提交 · da85a8d6f07f91948bddbd9cd100cfdbe7e95a9c · testgroup / pytensor

19 6月, 2010 3 次提交
- added a new optimization for tensor of pow. Disabled by default as this cause… · da85a8d6
  由 Frederic Bastien 提交于 6月 18, 2010
```
added a new optimization for tensor of pow. Disabled by default as this cause slow down on gpu in the current form. To enable it use the flag experimental.pow.
```
  da85a8d6
- added benchmark with numexpr. · 8b892674
  由 Frederic Bastien 提交于 6月 18, 2010
  
  8b892674
- modified dimensions in test_neighbours.py · f38b603c
  由 Simon Lemieux 提交于 6月 18, 2010
  
  f38b603c
17 6月, 2010 15 次提交
- Update test for softmax_opt gradient · 3662b679
  由 Pascal Lamblin 提交于 6月 16, 2010
  
  3662b679
- Comment out debug print · 6958dc9a
  由 Pascal Lamblin 提交于 6月 16, 2010
  
  6958dc9a
- New optimization local_softmax_grad · 7fe1ba4d
  由 Pascal Lamblin 提交于 6月 16, 2010
```
dy*sm - DimShuffle{0,'x'}(sum{1}(dy*sm))*sm -> softmax_grad(dy,sm)
```
  7fe1ba4d
- Update first test of softmax_opt · ed354d0e
  由 Pascal Lamblin 提交于 6月 16, 2010
  
  ed354d0e
- New optimization: sum(a / dimshuffle{...}(b), axis=l) -> sum(a, axis=l) / b · 127f1000
  由 Pascal Lamblin 提交于 6月 16, 2010
  
  127f1000
- Small changes in the test for softmax optimization · bae2c6df
  由 Pascal Lamblin 提交于 6月 16, 2010
  
  bae2c6df
- tensor.Canonize - adding external simplifiers to permit adding optimizations to… · dc02ed59
  由 James Bergstra 提交于 6月 16, 2010
```
tensor.Canonize - adding external simplifiers to permit adding optimizations to the local_canonicalizers / canonizers
```
  dc02ed59
- Adding commented-out softmax optimizations. They are under construction. · de37ce49
  由 James Bergstra 提交于 6月 16, 2010
  
  de37ce49
- merge · 0257b2b6
  由 James Bergstra 提交于 6月 16, 2010
  
  0257b2b6
- adding some new tests for the softmax optimization · 8c8e27d6
  由 James Bergstra 提交于 6月 16, 2010
  
  8c8e27d6
- merge · e1c41c79
  由 Simon Lemieux 提交于 6月 16, 2010
  
  e1c41c79
- modified neighbours op so that its inputs doesn't need to be contiguous · 3b53e332
  由 Simon Lemieux 提交于 6月 16, 2010
  
  3b53e332
- added eq and hash for tensor.Dot · 0265894e
  由 James Bergstra 提交于 6月 16, 2010
  
  0265894e
- merge · fefdf2b1
  由 Simon Lemieux 提交于 6月 16, 2010
  
  fefdf2b1
- neighbours seems to work on gpu · 8da1f256
  由 Simon Lemieux 提交于 6月 16, 2010
  
  8da1f256
15 6月, 2010 3 次提交
- remove not used stuff. · 81ae8ae9
  由 Frederic Bastien 提交于 6月 14, 2010
  
  81ae8ae9
- small change. · 0d268386
  由 Frederic Bastien 提交于 6月 14, 2010
  
  0d268386
- adding gpu code for neighbours.py · 21d76325
  由 Simon Lemieux 提交于 6月 14, 2010
  
  21d76325
14 6月, 2010 2 次提交
- make code versionable. · 946a8fa5
  由 Frederic Bastien 提交于 6月 14, 2010
  
  946a8fa5
- make test more complete, but the new test fail! · 790462b6
  由 Frederic Bastien 提交于 6月 14, 2010
  
  790462b6
12 6月, 2010 3 次提交
- merge · 4ba6a10a
  由 Razvan Pascanu 提交于 6月 11, 2010
  
  4ba6a10a
- aliasing bug reported by Ian fixed .. · 655ecd92
  由 Razvan Pascanu 提交于 6月 11, 2010
  
  655ecd92
- adding simplified neighbours op for cpu · dfc04612
  由 Simon Lemieux 提交于 6月 11, 2010
  
  dfc04612
11 6月, 2010 14 次提交
- made the gpu optimize phase earlier to have better performance in some case. · 56f5d44a
  由 Frederic Bastien 提交于 6月 11, 2010
  
  56f5d44a
- make Reshape.infer_shape work in more case. · 6b2b5132
  由 Frederic Bastien 提交于 6月 11, 2010
  
  6b2b5132
- make the optimizer local_inplace_setsubtensor work for GpuIncSubtensor too. · 2da94091
  由 Frederic Bastien 提交于 6月 11, 2010
  
  2da94091
- make the GpuAlloc work in more case. · b8a75301
  由 Frederic Bastien 提交于 6月 11, 2010
  
  b8a75301
- small clean up that make CudaNdarray_setitem it a little faster. · 5cc01161
  由 Frederic Bastien 提交于 6月 11, 2010
  
  5cc01161
- Add sanity check on seed in rng_mrg, to ensure the results' validity · 12fa8f65
  由 Pascal Lamblin 提交于 6月 10, 2010
  
  12fa8f65
- Remove one of the last references to weave · f3fedb46
  由 Pascal Lamblin 提交于 6月 10, 2010
  
  f3fedb46
- Added free(newdims) when exiting before the end of the function. · 70965d0d
  由 Pascal Lamblin 提交于 6月 10, 2010
  
  70965d0d
- add __str__ to Join and GpuJoin · 6b25596e
  由 Frederic Bastien 提交于 6月 10, 2010
  
  6b25596e
- merge · 303885e5
  由 James Bergstra 提交于 6月 10, 2010
  
  303885e5
- cmodule - reversed order of includes to make a nonstandard numpy installation work as well · e35a57c3
  由 James Bergstra 提交于 6月 10, 2010
  
  e35a57c3
- optimize Alloc to GpuAlloc more often. · c6d5ee22
  由 Frederic Bastien 提交于 6月 10, 2010
  
  c6d5ee22
- added GpuAlloc and opt to put them into the graph and test. · 3c51a2f1
  由 Frederic Bastien 提交于 6月 10, 2010
  
  3c51a2f1
- pfunc - fix to helper that guessed at which node is responsible for a wrong dtype in an update · a163846d
  由 James Bergstra 提交于 6月 10, 2010
  
  a163846d