提交 · 67e14ae056d09e63a24b77dd4b247fc9f041d369 · testgroup / pytensor

11 5月, 2016 34 次提交
- Pre apply opt. · 67e14ae0
  由 Frederic Bastien 提交于 5月 11, 2016
  
  67e14ae0
- inplace elemwise optimizer speed up. · 644d8bce
  由 Frederic Bastien 提交于 5月 11, 2016
  
  644d8bce
- Add optimizer profiling information. · fce0b82f
  由 Frederic Bastien 提交于 5月 10, 2016
  
  fce0b82f
- Create less temporary node. · c657bad8
  由 Frederic Bastien 提交于 5月 10, 2016
  
  c657bad8
- Try to make less temp node when moving to the GPU. · 12e58299
  由 Frederic Bastien 提交于 5月 10, 2016
  
  12e58299
- Merge pull request #4463 from nouiz/scan_reintroduced_benchmark · 8c58dfb8
  由 Frédéric Bastien 提交于 5月 10, 2016
```
Scan reintroduced benchmark
```
  8c58dfb8
- Merge pull request #4488 from abergeron/disable_preallocate · f512a560
  由 Frédéric Bastien 提交于 5月 10, 2016
```
Make negative values for gpuarray.preallocate disable the allocation cache completely.
```
  f512a560
- Fix typo in doc. · 0944780b
  由 Arnaud Bergeron 提交于 5月 10, 2016
  
  0944780b
- 1 means preallocate as much as possible. · d6119644
  由 Arnaud Bergeron 提交于 5月 10, 2016
  
  d6119644
- pep8 and remove a newlly added fast_compile tag. · a312daf1
  由 Frederic Bastien 提交于 5月 10, 2016
  
  a312daf1
- Skip a test in FAST_COMPILE and small refactor · 447ab781
  由 Frederic Bastien 提交于 5月 09, 2016
  
  447ab781
- Fix test by implementing the case with MakeVector · 0f82ee2d
  由 Frederic Bastien 提交于 5月 08, 2016
  
  0f82ee2d
- small typo fix · 7fc086d9
  由 Frederic Bastien 提交于 5月 06, 2016
  
  7fc086d9
- tmp · c227d7c0
  由 Frederic Bastien 提交于 5月 05, 2016
  
  c227d7c0
- Add code to print the cudnn algo used. · 94abb92a
  由 Frederic Bastien 提交于 5月 03, 2016
  
  94abb92a
- Allow to investigate more · 920194a7
  由 Frederic Bastien 提交于 5月 03, 2016
  
  920194a7
- Fix crash recently introduced · ac1660e5
  由 Frederic Bastien 提交于 5月 03, 2016
  
  ac1660e5
- Add Theano flag scan.debug=False. · aa2a430b
  由 Frederic Bastien 提交于 5月 03, 2016
  
  aa2a430b
- Add opt Elemwise[EQ](Subtensor(Shape(x)), -N) -> Elemwise[zeros](X) that work… · 70e866cd
  由 Frederic Bastien 提交于 5月 03, 2016
```
Add opt Elemwise[EQ](Subtensor(Shape(x)), -N) -> Elemwise[zeros](X) that work around this comment by removing the re-inserted scan: https://github.com/Theano/Theano/pull/4356/files#r61901454
```
  70e866cd
- Code clean up · cc28e44d
  由 Frederic Bastien 提交于 5月 03, 2016
  
  cc28e44d
- Small code simplication · add24191
  由 Frederic Bastien 提交于 5月 03, 2016
  
  add24191
- Better error message · 4b23242e
  由 Frederic Bastien 提交于 4月 22, 2016
  
  4b23242e
- Keep the reason · 28a93a27
  由 Frederic Bastien 提交于 4月 22, 2016
  
  28a93a27
- when using traceback.compile_limit, keep internal stack trace. · 339b6aba
  由 Frederic Bastien 提交于 5月 03, 2016
  
  339b6aba
- Make TensorConstant.equals() work with python int and float · ac97e23a
  由 Frederic Bastien 提交于 5月 03, 2016
  
  ac97e23a
- Code refactoring to create variable only once. · 23ccd2ac
  由 Frederic Bastien 提交于 5月 03, 2016
  
  23ccd2ac
- Better error mesg · b0115248
  由 Frederic Bastien 提交于 5月 03, 2016
  
  b0115248
- Add Theano flag traceback.compile_limit · 840352dd
  由 Frederic Bastien 提交于 5月 03, 2016
  
  840352dd
- Add documentation for gpuarray.preallocate. · a8719c76
  由 Arnaud Bergeron 提交于 5月 10, 2016
  
  a8719c76
- Bump the version of GpuArrayType. · f925a74e
  由 Arnaud Bergeron 提交于 5月 10, 2016
```
This will force a recompile of everything and get rid of old modules
that refer to the previous locations of the Ops, which causes a
warning when they are reloaded.
```
  f925a74e
- Change the way the cudnn version is printed and also show how much memory is preallocated. · 605da818
  由 Arnaud Bergeron 提交于 5月 10, 2016
  
  605da818
- Make negative values for gpuarray.preallocate disable the allocation · 57f5c7c7
  由 Arnaud Bergeron 提交于 5月 09, 2016
```
cache completely.
```
  57f5c7c7
- Merge pull request #4454 from slefrancois/gpu_out_sandbox · 69338f33
  由 abergeron 提交于 5月 10, 2016
```
Move new GPU backend out of sandbox
```
  69338f33
- Merge pull request #4492 from nouiz/fix_crash_gpufft · ed6c966d
  由 Frédéric Bastien 提交于 5月 10, 2016
```
[CRASH,REGRESSION] Fix crash gpufft
```
  ed6c966d
10 5月, 2016 6 次提交
- Shorter note · aef92b9c
  由 Frederic Bastien 提交于 5月 10, 2016
  
  aef92b9c
- Add a note that NanGuardMode isn't super fast · b9afaaa6
  由 Frederic Bastien 提交于 5月 10, 2016
  
  b9afaaa6
- Make an example ignore floatX. We didn'T talk about it yet! · 319e50df
  由 Frederic Bastien 提交于 5月 10, 2016
  
  319e50df
- Add sandbox/cuda/init to flake8 whitelist · 319382b5
  由 slefrancois 提交于 5月 10, 2016
  
  319382b5
- Fix gpufft regression that appear when doing flake8 fix. · 7fbbfd14
  由 Frederic Bastien 提交于 5月 10, 2016
  
  7fbbfd14
- Remove sandbox/cuda/test from flake8 whitelist · 2da7c932
  由 slefrancois 提交于 5月 10, 2016
  
  2da7c932