提交 · 118d9fb283d35bd44eaae5a64cd880ccc33561b4 · testgroup / pytensor · GitLab

06 5月, 2015 40 次提交
- Fix typo in scalar/basic.py. · 118d9fb2
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  118d9fb2
- Convert heretics. · 196303a5
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  196303a5
- Bump GpuElemwise version to use the new code. · 54b9af82
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  54b9af82
- Fix softmax opt to not restrict itself to {float32, float64}. · ea77e606
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  ea77e606
- Remove extra spaces · b8bd441f
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  b8bd441f
- Make GpuCrossentropyWithSoftmax ... Dx work with float16. · 7b18775f
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  7b18775f
- Stop hiding error messages with errors about constant copying. · 1c73c350
  由 Arnaud Bergeron 提交于 4月 29, 2015
  
  1c73c350
- Add the actual fp16 code for GpuCAReduceCuda. · eb10bddb
  由 Arnaud Bergeron 提交于 4月 30, 2015
  
  eb10bddb
- Document the '_f16_ok' attribute. · 95b42d18
  由 Arnaud Bergeron 提交于 4月 29, 2015
  
  95b42d18
- Fix broken condition checking for float16 in Composite. · 67812dbc
  由 Arnaud Bergeron 提交于 4月 29, 2015
  
  67812dbc
- Update comment about scalar.upcast() · 1c0cf112
  由 Arnaud Bergeron 提交于 4月 29, 2015
  
  1c0cf112
- Flake8 for gpuarray/opt.py · 5c9ae8ae
  由 Arnaud Bergeron 提交于 4月 29, 2015
  
  5c9ae8ae
- Add a check at C code generation time for ops that are not explicitely · 8d0b398d
  由 Arnaud Bergeron 提交于 4月 29, 2015
```
tagged for float16 support and disable their C code.
```
  8d0b398d
- Fix the broken test for float16 inputs/outputs in Elemwise and bump the Elemwise… · 708efc22
  由 Arnaud Bergeron 提交于 4月 23, 2015
```
Fix the broken test for float16 inputs/outputs in Elemwise and bump the Elemwise version to make no bad code is used.
```
  708efc22
- Bump the C version for the modified Crossentropy op. · fec62ac0
  由 Arnaud Bergeron 提交于 4月 23, 2015
  
  fec62ac0
- Don't duplicate the kernel code in GpuEye for float16. · c3f03aab
  由 Arnaud Bergeron 提交于 4月 23, 2015
  
  c3f03aab
- Don't autocast to float16 unless config.floatX == 'float16' · ed222a6d
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  ed222a6d
- Make test_binomial yield for sub-tests to ease identifying the very long test. · 1a4a1c96
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  1a4a1c96
- Remove unused import for flake8 · 73072ca5
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  73072ca5
- Add float16 to the complex corner case. · bc2d1a17
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  bc2d1a17
- Fix logic of the float16 check. · 0ae495e1
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  0ae495e1
- Remove hack in gpuarray/ to replace Cast by Cast16 · dd98172a
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  dd98172a
- Disable C code for all Elemwise that interact with float16 · d43f6df9
  由 Arnaud Bergeron 提交于 4月 22, 2015
  
  d43f6df9
- Fix last failing test · 9111dd3d
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  9111dd3d
- Make mean return the right type and keep the accumulator at least float32 if possible. · 45e52dd0
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  45e52dd0
- The bug mentioned was fixed in numpy 1.6 and we support 1.6.2 at minimum. · 9c9a5a51
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  9c9a5a51
- Spread the good word and add a connection_pattern. · b73adff1
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  b73adff1
- Fix a bunch of casting policy problems. · b20cca75
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  b20cca75
- Get rid of ghetto 2.4 code. · 0fea0093
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  0fea0093
- Update casting under numpy+floatX since that is tested. · 9a8eda51
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  9a8eda51
- Fix flake8 format. · ff085441
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  ff085441
- Skip tests in float16 for sparse. · 55b85779
  由 Arnaud Bergeron 提交于 4月 21, 2015
  
  55b85779
- Update floatX doc. · 79c49197
  由 Arnaud Bergeron 提交于 4月 20, 2015
  
  79c49197
- Add the float16 dtype and make sure it isn't lost to casting. · 234ffeab
  由 Arnaud Bergeron 提交于 4月 20, 2015
  
  234ffeab
- Make float16 a recognized type (no casting as it would break on CPU). · fd7655aa
  由 Arnaud Bergeron 提交于 4月 20, 2015
  
  fd7655aa
- Add support for float16 to one of the versions of GpuCrossentropy... · abd0a0fc
  由 Arnaud Bergeron 提交于 4月 16, 2015
  
  abd0a0fc
- Fix Elemwise for float16 inputs/outputs on GPU · cc7c365c
  由 Arnaud Bergeron 提交于 4月 16, 2015
  
  cc7c365c
- Make local_dot_to_dot22 stop failing for float16. · d3bba908
  由 Arnaud Bergeron 提交于 4月 16, 2015
  
  d3bba908
- Add float16 support for GpuEye. · 2c1d697c
  由 Arnaud Bergeron 提交于 4月 16, 2015
  
  2c1d697c
- Allow floatX=float16 · e6dc5e8d
  由 Arnaud Bergeron 提交于 4月 10, 2015
  
  e6dc5e8d