提交 · 0bd576e11dfe7f18437b43934db6e4f22a0bc4e2 · testgroup / pytensor

25 5月, 2010 3 次提交
- New tests, and reference results generated from the original java code · 0bd576e1
  由 Pascal Lamblin 提交于 5月 25, 2010
  
  0bd576e1
- New keyword to explicitly choose a number of streams in uniform() · 0e861a54
  由 Pascal Lamblin 提交于 5月 25, 2010
  
  0e861a54
- wrong argument name · 008ce348
  由 Razvan Pascanu 提交于 5月 24, 2010
  
  008ce348
24 5月, 2010 3 次提交
- yet another n_steps =1 bug fixed · aaf14063
  由 Razvan Pascanu 提交于 5月 24, 2010
  
  aaf14063
- added some more tests of log_add-related functionality... they do not pass but soon they should · 0198fa75
  由 James Bergstra 提交于 5月 23, 2010
  
  0198fa75
- added log_add optimization · 28e5b5cd
  由 James Bergstra 提交于 5月 23, 2010
  
  28e5b5cd
23 5月, 2010 1 次提交
- Do not fail if numpy generates 0 or 1 · 9677a10f
  由 Pascal Lamblin 提交于 5月 22, 2010
  
  9677a10f
22 5月, 2010 1 次提交
- remove extra outdated comments + fixed bug in case return_steps =1, and n_steps =1 or -1 · 1b2c02ae
  由 Razvan Pascanu 提交于 5月 21, 2010
  
  1b2c02ae
21 5月, 2010 2 次提交
- remove debug prints · 5d9d2303
  由 Razvan Pascanu 提交于 5月 21, 2010
  
  5d9d2303
- scan now makes copies of the other arguments · 3c268cc9
  由 Razvan Pascanu 提交于 5月 20, 2010
  
  3c268cc9
20 5月, 2010 6 次提交
- fix MRG normal distribution · 4f88815c
  由 Frederic Bastien 提交于 5月 20, 2010
  
  4f88815c
- don't let the mrg random generator for normal distribution upcast float32 to float64 · fefdb397
  由 Frederic Bastien 提交于 5月 19, 2010
  
  fefdb397
- generate subtensor only if their is a stride. · d56de3e6
  由 Frederic Bastien 提交于 5月 19, 2010
  
  d56de3e6
- disable some test case that make the tests fail due to rounding error. · bcb89e9b
  由 Frederic Bastien 提交于 5月 19, 2010
  
  bcb89e9b
- get the nvcc version · 1a9022e8
  由 Frederic Bastien 提交于 5月 19, 2010
  
  1a9022e8
- added comment · d12dafcc
  由 Frederic Bastien 提交于 5月 19, 2010
  
  d12dafcc
19 5月, 2010 1 次提交
- Remove experiemental conv_2d_offset code. · 0a62c9df
  由 Pierre-Antoine Manzagol 提交于 5月 19, 2010
  
  0a62c9df
18 5月, 2010 6 次提交
- show the float32 only limitation when we advertise for theano. · b3b502cd
  由 Frederic Bastien 提交于 5月 18, 2010
  
  b3b502cd
- make CudaNdarray_inplace_add and CudaNdarray_add work with tensor of 0 elements and test them. · e4e88af1
  由 Frederic Bastien 提交于 5月 17, 2010
  
  e4e88af1
- make GpuElemwise work with empty tensor. Add tests for this. · 6664a048
  由 Frederic Bastien 提交于 5月 17, 2010
  
  6664a048
- fix the display of error in some corner case in debug_mode · 6922a69c
  由 Frederic Bastien 提交于 5月 17, 2010
  
  6922a69c
- fix the last commit in DEBUG_MODE. · 33bd7278
  由 Frederic Bastien 提交于 5月 17, 2010
```
Make the copy to the cpu from the gpu work for empty cuda array.
Make the perform of Sum work in python when their is more then 1 dimenssions
```
  33bd7278
- make GpuSum work correctly with tensor of 0 element. · de9cde80
  由 Frederic Bastien 提交于 5月 17, 2010
  
  de9cde80
17 5月, 2010 3 次提交
- added test for nvcc 3.0 and 3.1beta bug. · bbbe245d
  由 Frederic Bastien 提交于 5月 17, 2010
  
  bbbe245d
- fix for corner case for n_steps = 1 in some instances · 7e32b612
  由 Razvan Pascanu 提交于 5月 16, 2010
  
  7e32b612
- fix inputs bug in scan · 5df42b9c
  由 Razvan Pascanu 提交于 5月 16, 2010
  
  5df42b9c
15 5月, 2010 1 次提交
- fix a bug in nvcc 3.0 · 246b7287
  由 Frederic Bastien 提交于 5月 14, 2010
  
  246b7287
14 5月, 2010 4 次提交
- in fprint_CudaNdarray, print the value on the device · 6242157d
  由 Frederic Bastien 提交于 5月 14, 2010
  
  6242157d
- warn if we call CudaNdarray_dealloc but their is too many ref count to it. · 3767c435
  由 Frederic Bastien 提交于 5月 14, 2010
  
  3767c435
- removed duplicate var that is not used. · 714ca7d7
  由 Frederic Bastien 提交于 5月 14, 2010
  
  714ca7d7
- better error msg. · 1dcc4047
  由 Frederic Bastien 提交于 5月 13, 2010
  
  1dcc4047
12 5月, 2010 5 次提交
- add missing import · cb42e890
  由 Frederic Bastien 提交于 5月 12, 2010
  
  cb42e890
- added GpuSum for pattern 0011 · 215a3ae6
  由 Frederic Bastien 提交于 5月 11, 2010
  
  215a3ae6
- bump numpy version requirement as we found a memory leark. · 63329761
  由 Frederic Bastien 提交于 5月 11, 2010
  
  63329761
- make tests test_rng_mrg.py:test_rng0 work in FAST_COMPILE and DEBUG_MODE and enabled those mode. · adca49c9
  由 Frederic Bastien 提交于 5月 11, 2010
  
  adca49c9
- give a warning if we use a numpy version that have a memory leak and that we… · dbc3e752
  由 Frederic Bastien 提交于 5月 11, 2010
```
give a warning if we use a numpy version that have a memory leak and that we exeecute code that will hit that leak.
```
  dbc3e752
11 5月, 2010 4 次提交
- put the constant folding optimization into fast_compile · 14561051
  由 Frederic Bastien 提交于 5月 11, 2010
  
  14561051
- fix buildbot. · 5f0a1536
  由 Frederic Bastien 提交于 5月 10, 2010
  
  5f0a1536
- fix syntax error. · 8b5986f9
  由 Frederic Bastien 提交于 5月 10, 2010
  
  8b5986f9
- moved test of rng_mrg to a separate file and fix one of the tests. · f41011a9
  由 Frederic Bastien 提交于 5月 10, 2010
  
  f41011a9