提交 · cd81283eb52607cea1274bcd1d67b33a830606de · testgroup / pytensor

05 11月, 2013 3 次提交
- Use numpy's officially supported api for retrieving BLAS compilation options · cd81283e
  由 huyng 提交于 11月 04, 2013
  
  cd81283e
- Merge pull request #1585 from nouiz/alloc_unbroadcast_grad · 85c9686e
  由 Pascal Lamblin 提交于 11月 04, 2013
```
Fix grad of Alloc when we unbroadcast an input.
```
  85c9686e
- Make python setup.py and pip install Theano, install *.pkl files. · f95390d0
  由 Frederic 提交于 11月 04, 2013
```
This is needed for a test.
```
  f95390d0
04 11月, 2013 1 次提交

Fix grad of Alloc when we unbroadcast an input. · 924e9d03

由提交于 11月 04, 2013

In some cases it was causing an error in the grad related to
broadcasting. In the test it case bad shape.

924e9d03

01 11月, 2013 1 次提交
- Merge pull request #1583 from nouiz/gpuflatten_skip_test · 2d492fd2
  由 Pascal Lamblin 提交于 11月 01, 2013
```
Gpuflatten skip test
```
  2d492fd2
31 10月, 2013 35 次提交
- Fix skipped test. · 2d63ed3b
  由 Frederic 提交于 10月 30, 2013
  
  2d63ed3b
- Transform all Flatten op to the new GpuReshape · fbea8820
  由 Frederic 提交于 10月 30, 2013
  
  fbea8820
- Merge pull request #1582 from abergeron/compyte2-rb · 8bd900f8
  由 Frédéric Bastien 提交于 10月 31, 2013
```
Compyte2 rebase
```
  8bd900f8
- Fix small nits introduced by the merge. · 217b616b
  由 Arnaud Bergeron 提交于 10月 30, 2013
  
  217b616b
- Use the versioning info from gpuarray to avoid unloadable cached modules. · b3f66f86
  由 Arnaud Bergeron 提交于 10月 30, 2013
  
  b3f66f86
- Change the test in op_lifter to be really OP-specific. · 9f6f3579
  由 Arnaud Bergeron 提交于 10月 30, 2013
  
  9f6f3579
- Move Flatten to the new GPU back-end by reusing reshape. · 9f48c753
  由 Frederic 提交于 10月 30, 2013
  
  9f48c753
- Add GpuReshape in the new gpu back-end. · 2284a814
  由 Frederic 提交于 10月 30, 2013
  
  2284a814
- Precompile some variant of Reduction kernel during Theano compilation. · 3aedd9be
  由 Frederic 提交于 10月 30, 2013
  
  3aedd9be
- Fix wrong indices order in GpuSubtensor c_code(). · 1d2a81bb
  由 Arnaud Bergeron 提交于 10月 25, 2013
  
  1d2a81bb
- Add tests for blas and fix an missing import in opt.py. · 77a62e79
  由 Arnaud Bergeron 提交于 10月 23, 2013
  
  77a62e79
- Have different instances of the inputs to support inplace ops. · dbeedc15
  由 Arnaud Bergeron 提交于 10月 23, 2013
  
  dbeedc15
- Use TestOptimizationMixin to provide better error messages when the op is not properly optimized. · eb2f4398
  由 Arnaud Bergeron 提交于 10月 23, 2013
  
  eb2f4398
- Revise makeTester to make it easier to test Gpu ops. · 21581731
  由 Arnaud Bergeron 提交于 10月 23, 2013
  
  21581731
- pep8 fixes · 890b9af0
  由 Arnaud Bergeron 提交于 10月 23, 2013
  
  890b9af0
- Replace the EquilibriumOptimizer by a in2out since it should be faster. · 2b1d70cc
  由 Arnaud Bergeron 提交于 10月 18, 2013
```
Fix mega typo in the Gemm -> GpuGemm optimizer.
```
  2b1d70cc
- Add GpuGemm op. · 5d7a1999
  由 Arnaud Bergeron 提交于 10月 16, 2013
  
  5d7a1999
- Check that the op is one of those that we support before replacing reduce Ops. · 5649f8b6
  由 Arnaud Bergeron 提交于 10月 08, 2013
  
  5649f8b6
- small compilation crash fix. · 894692ea
  由 Frederic 提交于 10月 08, 2013
  
  894692ea
- Fix gemv to accept both float32 and float64 and reenable optimization to insert it in the graph. · 02468f85
  由 Arnaud Bergeron 提交于 10月 07, 2013
  
  02468f85
- Fix tests for GpuCAReduce and a type error in some nan cases. · 52dcfe33
  由 Arnaud Bergeron 提交于 10月 01, 2013
  
  52dcfe33
- Remove usage of new_GpuArray() in favor of the new pygpu_<> API. · 618b2d94
  由 Arnaud Bergeron 提交于 10月 01, 2013
  
  618b2d94
- Add a test for GpuCAReduce · 33fae103
  由 Arnaud Bergeron 提交于 9月 27, 2013
  
  33fae103
- Fix problems with reduction. · 955248cc
  由 Arnaud Bergeron 提交于 9月 27, 2013
  
  955248cc
- Fix the cuda subtensor tests that were broken by my change to the 0_dims test in tensor. · 3dc03868
  由 Arnaud Bergeron 提交于 9月 27, 2013
  
  3dc03868
- Fix test that is searching for an error text. · 50245d71
  由 Arnaud Bergeron 提交于 9月 26, 2013
  
  50245d71
- Fix GpuSubtensor so that it covers all the cases properly. · 07e95861
  由 Arnaud Bergeron 提交于 9月 26, 2013
  
  07e95861
- Add missing return. · 776448dc
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  776448dc
- Try to make sure that the support code is not duplicated due to whitespace differences. · 4061f916
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  4061f916
- Make sure we don't try to optimize nodes that have a subclass of ours when doing fusion. · 0f8d0341
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  0f8d0341
- Use the proper way to init T_subtensor. · e7feb8e3
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  e7feb8e3
- Fix the 0-dim test to use the proper op. · 25ba75cb
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  25ba75cb
- Add a python implementation of GpuGemv. · da2d163e
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  da2d163e
- This should avoid SupportCodeErrors due to mismatching whitespace. · 8c278a05
  由 Arnaud Bergeron 提交于 9月 25, 2013
  
  8c278a05
- Disable the GpuGemv opt since the Op does not work yet. · 9708b02d
  由 Arnaud Bergeron 提交于 9月 23, 2013
  
  9708b02d