提交 · 8cdf8e6278168ed8a755fcd6c74006a83ebfa048 · testgroup / pytensor

23 1月, 2016 2 次提交
- BatchedDot: satisfy flake8's perverse desires · 8cdf8e62
  由 Tim Cooijmans 提交于 1月 22, 2016
  
  8cdf8e62
- BatchedDot: avoid repeated indexing · d8538e4a
  由 Tim Cooijmans 提交于 1月 22, 2016
  
  d8538e4a
13 1月, 2016 4 次提交
- BatchedDot: leave first dimension broadcastable if it was broadcastable for either of the inputs · 8f65b441
  由 Tim Cooijmans 提交于 1月 12, 2016
  
  8f65b441
- BatchedDot: remove reference to Dot · e527643a
  由 Tim Cooijmans 提交于 1月 12, 2016
  
  e527643a
- BatchedDot: remove complex test cases as we will not support complex types for now · 878633ed
  由 Tim Cooijmans 提交于 1月 12, 2016
  
  878633ed
- BatchedDot: upcast inputs if they are of mixed types · 25ca839d
  由 Tim Cooijmans 提交于 1月 12, 2016
  
  25ca839d
12 1月, 2016 5 次提交
- BatchedDot: small improvements · 8cd678da
  由 Tim Cooijmans 提交于 1月 11, 2016
  
  8cd678da
- BatchedDot: fix more bad pointer arithmetic · 0b89695c
  由 Tim Cooijmans 提交于 1月 11, 2016
  
  0b89695c
- BatchedDot: reduce duplication · 78bbb561
  由 Tim Cooijmans 提交于 1月 11, 2016
  
  78bbb561
- BatchedDot: fix bad pointer arithmetic · 1ac274a9
  由 Tim Cooijmans 提交于 1月 11, 2016
  
  1ac274a9
- BatchedDot: fix potential bug · c56c2f70
  由 Tim Cooijmans 提交于 1月 11, 2016
  
  c56c2f70
18 12月, 2015 19 次提交
- BatchedDot: PyArray_Newshape complains about NPY_KEEPORDER when it shouldn't · 0f03bac6
  由 Tim Cooijmans 提交于 12月 17, 2015
  
  0f03bac6
- BatchedDot: generalize matrix-matrix code to matrix-vector, vector-matrix, vector-vector products · 9a3f6681
  由 Tim Cooijmans 提交于 12月 17, 2015
  
  9a3f6681
- BatchedDot: first stab at C implementation · 4ce8a480
  由 Tim Cooijmans 提交于 12月 16, 2015
  
  4ce8a480
- BatchedDot: fix comments and remove assertion · cccef96c
  由 Tim Cooijmans 提交于 12月 15, 2015
  
  cccef96c
- BatchedDot: fix optimization to handle matrices · a3fc110c
  由 Tim Cooijmans 提交于 12月 01, 2015
  
  a3fc110c
- BatchedDot: remove unnecessary __str__ · 82aac7d9
  由 Tim Cooijmans 提交于 12月 01, 2015
  
  82aac7d9
- BatchedDot: remove reference to Dot · a915c821
  由 Tim Cooijmans 提交于 12月 01, 2015
  
  a915c821
- BatchedDot: trust that gradients are floats · 62e2bf52
  由 Tim Cooijmans 提交于 12月 01, 2015
  
  62e2bf52
- BatchedDot: fix grad · cc8c9e4c
  由 Tim Cooijmans 提交于 11月 26, 2015
  
  cc8c9e4c
- BatchedDot: more diverse tests · 2202eb41
  由 Tim Cooijmans 提交于 11月 25, 2015
  
  2202eb41
- BatchedDot: catch bad inputs · bee1a7a0
  由 Tim Cooijmans 提交于 11月 25, 2015
  
  bee1a7a0
- BatchedDot: write docstrings · 91fe0a2b
  由 Tim Cooijmans 提交于 11月 16, 2015
  
  91fe0a2b
- introduce optimization to move BatchedDot to GPU · 1a285716
  由 Tim Cooijmans 提交于 10月 23, 2015
  
  1a285716
- BatchedDot: provide dtype in numpy reference implementation · 7d1c9917
  由 Tim Cooijmans 提交于 10月 23, 2015
  
  7d1c9917
- BatchedDot: test and fix grad · ed4e0679
  由 Tim Cooijmans 提交于 10月 23, 2015
  
  ed4e0679
- BatchedDotOp: simplify tests · e68999e8
  由 Tim Cooijmans 提交于 10月 23, 2015
  
  e68999e8
- BatchedDot: implement and test · f6c48641
  由 Tim Cooijmans 提交于 10月 16, 2015
  
  f6c48641
- BatchedDotOp: implement and test infer_shape · 713fd0fe
  由 Tim Cooijmans 提交于 10月 09, 2015
  
  713fd0fe
- Merge pull request #3788 from nouiz/carriepl-v4 · 4ad36ddc
  由 Frédéric Bastien 提交于 12月 17, 2015
```
Rebased cudnn v4
```
  4ad36ddc
17 12月, 2015 1 次提交
- Merge pull request #3786 from lamblin/abstract_conv_infershape · 22db3930
  由 Frédéric Bastien 提交于 12月 16, 2015
```
Abstract conv: fix tests and add infer_shape
```
  22db3930
16 12月, 2015 9 次提交
- Merge pull request #3787 from oeway/master · 725b7a3f
  由 Frédéric Bastien 提交于 12月 16, 2015
```
fixed cula runtime error (libcula.so not found)
```
  725b7a3f
- Flake8 · 32e113c1
  由 carriepl 提交于 11月 25, 2015
  
  32e113c1
- Make dnn conv in gpuarray work with AbstractConv · ced46b7d
  由 carriepl 提交于 11月 25, 2015
  
  ced46b7d
- Standardize documentation of the 'precision' parameter · aafc23aa
  由 carriepl 提交于 11月 12, 2015
  
  aafc23aa
- Update and standardize error messages · 7d44a89c
  由 carriepl 提交于 11月 12, 2015
  
  7d44a89c
- Clean up V2 code · 524438b6
  由 carriepl 提交于 11月 12, 2015
  
  524438b6
- Avoid using method cudnnGetConvolution2dDescriptor() method · 877f4210
  由 carriepl 提交于 11月 10, 2015
  
  877f4210
- Make deprecated dnn flags raise more meaningfull error messages. · ebd97785
  由 carriepl 提交于 11月 10, 2015
  
  ebd97785
- Add 'as_input' option to dnn.conv.precision flag · 32f11551
  由 carriepl 提交于 11月 10, 2015
  
  32f11551