提交 · 306c850dff74c993413a33c13d34ed4511f97de7 · testgroup / pytensor

23 6月, 2010 12 次提交
- added check in test. · 306c850d
  由 Frederic Bastien 提交于 6月 23, 2010
  
  306c850d
- added missing import and make test a little faster. · 49435060
  由 Frederic Bastien 提交于 6月 23, 2010
  
  49435060
- make a GPU convolution kernel more precise to don't have test fail due to random data. · 8f094db5
  由 Frederic Bastien 提交于 6月 22, 2010
  
  8f094db5
- print exception caught in test. · f7778651
  由 Frederic Bastien 提交于 6月 22, 2010
  
  f7778651
- fix some test case of fusion badbly tested on cpu and fix some cuda test disabled by default. · 2ad63271
  由 Frederic Bastien 提交于 6月 22, 2010
  
  2ad63271
- fix test as now we don't limit the number of fused op on the cpu. · 57e2b381
  由 Frederic Bastien 提交于 6月 22, 2010
  
  57e2b381
- separate the grad test to another test. · a864aeef
  由 Frederic Bastien 提交于 6月 22, 2010
  
  a864aeef
- merge · 08578830
  由 James Bergstra 提交于 6月 22, 2010
  
  08578830
- Added an Optimizer that prints the state of the env to stdout. · ed1309f5
  由 James Bergstra 提交于 6月 22, 2010
```
It is not enabled by default by any optimizer.
```
  ed1309f5
- Uncommenting softmax optimization · 8d9da724
  由 James Bergstra 提交于 6月 22, 2010
  
  8d9da724
- make cuda_available==True only if their is a cuda card available and we have… · 34bb4b94
  由 Frederic Bastien 提交于 6月 22, 2010
```
make cuda_available==True only if their is a cuda card available and we have been able to compile/load the cuda_ndarray.cu file.

It can happen that we compile the file but that their is no card.
```
  34bb4b94
- make test a faster. · 71fc4a1b
  由 Frederic Bastien 提交于 6月 22, 2010
  
  71fc4a1b
22 6月, 2010 14 次提交
- removing debugprints · 22226ae5
  由 James Bergstra 提交于 6月 22, 2010
  
  22226ae5
- merge · 706d92fa
  由 James Bergstra 提交于 6月 22, 2010
  
  706d92fa
- added info to a failing assert in the gemm_canonicalize test · 36582dac
  由 James Bergstra 提交于 6月 22, 2010
  
  36582dac
- Added test for many broadcastable patterns with dot · a02f01cf
  由 James Bergstra 提交于 6月 22, 2010
  
  a02f01cf
- Minor mods to Composite optimization · 24401f1c
  由 James Bergstra 提交于 6月 22, 2010
  
  24401f1c
- Fixed check_x_over_absX to not replace complex arguments · 9ca91ae5
  由 James Bergstra 提交于 6月 22, 2010
  
  9ca91ae5
- Fixed minor error in div_sum_dimshuffle optimization · 7397cdf4
  由 James Bergstra 提交于 6月 22, 2010
  
  7397cdf4
- Added code to tensor.blas to prevent emitting c code when dtype is complex · c55401a0
  由 James Bergstra 提交于 6月 22, 2010
  
  c55401a0
- Added Gemv to tensor.blas · 4d040292
  由 James Bergstra 提交于 6月 22, 2010
  
  4d040292
- Minor correction to TypeError messages in tensor.dot.make_node · ce7f9a78
  由 James Bergstra 提交于 6月 22, 2010
  
  ce7f9a78
- Added numpy version to c_code version · b5eccebd
  由 James Bergstra 提交于 6月 22, 2010
  
  b5eccebd
- revert a recent change that broke the Deep Learning Tutorial. · 6b58df5c
  由 Frederic Bastien 提交于 6月 21, 2010
  
  6b58df5c
- got rid of extraneous printing in test · 434b03b3
  由 Michael I Mandel 提交于 6月 21, 2010
  
  434b03b3
- added tests for x over abs(x) canonization · 63cf9ea0
  由 Michael I Mandel 提交于 6月 21, 2010
  
  63cf9ea0
21 6月, 2010 14 次提交
- merge · 0cc1268b
  由 James Bergstra 提交于 6月 21, 2010
  
  0cc1268b
- added debugprints · 3afc64ab
  由 James Bergstra 提交于 6月 21, 2010
  
  3afc64ab
- Removed two debug prints · 4591855f
  由 James Bergstra 提交于 6月 21, 2010
  
  4591855f
- Reverting Dumi's change to getitem, needs more thought. · 6b748cb2
  由 James Bergstra 提交于 6月 21, 2010
  
  6b748cb2
- added an optimization phase: specialize_device and put the new POW optimization into that phase. · 3b18fef4
  由 Frederic Bastien 提交于 6月 21, 2010
  
  3b18fef4
- fix import · b91efd13
  由 Frederic Bastien 提交于 6月 21, 2010
  
  b91efd13
- fix optimization. · cabe3a0f
  由 Frederic Bastien 提交于 6月 21, 2010
  
  cabe3a0f
- fix test in fast_compile. · 0fdd5295
  由 Frederic Bastien 提交于 6月 21, 2010
  
  0fdd5295
- fix tests in FAST_COMPILE · 8c6975e2
  由 Frederic Bastien 提交于 6月 21, 2010
  
  8c6975e2
- removed debug print · 95659370
  由 James Bergstra 提交于 6月 20, 2010
  
  95659370
- Made log1p sigmoid optimizations use the new skip_identies option in PatternSub · 3006d252
  由 James Bergstra 提交于 6月 20, 2010
  
  3006d252
- added optimization in tensor/blas to use dot22 for broadcasted tensors · 00697549
  由 James Bergstra 提交于 6月 20, 2010
  
  00697549
- corrected comment in tensor/opt · 33becd6e
  由 James Bergstra 提交于 6月 20, 2010
  
  33becd6e
- added skip_identities code to PatternSub · 8c94ba0c
  由 James Bergstra 提交于 6月 20, 2010
  
  8c94ba0c