提交 · 2ed29b2ba1e260a14a73bf31ef9ced7e83d4e2c3 · testgroup / pytensor

04 9月, 2015 1 次提交
- GpuReshape opt. · 2ed29b2b
  由 Cesar Laurent 提交于 9月 03, 2015
  
  2ed29b2b
03 9月, 2015 1 次提交
- Gpu reshape opt. · b8008818
  由 Cesar Laurent 提交于 9月 02, 2015
  
  b8008818
01 9月, 2015 2 次提交
- Merge pull request #3344 from nouiz/cycle · 87f8c5a1
  由 abergeron 提交于 8月 31, 2015
```
Deactivate merge of assert as it cause cycle in the graph
```
  87f8c5a1
- Merge pull request #3290 from mohammadpz/prod_dimshuffle_opt · ca8d85de
  由 carriepl 提交于 8月 31, 2015
```
Prod dimshuffle opt
```
  ca8d85de
31 8月, 2015 7 次提交
- Deactivate merge of assert as it cause cycle in the graph · d537addb
  由 Frederic 提交于 8月 31, 2015
  
  d537addb
- FusionOptimizer added for solve FAST_COMPILE issue · 85a7535d
  由 Mohammad Pezeshki 提交于 8月 31, 2015
  
  85a7535d
- more tests + comments · 59b8455c
  由 Mohammad Pezeshki 提交于 8月 27, 2015
  
  59b8455c
- Logical tests added · 0686893f
  由 Mohammad Pezeshki 提交于 8月 27, 2015
  
  0686893f
- compatible dimes are now fixed · 303cc80d
  由 Mohammad Pezeshki 提交于 8月 27, 2015
  
  303cc80d
- numerical tests added · ae53db81
  由 Mohammad Pezeshki 提交于 8月 17, 2015
  
  ae53db81
- optimization for prod added · 6aa9d88e
  由 Mohammad Pezeshki 提交于 8月 17, 2015
  
  6aa9d88e
29 8月, 2015 3 次提交
- Merge pull request #3267 from koningrobot/tensordot-as-dot · 856aa0b6
  由 Frédéric Bastien 提交于 8月 29, 2015
```
Implement batched_tensordot in terms of batched_dot
```
  856aa0b6
- Merge pull request #3334 from abergeron/delete_old_crap · 77f6b2be
  由 abergeron 提交于 8月 28, 2015
```
Delete old stuff
```
  77f6b2be
- Merge pull request #3288 from abergeron/nouiz_mixed · 7320e1b1
  由 abergeron 提交于 8月 28, 2015
```
Nouiz mixed
```
  7320e1b1
28 8月, 2015 26 次提交
- Fix some typos and phrasings. · 7f43e9f4
  由 Arnaud Bergeron 提交于 8月 28, 2015
  
  7f43e9f4
- Remove remnants of theano modules that were deleted in 0.7. · bcfe70c7
  由 Arnaud Bergeron 提交于 8月 28, 2015
  
  bcfe70c7
- Delete the old unmaintained copy of scan in sandbox. · 43d58c19
  由 Arnaud Bergeron 提交于 8月 28, 2015
  
  43d58c19
- Flake8 fix. · c79e0cfd
  由 Arnaud Bergeron 提交于 8月 17, 2015
  
  c79e0cfd
- hash tuple instead of xor them · 0b5ee2f1
  由 Frederic 提交于 8月 11, 2015
  
  0b5ee2f1
- Remove deprecated comment · 1bc1c0fa
  由 Frederic 提交于 8月 11, 2015
  
  1bc1c0fa
- If a reduce upcast the input, don't move it to the GPU. · 233e7813
  由 Frederic 提交于 8月 11, 2015
  
  233e7813
- Add config var NanGuardMode.{nan,inf,big}_is_error · 0b5aa217
  由 Frederic 提交于 8月 11, 2015
  
  0b5aa217
- Make NanGuardMode not raise an error with theano.sandbox.rng_mrg.GPU_mrg_uniform · 6c4738d6
  由 Frederic 提交于 8月 10, 2015
  
  6c4738d6
- Make flag mode=NanGuardMode work and make it user provided optimizer · fdbe417b
  由 Frederic 提交于 8月 10, 2015
  
  fdbe417b
- Update following code review · 5922a930
  由 Frederic 提交于 8月 10, 2015
  
  5922a930
- Doc about cudnn v3 · 5f3b4fab
  由 Frederic 提交于 8月 08, 2015
  
  5f3b4fab
- Don't make opt return errors with bad user graph. The run time error will be better. · b5aeafc9
  由 Frederic 提交于 8月 08, 2015
  
  b5aeafc9
- Remove useless gpu opt warning when dtype isn't float32 · b08d5cb7
  由 Frederic 提交于 8月 08, 2015
  
  b08d5cb7
- If the gpu isn't working don't raise an error uselessly · ce0a3f02
  由 Frederic 提交于 8月 08, 2015
  
  ce0a3f02
- Add assert to see better problems · d6b799f8
  由 Frederic 提交于 8月 08, 2015
  
  d6b799f8
- pep8 · ccfeeb2a
  由 Frederic 提交于 8月 08, 2015
  
  ccfeeb2a
- Don't try to move to the new back-end elemwise with multiple output(not yet supported) · 9790a08e
  由 Frederic 提交于 8月 08, 2015
  
  9790a08e
- remove print · 944fd8e8
  由 Frederic 提交于 8月 08, 2015
  
  944fd8e8
- Make inplace elemwise opt support multiple output · e4abbe90
  由 Frederic 提交于 8月 08, 2015
  
  e4abbe90
- Fix dtype for Elemwise.perform with multiple output · 1a6e03c6
  由 Frederic 提交于 8月 08, 2015
  
  1a6e03c6
- Make Composite raise an error for case not supported · 350edca0
  由 Frederic 提交于 8月 08, 2015
  
  350edca0
- Make GpuElemwise work with multiple output, (new back-end raise an error) · 41a8e89b
  由 Frederic 提交于 8月 08, 2015
  
  41a8e89b
- Better test · 2595fea0
  由 Frederic 提交于 8月 07, 2015
  
  2595fea0
- Allow transfer type to hardcode the output dtype · f9a2e45b
  由 Frederic 提交于 8月 07, 2015
  
  f9a2e45b
- Make opt not crash with multi output CPU elemwise · a579c1eb
  由 Frederic 提交于 8月 06, 2015
  
  a579c1eb