- 14 12月, 2011 4 次提交
-
-
由 Razvan Pascanu 提交于
-
由 Razvan Pascanu 提交于
parameters.
-
由 Razvan Pascanu 提交于
the node.
-
由 Razvan Pascanu 提交于
property
-
- 06 12月, 2011 14 次提交
-
-
由 Razvan Pascanu 提交于
The body of make_node has been updated to make use of the newly revised methods of scan to parse arguments, hopefully increasing clarity.
-
由 Razvan Pascanu 提交于
-
由 Razvan Pascanu 提交于
The make_node function used to check if inputs of the scan op have the same type. In order to do so it has to parse them, This comments indicate the convention in naming variables when this check is done (so that there is no confusion between the variables representing the inputs of the lambda functions, i.e. slices of the inputs and the actual inputs).
-
由 Razvan Pascanu 提交于
Scan arguments are ordered, and depending on the index, an argument can be a sequence, a state/output or a non sequence. The scan op has methods that given the full list of arguments returns only those that represent a certain category of inputs. The old functions used to look at self.inputs, self.outputs or node.inputs/node.outputs to determine those entries. This approach was restrictive, since all functions on scan most of the time work on clones of the original arguments. The new functions now take a list of arguments (could be either the original or clones).
-
由 nouiz 提交于
edited NEWS.txt
-
由 Ian Goodfellow 提交于
-
由 nouiz 提交于
Added new command 'theano-cache unlock'
-
由 Olivier Delalleau 提交于
This can be used to unlock the compilation directory without having to wait two minutes, or having to manually delete a file.
-
由 Olivier Delalleau 提交于
Minor fix in an theano/tensor/blas.py.
-
由 nouiz 提交于
First version of file for the release candidate.
-
由 Frederic 提交于
-
由 nouiz 提交于
Changed argmax dtype from int32 to int64
-
由 nouiz 提交于
Refactored GPU install instructions
-
由 Justin Bayer 提交于
-
- 05 12月, 2011 16 次提交
-
-
由 Olivier Delalleau 提交于
Also refactored a bit GPU tips, and removed the note about limitation to 4 GPUs (which is not true anymore).
-
由 Olivier Delalleau 提交于
Moved instructions from "Using the GPU" tutorial page to the intallation instructions. Added a link from installation instructions to the "Using the GPU" tutorial.
-
由 nouiz 提交于
Minor stuff
-
由 nouiz 提交于
Fix argmax grad
-
由 nouiz 提交于
Some test fixes
-
由 nouiz 提交于
Fixed some tests with compute_test_value = ignore
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
Now also testing gradient with axis=None (failing as of this commit).
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
If 'axis' is empty then x.ndim must be equal to zero, and thus we also have 'axis == range(x.ndim)'. If x.ndim in non-zero, there is no reason for axis to be empty, so better not catch this case.
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
Windows requires pickled files to be opened in binary mode. Also rewrote file operations to be safer.
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
-
- 04 12月, 2011 6 次提交
-
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
-
由 Razvan Pascanu 提交于
Fixed test_lazy.py and test_vm.py Everything looks fine.
-
由 Olivier Delalleau 提交于
-
由 Olivier Delalleau 提交于
Fixed test_vm.py:TestCallbacks.test_callback_with_ifelse.
-