提交 · 2e7311dc87a55ba7cb67a381c3ebbb653a3ee730 · testgroup / pytensor

07 4月, 2010 8 次提交
- added test for gpusoftmax · 2e7311dc
  由 Frederic Bastien 提交于 4月 07, 2010
  
  2e7311dc
- added property to mimix the old interface or the profiler that is used elsewhere in theano. · cb6f77cf
  由 Frederic Bastien 提交于 4月 07, 2010
  
  cb6f77cf
- lowered profiler overhead · c80d4a60
  由 Frederic Bastien 提交于 4月 07, 2010
  
  c80d4a60
- lowered profiler overhead · 2491a3f4
  由 Frederic Bastien 提交于 4月 06, 2010
  
  2491a3f4
- lowered the profiler overhead. · d44efbcd
  由 Frederic Bastien 提交于 4月 06, 2010
  
  d44efbcd
- added GpuSoftmaxWithBias to make a real speed up to the tutorial logistic_cg · 91b67a6d
  由 Frederic Bastien 提交于 4月 06, 2010
  
  91b67a6d
- better error message. · 80fc79ec
  由 Frederic Bastien 提交于 4月 06, 2010
  
  80fc79ec
- implement CudaNdarray.copy to have the same api as ndarray. · 970ea842
  由 Frederic Bastien 提交于 4月 06, 2010
  
  970ea842
06 4月, 2010 5 次提交
- put print into the logging system. · f9898c93
  由 Frederic Bastien 提交于 4月 06, 2010
  
  f9898c93
- typos reported by Ivo Danihelka · f5bad301
  由 Razvan Pascanu 提交于 4月 05, 2010
  
  f5bad301
- fixed type introduced in last commit · 93ffd843
  由 Razvan Pascanu 提交于 4月 05, 2010
  
  93ffd843
- added info about shape_padleft, shape_padright, addbroadcast, unbroadcast · b2abbe89
  由 Razvan Pascanu 提交于 4月 05, 2010
  
  b2abbe89
- Scan now uses Rebroadcast · 6a1ac4b0
  由 Razvan Pascanu 提交于 4月 05, 2010
  
  6a1ac4b0
05 4月, 2010 1 次提交
- More details about the 'f' option for Dimshuffle · 9243f3ef
  由 Razvan Pascanu 提交于 4月 05, 2010
  
  9243f3ef
04 4月, 2010 1 次提交
- new optimization for scan + new feature for Dimshuffle · 31904be5
  由 Razvan Pascanu 提交于 4月 03, 2010
  
  31904be5
02 4月, 2010 7 次提交
- rng_mrg - renamed 'prob' parameter of .binomial to 'p' · 3903e59f
  由 James Bergstra 提交于 4月 01, 2010
  
  3903e59f
- ConvOp - modified grad set unrolling params to None, and let __init__ set them. · 1cfa8fbc
  由 James Bergstra 提交于 4月 01, 2010
  
  1cfa8fbc
- gpu conv - template-unrolling depth logarithmic in kernel size rather than linear. · 9e4f9e45
  由 James Bergstra 提交于 4月 01, 2010
  
  9e4f9e45
- ConvOp select automatically the fastest c code algo when none are specified. · 8b9676e3
  由 Frederic Bastien 提交于 4月 01, 2010
```
The timming is done on maggie, so it could differ on other computer.
```
  8b9676e3
- Fred answered question about dot22scalar test · 94c8bc56
  由 James Bergstra 提交于 4月 01, 2010
  
  94c8bc56
- more clearn error. · edd6707d
  由 Frederic Bastien 提交于 4月 01, 2010
  
  edd6707d
- added some result that people can compare again. · 2ab2bbf3
  由 Frederic Bastien 提交于 4月 01, 2010
  
  2ab2bbf3
01 4月, 2010 5 次提交
- moved debug print at the good place · 27b9f8e2
  由 Frederic Bastien 提交于 4月 01, 2010
  
  27b9f8e2
- merge · be0902fb
  由 Razvan Pascanu 提交于 4月 01, 2010
  
  be0902fb
- typo · 0fe8b8f8
  由 Razvan Pascanu 提交于 4月 01, 2010
  
  0fe8b8f8
- always check the perform if their is no c_code. · 223cdae9
  由 Frederic Bastien 提交于 3月 31, 2010
  
  223cdae9
- changed the speed test of conv op c code parameter. · 79b125f3
  由 Frederic Bastien 提交于 3月 31, 2010
  
  79b125f3
31 3月, 2010 8 次提交
- Added first shot at normal distribution in mrg_rng, with tests adapted from the… · 5f6e6abe
  由 fsavard 提交于 3月 30, 2010
```
Added first shot at normal distribution in mrg_rng, with tests adapted from the ones James used for uniform
```
  5f6e6abe
- make the test more fine granular · adab6fba
  由 Frederic Bastien 提交于 3月 30, 2010
  
  adab6fba
- make test faster. · 285316c1
  由 Frederic Bastien 提交于 3月 30, 2010
  
  285316c1
- ConvOp - erased duplicate line (cut & paste err?) · 46db4371
  由 James Bergstra 提交于 3月 30, 2010
  
  46db4371
- nnet.conv - added comments to file __doc__ · 346ce04c
  由 James Bergstra 提交于 3月 30, 2010
  
  346ce04c
- Shape_i.perform - optimization. skip _asarray() when output is provided. · 884c824d
  由 James Bergstra 提交于 3月 30, 2010
  
  884c824d
- MakeVector.perform - optimization. modifying to skip asarray() when output space… · b931d52f
  由 James Bergstra 提交于 3月 30, 2010
```
MakeVector.perform - optimization. modifying to skip asarray() when output space is already allocated.
```
  b931d52f
- blas - bumping gemm_version after some confusion, just to be safe. · 69ec8fcb
  由 James Bergstra 提交于 3月 30, 2010
  
  69ec8fcb
29 3月, 2010 1 次提交
- removed some benchmarking code to DeepLearningBenchmarks on github · 6279b842
  由 James Bergstra 提交于 3月 28, 2010
  
  6279b842
30 3月, 2010 4 次提交
- Added new required installation step on Windows (proper blas.ldflags) · fb491a08
  由 Olivier Delalleau 提交于 3月 30, 2010
  
  fb491a08
- Fixed test to properly be SKIPPED instead of ERROR · 9bab8b59
  由 Olivier Delalleau 提交于 3月 30, 2010
  
  9bab8b59
- escape the char % in Composite op when we use it with the Mod op. · acbd8b54
  由 Frederic Bastien 提交于 3月 30, 2010
  
  acbd8b54
- added c_code_cache_version for Composite(used for fusion of elemwise) · 5c2a6cd7
  由 Frederic Bastien 提交于 3月 29, 2010
  
  5c2a6cd7