提交 · a684643ae4b65c15dd83786fce6bed2df9bbd67d · testgroup / pytensor

24 9月, 2014 5 次提交
- Fixed typo. · a684643a
  由 Dustin Webb 提交于 9月 19, 2014
  
  a684643a
- Incorporated reviewer comments. · 8a83fdcb
  由 Dustin Webb 提交于 9月 19, 2014
  
  8a83fdcb
- Fixed problem with cudnn softmax tests. This fix should be generalized for use… · b9609e87
  由 Dustin Webb 提交于 9月 18, 2014
```
Fixed problem with cudnn softmax tests. This fix should be generalized for use of the class. Also there are two cases it cannot handle.
```
  b9609e87
- Changed documentation comment. · 4948b7c8
  由 Dustin Webb 提交于 9月 18, 2014
```
Fixed eq and hash functions.
```
  4948b7c8
- Implemented cuDNN softmax but it is not currently passing tests. · f3b5fe3b
  由 Dustin Webb 提交于 9月 17, 2014
  
  f3b5fe3b
23 9月, 2014 8 次提交
- Merge pull request #2117 from abergeron/dnn · 239b6d80
  由 Frédéric Bastien 提交于 9月 23, 2014
```
Add support for subsampling and gradient to GpuDnnConv
```
  239b6d80
- Don't mark c_cleanup as optional, it is not. · dedc323d
  由 Arnaud Bergeron 提交于 9月 22, 2014
  
  dedc323d
- Add comment explaining empty c_cleanup. · 2f71fded
  由 Arnaud Bergeron 提交于 9月 22, 2014
  
  2f71fded
- Fix last few comments about the tests. · 3a0dcdbc
  由 Arnaud Bergeron 提交于 9月 22, 2014
  
  3a0dcdbc
- Add a test for CDataType and fix a bug that was calling freefunc too often. · b90abd33
  由 Arnaud Bergeron 提交于 9月 22, 2014
  
  b90abd33
- Merge pull request #2120 from npinto/patch-1 · 025d484e
  由 abergeron 提交于 9月 22, 2014
```
FIX: tutorial pycuda <> theano compatibility
```
  025d484e
- Add documentation and checks to CDataType. · 4f1edd17
  由 Arnaud Bergeron 提交于 9月 22, 2014
  
  4f1edd17
- Merge pull request #2119 from nouiz/warning · 9944efc5
  由 abergeron 提交于 9月 22, 2014
```
Warning
```
  9944efc5
21 9月, 2014 1 次提交
- FIX: tutorial pycuda <> theano compatilbity · a0df82f9
  由 Nicolas Pinto 提交于 9月 20, 2014
```
thunk needs lazy attribute
```
  a0df82f9
20 9月, 2014 9 次提交
- Also make sure it copies .h in MANIFEST.in · 539a9614
  由 Arnaud Bergeron 提交于 9月 19, 2014
  
  539a9614
- Fix the tests for the gradient case. · f8ab3cf1
  由 Arnaud Bergeron 提交于 9月 19, 2014
  
  f8ab3cf1
- Fake "correct" strides to satisfy the cudnn functions. · 3bdd3ce0
  由 Arnaud Bergeron 提交于 9月 19, 2014
  
  3bdd3ce0
- Outsource the creation of the cudnn convolution descriptor. · 1f4f4ff6
  由 Arnaud Bergeron 提交于 9月 19, 2014
  
  1f4f4ff6
- Add CDataType to return a C pointer. · 1a1dbe60
  由 Arnaud Bergeron 提交于 9月 19, 2014
```
This handles Capsules, Cobjects and python 3 compat.
```
  1a1dbe60
- Copy .h files also. · c3fd1cb8
  由 Arnaud Bergeron 提交于 9月 19, 2014
  
  c3fd1cb8
- Add timming of the compute min peak algo. · 024a1d61
  由 Frederic 提交于 9月 19, 2014
  
  024a1d61
- Tell how many profile are summed · c0d12f1b
  由 Frederic 提交于 9月 19, 2014
  
  c0d12f1b
- Fix opt warning in some cases. · 174e54f0
  由 Frederic 提交于 9月 19, 2014
  
  174e54f0
19 9月, 2014 3 次提交
- Fix not passing subsample to the gradients. · 3f849780
  由 Arnaud Bergeron 提交于 9月 18, 2014
  
  3f849780
- Merge pull request #2106 from RoyXue/GSoC2014_part2 · d173f8cb
  由 Frédéric Bastien 提交于 9月 18, 2014
```
Algorithm Speed Up
```
  d173f8cb
- Merge pull request #1760 from MarcCote/cumsum_gpu · 4c0ca769
  由 abergeron 提交于 9月 18, 2014
```
CumsumOp on GPU
```
  4c0ca769
18 9月, 2014 14 次提交
- Adressed @abergeron comment on copying variable sub and not using it. · a074daae
  由 Marc-Alexandre Cote 提交于 9月 17, 2014
  
  a074daae
- Put back __str__ and revert change to code cache version. · c7c25604
  由 Marc-Alexandre Cote 提交于 9月 17, 2014
  
  c7c25604
- Replaced __eq__(), __hash__() and __str__() by using __props__ instead · 92372b77
  由 Marc-Alexandre Cote 提交于 9月 16, 2014
  
  92372b77
- Make GpuCumsumOp available for optimization · 4b1eb7ae
  由 Marc-Alexandre Cote 提交于 8月 14, 2014
  
  4b1eb7ae
- Fix copy-paste error. · 79216c2c
  由 Marc-Alexandre Cote 提交于 8月 01, 2014
  
  79216c2c
- Use output's strides instead of assuming they are the same as the input's one. · 584520bc
  由 Marc-Alexandre Cote 提交于 6月 30, 2014
  
  584520bc
- Remove non-needed CUDA_THREAD_SYNC · e9c843bb
  由 Marc-Alexandre Cote 提交于 6月 26, 2014
  
  e9c843bb
- Added unit tests and comments. · 771d5c27
  由 Marc-Alexandre Cote 提交于 6月 25, 2014
  
  771d5c27
- Force cast of dim3's attributes to int. · 2ad36935
  由 Marc-Alexandre Cote 提交于 6月 23, 2014
  
  2ad36935
- smaller test size · 37b40abd
  由 Frederic 提交于 4月 17, 2014
  
  37b40abd
- Assert that the gpuop is used. · e3156b66
  由 Frederic 提交于 4月 17, 2014
  
  e3156b66
- Disable the perform on the gpu op. · 54d767c1
  由 Frederic 提交于 4月 17, 2014
  
  54d767c1
- pep8 · db7c8b58
  由 Frederic 提交于 4月 17, 2014
  
  db7c8b58
- Force float32 tests and use the mode for gpucumsum tests · 1e141e5c
  由 Frederic 提交于 4月 17, 2014
  
  1e141e5c