Age | Commit message (Collapse) | Author | |
---|---|---|---|
2014-11-25 | Merge pull request #107 from ajtulloch/classnllcriterion | Soumith Chintala | |
Support CUDA implementation of ClassNLLCriterion | |||
2014-11-24 | [Torch] Move test.lua to the top level | Andrew Tulloch | |
Summary: This removes some CMakeLists.txt manipulation. Test Plan: ``` luarocks make rocks/nn-scm-1.rockspec && th -lnn -e "nn.test()" ... Completed 710 asserts in 81 tests with 0 errors ``` | |||
2014-11-24 | Merge pull request #109 from deltheil/master | Soumith Chintala | |
doc: link to the official Torch7 repo | |||
2014-11-24 | doc: link to the official Torch7 repo | Cédric Deltheil | |
2014-11-22 | fix for euclidian patch from #108 | Soumith Chintala | |
2014-11-21 | Merge pull request #108 from ajtulloch/luacheck-config | Soumith Chintala | |
[Torch] LuaCheck config file and fixes to make lint clean | |||
2014-11-21 | Fix various unused variables in nn | Andrew Tulloch | |
2014-11-21 | Fixed bug from unused scale variable in Euclidean | Andrew Tulloch | |
2014-11-21 | Added Luacheckrc | Andrew Tulloch | |
2014-11-21 | Support CUDA implementation of ClassNLLCriterion | Andrew Tulloch | |
2014-11-18 | Merge pull request #106 from ajtulloch/sum-fix | Soumith Chintala | |
[Torch] Fix `Sum:updateGradInput()` in the zero-stride case. | |||
2014-11-18 | [Torch] Fix `Sum:updateGradInput()` in the zero-stride case. | Andrew Tulloch | |
2014-11-18 | Merge branch 'master' of git://github.com/torch/nn | Ronan Collobert | |
2014-11-18 | fix 1D case for Mean/Sum/Max/Min | Ronan Collobert | |
2014-11-18 | global variable fixes (found via th -g) | soumith | |
2014-11-18 | Merge pull request #102 from ajtulloch/jacobian-concurrency | Soumith Chintala | |
[Torch] Make Jacobian.testIO safe for concurrent runs | |||
2014-11-17 | generic/LogSoftMax speedups | Andrew Tulloch | |
2014-11-17 | Jacobian concurrency | Andrew Tulloch | |
2014-11-14 | bugfix for SpatialConvolutionMM and non-contiguous cases | Soumith Chintala | |
2014-11-14 | Merge pull request #100 from nicholas-leonard/dropout | Soumith Chintala | |
Dropout can use torch.CudaTensor:bernoulli() | |||
2014-11-13 | Dropout can use torch.CudaTensor:bernoulli() | Nicholas Leonard | |
2014-11-12 | Merge pull request #99 from nicholas-leonard/lookup | Soumith Chintala | |
LookupTable fixes and optimizations | |||
2014-11-11 | LookupTable fixes and optimizations | Nicholas Leonard | |
2014-11-11 | Merge pull request #97 from Aysegul/BatchModeFix | Soumith Chintala | |
Batchmode is added to SpatialConvolutionMap | |||
2014-11-10 | Merge pull request #96 from kaishengtai/fastlookuptable | Soumith Chintala | |
Performance improvement to LookupTable | |||
2014-11-10 | Performance improvement to LookupTable | Kai Sheng Tai | |
2014-11-10 | Batchmode is added to SpatialConvmap | Aysegul Dundar | |
2014-11-04 | Merge pull request #94 from nicholas-leonard/LookupTable | Soumith Chintala | |
removed global variable in LookupTable | |||
2014-11-04 | removed global variable in LookupTable | nicholas-leonard | |
2014-11-04 | Merge pull request #93 from nicholas-leonard/Add | Soumith Chintala | |
nn.Add works with batches | |||
2014-11-04 | nn.Add works with batches | nicholas-leonard | |
2014-10-29 | Merge pull request #90 from fidlej/topic_gradInput_size | Soumith Chintala | |
Updated CSubTable to have gradInput[2]:size() equal to input[2]:size(). | |||
2014-10-29 | Updated CSubTable to have gradInput[2]:size() equal to input[2]:size(). | Ivo Danihelka | |
2014-10-28 | fixing concat in the case when gradInput is not emmitted | soumith | |
2014-10-27 | Typo | Jonathan Hunt | |
2014-10-27 | adding scale to accGradParameters documentation | Soumith Chintala | |
2014-10-27 | Relax equality tests from exact precision 0. | GeorgOstrovski | |
Expecting exact precision 0 can lead to failing tests if underlying implementation changes, a high expected precision of say 1e-15 seem more appropriate (nntest.Power sometimes fails because of this) | |||
2014-10-27 | Faster unit tests | nicholas-leonard | |
2014-10-27 | adding addtional weights parameter to ClassNLLCriterion, which is 1D tensor ↵ | Yossi Biton | |
assigning different weight to each class | |||
2014-10-27 | modernized MarginCriterion | Sergey Zagoruyko | |
2014-10-27 | Fixed broken CosineDistance examples. | Nicholas Dronen | |
2014-10-27 | Corrected getParamaters for partial views | James Kirkpatrick | |
Module:getParameters was incorrectly overwriting parameters that were partial views on larger storages. | |||
2014-10-27 | DepthConcat | nicholas-leonard | |
2014-10-27 | SpatialConvolutionMM now support on-the-fly zero-padding. | Clement Farabet | |
Interface similar to cunn.SpatialConvolutionMM. To get a layer where osize == isize, do this: local ksize = 5 local padding = (ksize-1)/2 local n = nn.SpatialConvolutionMM(ni,no, ksize,ksize, 1,1, padding) Note: dW/dH still not supported on the CPU version (strides). | |||
2014-10-27 | added msg to recursive tester | Nicholas Leonard | |
2014-10-27 | Fixed PairwiseDistance example | Gunnar Aastrand Grimnes | |
2014-10-27 | Fix DotProduct example error | Grisson Wang | |
missing "{" "}". x=torch.Tensor(1,2,3) => x=torch.Tensor({1,2,3}) y=torch.Tensor(4,5,6) => y=torch.Tensor({4,5,6}) | |||
2014-09-23 | adding CONTRIBUTING guidelines | Soumith Chintala | |
2014-09-03 | Merge pull request #66 from timharley/master | Soumith Chintala | |
Give better error messages when trying to use the wrong kind of Tensor. | |||
2014-09-03 | Give better error messages when trying to use the wrong kind of Tensor. | Tim Harley | |