提交 · 0056cb74cf8978ad7e9328cc2140e811e34fe7ab · testgroup / pytensor

20 12月, 2014 12 次提交
- crash fix and better error message · 0056cb74
  由 Frederic 提交于 12月 16, 2014
  
  0056cb74
- better error message · f9788298
  由 Frederic 提交于 12月 16, 2014
  
  f9788298
- Fix crash in new cudnn R2 conv code in full mode · b2d72c7c
  由 Frederic 提交于 12月 15, 2014
  
  b2d72c7c
- better error message · 20be5a2e
  由 Frederic 提交于 12月 15, 2014
  
  20be5a2e
- Use the new name in the code. · dca13c09
  由 Frederic 提交于 12月 15, 2014
  
  dca13c09
- make conv cudnn work with R2 · 79804420
  由 Frederic 提交于 12月 15, 2014
  
  79804420
- Make dnn pooling compile again R2 · 0c962794
  由 Frederic 提交于 12月 10, 2014
  
  0c962794
- make cudnn softmax support R2 · 87640d7e
  由 Frederic 提交于 12月 10, 2014
  
  87640d7e
- cache the cudnn version · 487148a8
  由 Frederic 提交于 12月 10, 2014
  
  487148a8
- Use the dynamic dnn version too. · 705d0f28
  由 Frederic 提交于 12月 10, 2014
  
  705d0f28
- Add cuDNN version information. · 78b41288
  由 Frederic 提交于 12月 09, 2014
  
  78b41288
- Merge pull request #2357 from nouiz/cudnn_repair_r1 · 5895d5a2
  由 abergeron 提交于 12月 19, 2014
```
[CRASH] Cudnn repair r1 softmax
```
  5895d5a2
19 12月, 2014 5 次提交
- small code refactoring · 6af9d9a0
  由 Frederic 提交于 12月 19, 2014
  
  6af9d9a0
- [CRASH] cudnn softmax r1 · 5036835a
  由 Frederic 提交于 12月 19, 2014
  
  5036835a
- Merge pull request #2355 from nouiz/cudnn_repair_r1 · e34f6e59
  由 abergeron 提交于 12月 18, 2014
```
[CRASH] Repair crash with cudnn r-1 following the removing of struct_id
```
  e34f6e59
- Repair crash with cudnn r-1 following the removing of struct_id · d1d3451f
  由 Frederic Bastien 提交于 12月 18, 2014
  
  d1d3451f
- Merge pull request #2353 from nouiz/revert_type_constructor · 6a017e04
  由 abergeron 提交于 12月 18, 2014
```
Revert "Get rid of typeConstructor in Scan."
```
  6a017e04
18 12月, 2014 9 次提交
- Revert "Get rid of typeConstructor in Scan." · 743962d4
  由 Frederic 提交于 12月 18, 2014
```
This reverts commit 7d70c3a4.
```
  743962d4
- Merge pull request #2292 from ballasn/memoutput · 51f375bf
  由 carriepl 提交于 12月 18, 2014
```
Output StorageMap content in raise_with_op
```
  51f375bf
- Merge pull request #2352 from lamblin/fix_conv2d_stride_check · c166a9e7
  由 Frédéric Bastien 提交于 12月 18, 2014
```
Use numpy's function to check array is contiguous
```
  c166a9e7
- Merge pull request #2319 from rmcgibbo/expm · 2475957f
  由 Pascal Lamblin 提交于 12月 17, 2014
```
[ENH] Matrix exponential op and gradient (round 2!)
```
  2475957f
- Use numpy's function to check array is contiguous · 6685523c
  由 Pascal Lamblin 提交于 12月 17, 2014
```
In particular, strides for dimensions of length 1 could be anything
```
  6685523c
- Merge pull request #2336 from abergeron/scan_allow_gc · 8a5d41da
  由 Frédéric Bastien 提交于 12月 17, 2014
```
Separate allow_gc flag for scan
```
  8a5d41da
- Merge pull request #2331 from lamblin/fix_set_subtensor1_grad · beb09bb1
  由 Frédéric Bastien 提交于 12月 17, 2014
```
[BUG] Fix gradient of advanced_set_subtensor1
```
  beb09bb1
- Merge pull request #2346 from carriepl/use_ramdisk_in_travis · cb1e2e32
  由 Frédéric Bastien 提交于 12月 17, 2014
```
Use gcc.cxxflags=-pipe for travis tests
```
  cb1e2e32
- correct handling of int/complex dtypes in IncSubtensor.grad · ff9c4bdc
  由 Pascal Lamblin 提交于 12月 17, 2014
  
  ff9c4bdc
17 12月, 2014 10 次提交
- Don't call self.fn.free() unnecessarily. · ba517c71
  由 Arnaud Bergeron 提交于 12月 17, 2014
  
  ba517c71
- Of course you also need to modify the *other* scan. · e4449111
  由 Arnaud Bergeron 提交于 12月 16, 2014
  
  e4449111
- Use gcc.cxxflags=-pipe for travis tests · bf31b9b5
  由 Pierre Luc Carrier 提交于 12月 16, 2014
  
  bf31b9b5
- Forgot grad and R_op for allow_gc. · 31823494
  由 Arnaud Bergeron 提交于 12月 16, 2014
  
  31823494
- Propagate the allow_gc flag in scan optimizations. · d0bb8089
  由 Arnaud Bergeron 提交于 12月 16, 2014
```
I hope I got all the spots. This is insane.
```
  d0bb8089
- Update print output of storage map · 4df20428
  由 Nicolas Ballas 提交于 12月 16, 2014
  
  4df20428
- Add way to specify the allow_gc flags on individual scan instances. · b2bc4f62
  由 Arnaud Bergeron 提交于 12月 13, 2014
  
  b2bc4f62
- Fix the code written while way too tired. · 17d1e8c4
  由 Arnaud Bergeron 提交于 12月 13, 2014
  
  17d1e8c4
- Fix typos in comment. · 613b4e93
  由 Arnaud Bergeron 提交于 12月 12, 2014
  
  613b4e93
- Add a separate allow_gc flag for the Scans. · 5d051ce9
  由 Arnaud Bergeron 提交于 12月 12, 2014
  
  5d051ce9
16 12月, 2014 4 次提交
- Merge pull request #2344 from carriepl/test_too_long · 26a22806
  由 Frédéric Bastien 提交于 12月 16, 2014
```
Add print during unit test to help preventing timeouts
```
  26a22806
- Merge pull request #2342 from abergeron/multi_fixes · c190056b
  由 Frédéric Bastien 提交于 12月 16, 2014
```
Multi fixes (pieces of the multi-gpu code)
```
  c190056b
- Document the clone() method. · 7a5cb506
  由 Arnaud Bergeron 提交于 12月 16, 2014
  
  7a5cb506
- Add print during unit test to help preventing timeouts · 34d90438
  由 Pierre Luc Carrier 提交于 12月 16, 2014
  
  34d90438