autotune to the Cudnn conv algorithm selection.

  Show description for 6ff265e
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  6ff265e

  Copy full SHA for 6ff265e