Soumith Chintala soumith

41 commits

• Pushed 11 commits to facebook/fbcunn Dec 29 – Jan 3
• Pushed 7 commits to torch/cutorch Dec 29 – Jan 3
• Pushed 9 commits to torch/nn Dec 31 – Jan 2
• Pushed 6 commits to soumith/cudnn.torch Dec 31 – Jan 2
• Pushed 2 commits to torch/rocks Jan 1 – Jan 2
• Pushed 2 commits to soumith/convnet-benchmarks Jan 1
• Pushed 1 commit to torch/trepl Dec 31
• Pushed 1 commit to facebook/fbcuda Dec 29
• Pushed 1 commit to facebook/fbnn Dec 29
• Pushed 1 commit to soumith/cuda-convnet2.torch Dec 28

8 Pull Requests

1. Merged #95 adds a withDevice closure to easily run code on a given device Jan 3, 2015 torch/cutorch
2. Merged #132 refactoring all the common container code into nn.Container Jan 3, 2015 torch/nn
3. Merged #131 speedup and optimizations for SparseLinear Jan 3, 2015 torch/nn
4. Merged #130 another fix for making nn.Linear non-blocking Jan 3, 2015 torch/nn
5. Merged #129 added optional precision parameter to nn.Jacobian Jan 3, 2015 torch/nn
6. Merged #14 adding in-place ReLU Jan 1, 2015 soumith/cudnn.torch
7. Merged #94 more copy optimizations, and additions to unit tests Dec 31, 2014 torch/cutorch
8. Merged #93 very fast copy kernels for non-contiguous cases + test Dec 30, 2014 torch/cutorch

2 Issues reported

• Closed #1 cudnn ReLU now with zero memory Jan 2, 2015
• Closed #12 cudnn R2 memory errors Dec 29, 2014