Commits

Branch selector

 master

User selector

mingfeima

Datepicker

All time

Commit History

Commits on Dec 30, 2021

• add BFloat16 support for AdaptiveAvgPool2d on CPU (pytorch#56902)

  Show description for 401a6b6
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  401a6b6

  Copy full SHA for 401a6b6

Commits on Aug 30, 2021

• add channel last support for MaxUnpool2d (pytorch#49984)

  Show description for c5ed31e
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  c5ed31e

  Copy full SHA for c5ed31e

Commits on Aug 25, 2021

• add BFloat16 support for bernoulli and Dropout on CPU (pytorch#56372)

  Show description for b0782f0
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  b0782f0

  Copy full SHA for b0782f0

Commits on Aug 24, 2021

• add bf16 support for bucketize (pytorch#55588)

  Show description for 14d4723
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  14d4723

  Copy full SHA for 14d4723

• optimize BFloat16 elemwise operators CPU: sigmoid, sigmoid_backward, tanh_backward, addcmul, addcdiv (pytorch#55221)

  Show description for 94d6215
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  94d6215

  Copy full SHA for 94d6215

• fix batchnorm2d issue when input is non contiguous (pytorch#63392)

  Show description for d3be02d
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  d3be02d

  Copy full SHA for d3be02d

• add channels last for GroupNorm (pytorch#49821)

  Show description for 5b7cdc5
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  5b7cdc5

  Copy full SHA for 5b7cdc5

Commits on Jul 20, 2021

• add channels last support for thnn_conv2d (non-dilated) (pytorch#49582)

  Show description for 8b88c24
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  8b88c24

  Copy full SHA for 8b88c24

Commits on Jul 7, 2021

• MaxPool2d: use channels_last format for both output and indice when input is channels_last (pytorch#61245)

  Show description for 8bec478
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  8bec478

  Copy full SHA for 8bec478

Commits on Jun 30, 2021

• SumKernel (BFloat16): use float as accumulation type (pytorch#55217)

  Show description for 4f5c688
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  4f5c688

  Copy full SHA for 4f5c688

Commits on Jun 29, 2021

• add BFloat16 support for LayerNorm CPU (pytorch#55210)

  Show description for 652d911
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  652d911

  Copy full SHA for 652d911

Commits on Jun 25, 2021

• add channels last for AdapativeMaxPool2d (pytorch#48920)

  Show description for dd045ab
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  dd045ab

  Copy full SHA for dd045ab

Commits on Jun 24, 2021

• Optimize some redunction operators on CPU BFloat16 (pytorch#55202)

  Show description for 5a077bb
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  5a077bb

  Copy full SHA for 5a077bb

Commits on Jun 11, 2021

• optimize channels last for BatchNorm2d on CPU (pytorch#59286)

  Show description for f321856
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  f321856

  Copy full SHA for f321856

Commits on Jun 10, 2021

• add channels last support for AvgPool2d on CPU (pytorch#58725)

  Show description for bb19dc1
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  bb19dc1

  Copy full SHA for bb19dc1

Commits on May 14, 2021

• optimize channels last for BatchNorm2d on CPU (pytorch#48919)

  Show description for 0be334a
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  0be334a

  Copy full SHA for 0be334a

• add channels last support for AvgPool2d on CPU (pytorch#48918)

  Show description for 8ac0917
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  8ac0917

  Copy full SHA for 8ac0917

Commits on Apr 20, 2021

• add channels last for MaxPool2d (pytorch#56361)

  Show description for 1e03a25
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  1e03a25

  Copy full SHA for 1e03a25

Commits on Apr 2, 2021

• add channels last for MaxPool2d (pytorch#48917)

  Show description for f43eb59
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  f43eb59

  Copy full SHA for f43eb59

Commits on Mar 17, 2021

• fix nest openmp performance bug in thnn_conv2d (pytorch#52577)

  Show description for b936abd
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  b936abd

  Copy full SHA for b936abd

Commits on Dec 14, 2020

• add channels last for AdaptiveAvgPool2d (pytorch#48916)

  Show description for 690eaf9
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  690eaf9

  Copy full SHA for 690eaf9

Commits on Aug 12, 2020

• Optimize LayerNorm performance on CPU both forward and backward (pytorch#35750)

  Show description for 686705c
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  686705c

  Copy full SHA for 686705c

Commits on Jun 16, 2020

• optimize upsample performance linear mode on CPU (pytorch#34864)

  Show description for 4b5530d
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  4b5530d

  Copy full SHA for 4b5530d

Commits on May 20, 2020

• port masked_select from TH to ATen and optimize perf on CPU (pytorch#33269)

  Show description for fe66bdb
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  fe66bdb

  Copy full SHA for fe66bdb

Commits on May 19, 2020

• cover nn.Conv1d in mkldnn model conversion logic (pytorch#38528)

  Show description for 9ad14f6
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  9ad14f6

  Copy full SHA for 9ad14f6

Commits on Mar 4, 2020

• optimize UpSampleNearest 1d 2d and 3d performance on CPU (pytorch#31452)

  Show description for 39f78db
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  39f78db

  Copy full SHA for 39f78db

Commits on Feb 28, 2020

• Fix torch.cat() performance regression on single core CPU (pytorch#33534)

  Show description for c6d3012
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  c6d3012

  Copy full SHA for c6d3012

Commits on Feb 27, 2020

• port masked_fill from TH to ATen (pytorch#33330)

  Show description for cd0acf4
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  cd0acf4

  Copy full SHA for cd0acf4

Commits on Feb 11, 2020

• optimize cat performance on CPU with TensorIterator (pytorch#30806)

  Show description for 7b50e76
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  7b50e76

  Copy full SHA for 7b50e76

Commits on Jan 3, 2020

• optimize index_select performance on CPU with TensorIterator (pytorch#30598)

  Show description for 79e30ff
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  79e30ff

  Copy full SHA for 79e30ff

Commits on Jul 19, 2019

• BERT CPU performance optimization: use mkldnn for nn.Linear() when input is dense layout (pytorch#21851)

  Show description for 25f0dc3
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  25f0dc3

  Copy full SHA for 25f0dc3

Commits on Jul 12, 2019

• optimize topk on cpu using parallel and partial sort (pytorch#19736)

  Show description for 10c14ad
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  10c14ad

  Copy full SHA for 10c14ad

Commits on May 16, 2019

• Improve bmm performance on CPU by applying TensorAccessor (pytorch#20266)

  Show description for 21ef4cc
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  21ef4cc

  Copy full SHA for 21ef4cc

Commits on Apr 18, 2019

• Improve bmm() performance on CPU when input tensor is non-contiguous (pytorch#19338)

  Show description for b8fb6ea
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  b8fb6ea

  Copy full SHA for b8fb6ea

Commits on Oct 15, 2018

• fix masked_fill_ bug on non-contiguous tensor (pytorch#12594)

  Show description for 02695c1
  

  mingfeima

  authored and

  facebook-github-bot

  committed
  02695c1

  Copy full SHA for 02695c1

Pagination

Previous

Next