Commits

Branch selector

 master

User selector

apaszke

Datepicker

All time

Commit History

Commits on Mar 29, 2017

• Fix deadlock in autograd (#1140)

  

  apaszke

  authored and

  soumith

  committed
  8aa1cef

  Copy full SHA for 8aa1cef

Commits on Mar 24, 2017

• Pass NULL rinfo_ to btrifact by default (#1089)

  

  apaszke

  authored and

  soumith

  committed
  bf63254

  Copy full SHA for bf63254

• Revert "Add back zero fill for ger" (#1093)

  Show description for 282402d
  

  apaszke

  authored and

  soumith

  committed
  282402d

  Copy full SHA for 282402d

• Use zero instead of mul when beta == 0 in addr

  apaszke

  committed
  b4fe5ad

  Copy full SHA for b4fe5ad

• Make rinfo_ argument optional in btrifact

  apaszke

  committed
  e8196f9

  Copy full SHA for e8196f9

• Make rinfo_ optional in btrifact

  apaszke

  committed
  269b77a

  Copy full SHA for 269b77a

Commits on Mar 23, 2017

• Fix formula for stddevs grad in Normal function (#1076)

  

  apaszke

  authored and

  soumith

  committed
  ced0054

  Copy full SHA for ced0054

Commits on Mar 22, 2017

• Break only after the update in L-BFGS

  

  apaszke

  authored and

  soumith

  committed
  1e8cb82

  Copy full SHA for 1e8cb82

• Return total param norm from clip_grad_norm

  

  apaszke

  authored and

  soumith

  committed
  dd399a8

  Copy full SHA for dd399a8

• Fix torch.cat bugs

  Show description for faac0f5
  

  apaszke

  authored and

  soumith

  committed
  faac0f5

  Copy full SHA for faac0f5

• Make random_ exclusive and make generator kwarg only in all random

  Show description for c36f47b
  

  apaszke

  authored and

  soumith

  committed
  c36f47b

  Copy full SHA for c36f47b

• Fix size mismatch in CosineEmbeddingLoss backward

  

  apaszke

  authored and

  soumith

  committed
  3d1888c

  Copy full SHA for 3d1888c

• Narrow V when returning only some right singular vectors

  apaszke

  committed
  0056b08

  Copy full SHA for 0056b08

• Cast accumulator in LookupTable renorm to accreal

  apaszke

  committed
  bd0df61

  Copy full SHA for bd0df61

Commits on Mar 17, 2017

• Fixes for Prod and Expand functions (#1026)

  Show description for 7e46eb1
  

  apaszke

  authored and

  soumith

  committed
  7e46eb1

  Copy full SHA for 7e46eb1

Commits on Mar 12, 2017

• Allow submodules and parameters to shadow attrs on assignment

  

  apaszke

  authored and

  soumith

  committed
  d602b3a

  Copy full SHA for d602b3a

• Fix memory leak in torch.from_numpy

  

  apaszke

  authored and

  soumith

  committed
  f531d98

  Copy full SHA for f531d98

• Remove some unnecessary AutoGPU calls

  

  apaszke

  authored and

  soumith

  committed
  6bdd5ec

  Copy full SHA for 6bdd5ec

• Fix Embedding bug when max_norm was used

  

  apaszke

  authored and

  soumith

  committed
  bfbde9d

  Copy full SHA for bfbde9d

Commits on Mar 3, 2017

• Fix issues with lazy grad initialization (#912)

  

  apaszke

  authored and

  soumith

  committed
  c238ee3

  Copy full SHA for c238ee3

• Fix slicing with step (#905)

  

  apaszke

  authored and

  soumith

  committed
  490c15f

  Copy full SHA for 490c15f

Commits on Mar 1, 2017

• Add Nesterov Momentum (#887)

  apaszke

  authored
  3277d83

  Copy full SHA for 3277d83

• Allow backprop through cuDNN RNN in eval mode

  Show description for 1487278
  apaszke

  committed
  1487278

  Copy full SHA for 1487278

• Handle duplicate backward roots in autograd

  apaszke

  committed
  977630b

  Copy full SHA for 977630b

• Add support for variable length sequences in RNNs (#873)

  apaszke

  authored
  da72583

  Copy full SHA for da72583

Commits on Feb 28, 2017

• Expose torch.HalfTensor

  

  apaszke

  authored and

  soumith

  committed
  67f9455

  Copy full SHA for 67f9455

Commits on Feb 27, 2017

• Fix one more reference cycle and ensure correct flag propagation (#868)

  

  apaszke

  authored and

  soumith

  committed
  502ebed

  Copy full SHA for 502ebed

• Improve autograd memory usage (#859)

  

  apaszke

  authored and

  soumith

  committed
  2b23712

  Copy full SHA for 2b23712

Commits on Feb 26, 2017

• Make Optimizer.load_state_dict use __setstate__

  apaszke

  committed
  bd7a5ad

  Copy full SHA for bd7a5ad

• Fall back to indexing compatible with numpy

  apaszke

  committed
  1f6f82d

  Copy full SHA for 1f6f82d

• Allow using expand to broadcast tensors

  apaszke

  committed
  1f89399

  Copy full SHA for 1f89399

• Add docs for ModuleList and ParameterList

  apaszke

  committed
  b3d41a5

  Copy full SHA for b3d41a5

• Reshape grad_output in basic ops

  apaszke

  committed
  fec2d49

  Copy full SHA for fec2d49

• Fix for Long and Byte tensor indexing of Variables

  apaszke

  committed
  86ee75f

  Copy full SHA for 86ee75f

• Prevent creation of reference cycles with leaf Variables that don't require grad

  Show description for 3194191
  apaszke

  committed
  3194191

  Copy full SHA for 3194191

Pagination

Previous

Next