Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Sep 11, 2017

• fix alignment warning

  

  Natalia Gimelshein

  authored and

  soumith

  committedSep 11, 2017
  5e9d86c

Commits on Sep 10, 2017

• Optimize pow for different exponents and add tests

  

  fmassa

  authored and

  soumith

  committedSep 10, 2017
  bc651e4

Commits on Sep 7, 2017

• fix static linkage and make THD statically linked

  soumith

  committedSep 7, 2017
  326b5a5

Commits on Aug 29, 2017

• Fix grid size for batch cat tensor now that getApplyGrid has been changed.

  

  csarofeen

  authored and

  soumith

  committedAug 29, 2017
  997b8b1

Commits on Aug 27, 2017

• Allowing larger grids for THCApply shows improved performance.

  

  csarofeen

  authored and

  soumith

  committedAug 27, 2017
  d891ff3

Commits on Aug 25, 2017

• Fix typos.

  

  cdluminate

  authored and

  soumith

  committedAug 25, 2017
  34cb262

• add ones_like and zeros_like

  

  alykhantejani

  authored and

  soumith

  committedAug 25, 2017
  e42a37e

• cuda 9 hgemm fix

  soumith

  committedAug 25, 2017
  d0bb7e1

• Updates for CUDA 9

  

  csarofeen

  authored and

  soumith

  committedAug 25, 2017
  a9f950a

Commits on Aug 16, 2017

• fixing the bug with squeezing a singleton dimension in torch.min and torch.max

  

  aosokin

  authored and

  soumith

  committedAug 16, 2017
  2de2d2b

• Add CUDA version of eye

  

  fmassa

  authored and

  soumith

  committedAug 16, 2017
  32a9708

Commits on Aug 15, 2017

• accumulate in accType for reductions over dimensions

  

  Natalia Gimelshein

  authored and

  soumith

  committedAug 15, 2017
  cec53c3

• Support __neg__, .neg(), and neg_() for Long, Int, Short tensor types.

  

  gchanan

  authored and

  soumith

  committedAug 15, 2017
  0252bcd

Commits on Aug 10, 2017

• call gemmStridedBatched for cuda >=8 to avoid calling kernels to set up pointers (#794)

  

  ngimel

  authored and

  soumith

  committedAug 10, 2017
  7462a22

Commits on Aug 4, 2017

• move normal variants to TH/THC

  

  killeent

  authored and

  soumith

  committedAug 4, 2017
  9dab9de

Commits on Jul 21, 2017

• Fix torch.inverse when magma is not available

  

  colesbury

  authored and

  soumith

  committedJul 21, 2017
  e29a4eb

Commits on Jul 19, 2017

• Add CUDA support for arange

  

  fmassa

  authored and

  soumith

  committedJul 19, 2017
  ea2cced

• add explicit BLAS linkage to THC when linked against magma (in binary build)

  soumith

  committedJul 19, 2017
  07939f7

• move to model with cuda indexing tensors for cuda tensor adv indexing

  

  killeent

  authored and

  soumith

  committedJul 19, 2017
  5227de5

Commits on Jul 18, 2017

• fix baddbmm for expanded tensors

  

  Natalia Gimelshein

  authored and

  soumith

  committedJul 18, 2017
  a0799aa

Commits on Jul 17, 2017

• fix cwrap

  soumith

  committedJul 17, 2017
  caf84f3

Commits on Jul 14, 2017

• fix cwrap for std/var

  soumith

  committedJul 14, 2017
  f1cad19

• Wrap unbiased flag in var, std, varall, stdall

  

  lantiga

  authored and

  soumith

  committedJul 14, 2017
  7353581

Commits on Jul 13, 2017

• add launch_bounds to greedy kernels

  

  Natalia Gimelshein

  authored and

  soumith

  committedJul 13, 2017
  5d2ce6d

• Advanced Indexing: Calculate linear offsets directly on the GPU when working with CUDA Tensors

  

  killeent

  authored and

  soumith

  committedJul 13, 2017
  f7fc564

• Check for shared_mem size in multinomial single-sample implementation

  

  BestSonny

  authored and

  soumith

  committedJul 13, 2017
  d62cf53

Commits on Jul 12, 2017

• Avoid two unnecessary copies in addmm backward

  

  colesbury

  authored and

  soumith

  committedJul 12, 2017
  d175269

Commits on Jul 11, 2017

• Alias multinomial sampling in Cuda (#784)

  

  amartya18x

  authored and

  soumith

  committedJul 11, 2017
  4e085cd

Commits on Jul 4, 2017

• add missing definition

  soumith

  committedJul 4, 2017
  0ea5868

• Have median reduce over all dims and return just the value when dim is not provided

  

  lantiga

  authored and

  soumith

  committedJul 4, 2017
  a5fc264

Commits on Jul 3, 2017

• Add a nonContigDim reduction kernel to improve latency for small tensors. (#768)

  

  csarofeen

  authored and

  soumith

  committedJul 3, 2017
  707f94c

• Make reduction functors accept only constant arguments (#753)

  

  ngimel

  authored and

  soumith

  committedJul 3, 2017
  0e5ccc3

Commits on Jun 29, 2017

• Warp intrinsic fixes (#785)

  

  ngimel

  authored and

  soumith

  committedJun 29, 2017
  35d484c

Commits on Jun 26, 2017

• support more than 8 gpus (#774)

  

  szagoruyko

  authored and

  soumith

  committedJun 26, 2017
  da935af

• Fp16 fixes for CUDA 9 (#783)

  

  csarofeen

  authored and

  soumith

  committedJun 26, 2017
  653811f

Pagination

PreviousNext