提交 · a06016e8a363dd1f282ae48f8577b74ffd546072 · testgroup / pytensor

10 2月, 2014 15 次提交
- Fix import order to be pep8 · a06016e8
  由 Frederic 提交于 2月 07, 2014
  
  a06016e8
- Add a tentative documentation · fefe6206
  由 AlOa 提交于 2月 07, 2014
  
  fefe6206
- Chage op name in test scripts and add env flag · cdb4ab21
  由 AlOa 提交于 2月 01, 2014
  
  cdb4ab21
- Adding scripts to test the speedup(slowdonw) using openmp in elemwise ops · 74a105e8
  由 AlOa 提交于 1月 21, 2014
  
  74a105e8
- Add tests for elemwise op with matrix and vectof of size greater than openmp_elemwise_minsize · 32fe7e5c
  由 AlOa 提交于 1月 13, 2014
  
  32fe7e5c
- Add openmp_elemsize_minzise to documentation · 646287d5
  由 AlOa 提交于 1月 10, 2014
  
  646287d5
- Differents make_loop functions for CAReduce and Elemwise · fe7b8e98
  由 AlOa 提交于 1月 08, 2014
  
  fe7b8e98
- Chage openmp_minsize to openmp_elemwise_minsize · 95af7fce
  由 AlOa 提交于 1月 07, 2014
  
  95af7fce
- Correct variables types · 116b4e93
  由 AlOa 提交于 1月 07, 2014
  
  116b4e93
- Add code for scalar case(No loop) · 7d87c82c
  由 AlOa 提交于 1月 07, 2014
  
  7d87c82c
- Add openmp to elemwise non contiguous case · 375b82ef
  由 AlOa 提交于 1月 07, 2014
  
  375b82ef
- Using directly self.openmp as parameter to add to c code chace · 0a92ffef
  由 AlOa 提交于 1月 07, 2014
  
  0a92ffef
- Modified c code cache so one can havei c code · 4664fefa
  由 AlOa 提交于 1月 05, 2014
```
with and without openmp
chaced
```
  4664fefa
- Add openmp to elemwise contiguous case · 2a2353b3
  由 AlOa 提交于 1月 05, 2014
  
  2a2353b3
- Add new config variable openmp_minsize · 85006a60
  由 AlOa 提交于 1月 03, 2014
  
  85006a60
01 2月, 2014 3 次提交
- Merge pull request #1717 from nouiz/faster_opt · 6253b797
  由 abergeron 提交于 1月 31, 2014
```
Faster opt
```
  6253b797
- Better manual openblas instruction. · a4b14d66
  由 Frederic 提交于 1月 31, 2014
  
  a4b14d66
- Add more error information. · a7dec266
  由 Frederic 提交于 1月 31, 2014
  
  a7dec266
31 1月, 2014 5 次提交
- Merge pull request #1716 from nouiz/err_msg · 9df815dc
  由 abergeron 提交于 1月 31, 2014
```
When an apply node execution raise an error, print the inputs values when they are scalars.
```
  9df815dc
- Generalize the error message to help work around a travis failure. · 51f6b75a
  由 Frederic 提交于 1月 31, 2014
  
  51f6b75a
- Fix test since now we introduce less neg node. · 9fa8a6ff
  由 Frederic 提交于 1月 30, 2014
  
  9fa8a6ff
- client of variable can be the string 'output' · 6fd91c68
  由 Frederic 提交于 1月 30, 2014
  
  6fd91c68
- Merge pull request #1710 from nouiz/gpu_reduce · 06cdb1fa
  由 abergeron 提交于 1月 30, 2014
```
new Gpu reduce based on the version in Theano that is much faster.
```
  06cdb1fa
30 1月, 2014 13 次提交
- Rename GpuCAReduce to GpuCAReduceCuda and reuse the name GpuCAReduce for… · 1121fe45
  由 Frederic 提交于 1月 30, 2014
```
Rename GpuCAReduce to GpuCAReduceCuda and reuse the name GpuCAReduce for GpuCAReduceCPY to allow reloading older pickled files.
```
  1121fe45
- When an apply node execution raise an error, print the inputs values when they are scalar. · 57fed426
  由 Frederic 提交于 1月 30, 2014
  
  57fed426
- Port typo fix from gpuarray files by Olivier D. to the old back-end. · a2afb9dd
  由 Frederic 提交于 1月 30, 2014
  
  a2afb9dd
- Fix opt crash with other dtype then float32, the code isn't ready. · 111400e4
  由 Frederic 提交于 1月 29, 2014
  
  111400e4
- Merge pull request #1700 from MarcCote/cumsum · 262f59a2
  由 Frédéric Bastien 提交于 1月 29, 2014
```
Added the cumsum and cumprod functions similar to numpy's ones.
```
  262f59a2
- remove GpuShape and associated opt as now we reuse the Shape Op with specific c code. · 047c8913
  由 Frederic 提交于 1月 29, 2014
  
  047c8913
- fix/update comment/docstring. · dce1d289
  由 Frederic 提交于 1月 29, 2014
  
  dce1d289
- remove commented code. · 1f79dbf1
  由 Frederic 提交于 1月 29, 2014
  
  1f79dbf1
- local_mul_specialize now don't introcude neg node when this will add more node then needed. · 9ec45737
  由 Frederic 提交于 1月 29, 2014
  
  9ec45737
- remove local_mul_to_neg as it doing a subset of local_mul_specialize · ea3dcffc
  由 Frederic 提交于 1月 29, 2014
  
  ea3dcffc
- Remove check_preallocated_output and use default mode for theano.function. · 5c504099
  由 Marc-Alexandre Cote 提交于 1月 29, 2014
  
  5c504099
- local_mul_to_neg now only apply when it won't add node to the graph · 1cf4a13c
  由 Frederic 提交于 1月 29, 2014
  
  1cf4a13c
- Pass the reason parameter down. · cc47a946
  由 Frederic 提交于 1月 29, 2014
  
  cc47a946
29 1月, 2014 4 次提交
- Remove useless fct call · 8b62a1a2
  由 Frederic 提交于 1月 29, 2014
  
  8b62a1a2
- Fix crashes introduced in this PR. · b68d4a87
  由 Frederic 提交于 1月 29, 2014
  
  b68d4a87
- Faster opt by not doing useless stuff. · e8aff982
  由 Frederic 提交于 1月 29, 2014
```
This make the Canonizer take 4x less time in a test case.
```
  e8aff982
- Print more info in the optimizer profiling · 2b7b0305
  由 Frederic 提交于 1月 29, 2014
  
  2b7b0305