Commits

Branch selector

 master

User selector

apaszke

Datepicker

All time

Commit History

Commits on May 25, 2017

• Add scatterAdd

  

  apaszke

  authored and

  soumith

  committed
  7b578dd

  Copy full SHA for 7b578dd

• Add scatterAdd

  

  apaszke

  authored and

  soumith

  committed
  3f1f3f9

  Copy full SHA for 3f1f3f9

• Add scatterAdd

  

  apaszke

  authored and

  soumith

  committed
  bd705d3

  Copy full SHA for bd705d3

Commits on May 24, 2017

• Improve handling of graph roots in autograd engine (pytorch#1635)

  

  apaszke

  authored and

  soumith

  committed
  fa93653

  Copy full SHA for fa93653

Commits on May 21, 2017

• Update build status matrix

  apaszke

  committed
  0c5598c

  Copy full SHA for 0c5598c

Commits on May 17, 2017

• Bug fixes (pytorch#1573)

  Show description for c573d53
  

  apaszke

  authored and

  colesbury

  committed
  c573d53

  Copy full SHA for c573d53

Commits on May 15, 2017

• Add F.cosine_similarity (pytorch#1502)

  

  apaszke

  authored and

  soumith

  committed
  6b84dc2

  Copy full SHA for 6b84dc2

Commits on May 10, 2017

• Fix Linear function

  apaszke

  committed
  5bb1348

  Copy full SHA for 5bb1348

• Fix comparison functions

  apaszke

  committed
  a86adf4

  Copy full SHA for a86adf4

• Expose variable attribute of AccumulateGrad

  apaszke

  committed
  1c304a9

  Copy full SHA for 1c304a9

• Don't modify non-volatile grads in zero_grad

  apaszke

  committed
  feef54e

  Copy full SHA for feef54e

• Minor fix in Prod backward

  apaszke

  committed
  5026209

  Copy full SHA for 5026209

• Add new flags to Variable.backward

  apaszke

  committed
  e722038

  Copy full SHA for e722038

• Replace retain_variables with retain_graph

  apaszke

  committed
  9fa0e40

  Copy full SHA for 9fa0e40

• Improve output wrapping logic in autograd

  apaszke

  committed
  35cf380

  Copy full SHA for 35cf380

Commits on May 6, 2017

• Add a note about THPFunction_asFunction

  apaszke

  committed
  5f15a9e

  Copy full SHA for 5f15a9e

Commits on May 3, 2017

• Fix memory leak introduced by 72e8190 (pytorch#1464)

  

  apaszke

  authored and

  soumith

  committed
  e9d648c

  Copy full SHA for e9d648c

• Use at most one shared_ptr block at a time to manage THPFunctions (pytorch#1454)

  Show description for 72e8190
  

  apaszke

  authored and

  soumith

  committed
  72e8190

  Copy full SHA for 72e8190

• Fix typo in autograd docs

  apaszke

  committed
  e1278d4

  Copy full SHA for e1278d4

Commits on May 1, 2017

• Fix bugs, rename differentiate to grad, make it more flexible

  

  apaszke

  authored and

  soumith

  committed
  5c74534

  Copy full SHA for 5c74534

• Bug fixes

  

  apaszke

  authored and

  soumith

  committed
  87164f5

  Copy full SHA for 87164f5

• Fix memory issues with Conv and BatchNorm

  

  apaszke

  authored and

  soumith

  committed
  267e7c0

  Copy full SHA for 267e7c0

• Add torch.autograd.differentiate

  

  apaszke

  authored and

  soumith

  committed
  e5db8f9

  Copy full SHA for e5db8f9

• Convert some of the functions to new format

  Show description for 20aa5b0
  

  apaszke

  authored and

  soumith

  committed
  20aa5b0

  Copy full SHA for 20aa5b0

• Add support for the new Function format

  

  apaszke

  authored and

  soumith

  committed
  de9998e

  Copy full SHA for de9998e

• Make Variables not subclass Function anymore

  Show description for 702a2e3
  

  apaszke

  authored and

  soumith

  committed
  702a2e3

  Copy full SHA for 702a2e3

• Refactor attribute names in autograd

  

  apaszke

  authored and

  soumith

  committed
  2ca787f

  Copy full SHA for 2ca787f

• Add base for CUDA allReduce and broadcast in DataChannelGloo

  apaszke

  committed
  4ebf3ff

  Copy full SHA for 4ebf3ff

• Rebase fixes

  apaszke

  committed
  ac3ba9a

  Copy full SHA for ac3ba9a

• Initial gloo bindings

  apaszke

  committed
  7e8830c

  Copy full SHA for 7e8830c

• Don't build tests by default

  apaszke

  committed
  f891d9b

  Copy full SHA for f891d9b

• Use TCP_NODELAY for data sockets

  apaszke

  committed
  1391ff9

  Copy full SHA for 1391ff9

• Always loop over all possible addresses in worker

  apaszke

  committed
  43019bd

  Copy full SHA for 43019bd

Commits on Apr 29, 2017

• Parallelize TensorMethods.cpp builds (pytorch#1400)

  

  apaszke

  authored and

  soumith

  committed
  9169f60

  Copy full SHA for 9169f60

• Use THCUNN backward kernels for Tanh and Sigmoid in Autograd (pytorch#1399)

  

  apaszke

  authored and

  soumith

  committed
  457d78a

  Copy full SHA for 457d78a

Pagination

Previous

Next