提交 · b489422d370f18eb60add07937120187fcebd47f · testgroup / pytensor · GitLab

01 3月, 2010 2 次提交
- Always include ShapeOpt and specialize optimizers in test_local_track_shape_i · b489422d
  由 Pascal Lamblin 提交于 2月 28, 2010
```
Makes test pass in FAST_COMPILE.
```
  b489422d
- Add optimizations to the mode used, to make test pass in FAST_COMPILE · e099538a
  由 Pascal Lamblin 提交于 2月 28, 2010
  
  e099538a
28 2月, 2010 4 次提交
- Include 'fast_run' optimizations to make test pass in FAST_COMPILE · d906e876
  由 Pascal Lamblin 提交于 2月 28, 2010
  
  d906e876
- Use mode=m for functions built in test_hostfromgpu_shape_i. · 4f2a0f22
  由 Pascal Lamblin 提交于 2月 28, 2010
```
This makes the test pass with FAST_COMPILE
```
  4f2a0f22
- Change test, because it was building an incorrect graph. · 5cfa6e2b
  由 Pascal Lamblin 提交于 2月 27, 2010
```
It makes no sense to have a softmax_grad node with inputs that are not
matrices of the same size. C implementation would have complained.
T_SoftmaxGrad.test_infer_shape now passes in FAST_COMPILE.
```
  5cfa6e2b
- added more informative shape error message to gemm · 5df0aeca
  由 James Bergstra 提交于 2月 27, 2010
  
  5df0aeca
27 2月, 2010 19 次提交
- Disabled registration of local_shape_i_lift, which does not seem to be needed anymore. · 05c80ca0
  由 Pascal Lamblin 提交于 2月 27, 2010
  
  05c80ca0
- New test for ShapeOptimizer and local_track_shape_i. · 370d244c
  由 Pascal Lamblin 提交于 2月 27, 2010
  
  370d244c
- Fix bug in local_track_shape_i: we were not scheduling current node when needed. · a961624d
  由 Pascal Lamblin 提交于 2月 27, 2010
  
  a961624d
- Cosmetic changes · a9461657
  由 Pascal Lamblin 提交于 2月 27, 2010
  
  a9461657
- merge · 5281bc31
  由 James Bergstra 提交于 2月 26, 2010
  
  5281bc31
- Shape_i - requires Variable argument to make_node, doesnt convert with as_tensor · 5454b126
  由 James Bergstra 提交于 2月 26, 2010
  
  5454b126
- added test for the bug with CrossentropySoftmax1HotWithBias. · b1c5a558
  由 Frederic Bastien 提交于 2月 26, 2010
  
  b1c5a558
- fix test for HostFromGpu.infer_shape and GpuFromHost.infer_shape. · 4a825957
  由 Frederic Bastien 提交于 2月 26, 2010
  
  4a825957
- give the same default value to gpu op paramter then cpu op. · f89e86ef
  由 Frederic Bastien 提交于 2月 26, 2010
  
  f89e86ef
- added test for some op.infer_shape · 5753e717
  由 Frederic Bastien 提交于 2月 26, 2010
  
  5753e717
- make better test. · 67d88d11
  由 Frederic Bastien 提交于 2月 26, 2010
  
  67d88d11
- added some tests for infer_shape fct. · 48bc80fb
  由 Frederic Bastien 提交于 2月 26, 2010
  
  48bc80fb
- merge · a3b4fd8c
  由 James Bergstra 提交于 2月 26, 2010
  
  a3b4fd8c
- ShapeFeature - better diagnostic information when unpacking infer_shape rval fails · db72e2b8
  由 James Bergstra 提交于 2月 26, 2010
  
  db72e2b8
- NNet - added some long int casts to fprintf calls · c1aeb86c
  由 James Bergstra 提交于 2月 26, 2010
  
  c1aeb86c
- Slightly more detailed error · 8267bcc5
  由 Pascal Lamblin 提交于 2月 26, 2010
  
  8267bcc5
- Cast return value of infer_shapes to int64 in ARange.infer_shape · 788b52f6
  由 Pascal Lamblin 提交于 2月 26, 2010
  
  788b52f6
- crippling setuptools until it is verified and documented in installation docs · 3f6cb8f9
  由 James Bergstra 提交于 2月 26, 2010
  
  3f6cb8f9
- removing debugging statements from ShapeFeature and local_track_shape · c2af113d
  由 James Bergstra 提交于 2月 26, 2010
  
  c2af113d
26 2月, 2010 15 次提交
- some fix to local_track_shape_i. · a41e0dc0
  由 Frederic Bastien 提交于 2月 26, 2010
  
  a41e0dc0
- removed old code not used anymore · e92cdf6d
  由 Frederic Bastien 提交于 2月 26, 2010
  
  e92cdf6d
- make var.shape use the new shape feature to always return a shape object. · bd7d8d2e
  由 Frederic Bastien 提交于 2月 26, 2010
  
  bd7d8d2e
- corrected bug with map · 2fc8a039
  由 rman@rpad 提交于 2月 26, 2010
  
  2fc8a039
- merge · b6b6449a
  由 James Bergstra 提交于 2月 25, 2010
  
  b6b6449a
- ShapeFeature DRAFT of how to update shape_i's when variables are replaced during · 961d94ce
  由 James Bergstra 提交于 2月 25, 2010
```
optimizations.
```
  961d94ce
- workaround for numpy bug in binomial · ba72dc0e
  由 James Bergstra 提交于 2月 25, 2010
  
  ba72dc0e
- Fix a bug recently introduced in local_advanced_indexing_crossentropy_onehot_grad · 4161a309
  由 Pascal Lamblin 提交于 2月 26, 2010
  
  4161a309
- fix MaxAndArgMax to work when the axis is not a constant. · 9a5c5861
  由 Frederic Bastien 提交于 2月 25, 2010
  
  9a5c5861
- implemented MaxAndArgMax.infer_shape and test. I discovered that we can't make a max on a matrix! · 9dac94ad
  由 Frederic Bastien 提交于 2月 25, 2010
  
  9dac94ad
- remove obsoleted optimization local_shape_lift_softmax · 64178ac8
  由 Pascal Lamblin 提交于 2月 25, 2010
  
  64178ac8
- added the cumulatif of % column into the profile mode and made the column a… · 779853d1
  由 Frederic Bastien 提交于 2月 25, 2010
```
added the cumulatif of % column into the profile mode and made the column a little bit more aligned.
```
  779853d1
- Optimize special cases of ARange.infer_shapes, when start=0 or step=1. · 39b45c6a
  由 Pascal Lamblin 提交于 2月 25, 2010
  
  39b45c6a
- Various tests for softmax with a vector (or sum of vectors) as input. · ad7937d2
  由 Pascal Lamblin 提交于 2月 25, 2010
  
  ad7937d2
- Allow optimization to work on xent(softmax(vector)) · de4b6c60
  由 Pascal Lamblin 提交于 2月 25, 2010
  
  de4b6c60