Welcome to mirror list, hosted at ThFree Co, Russian Federation.

github.com/soumith/cudnn.torch.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
AgeCommit message (Expand)Author
2015-08-19streams fixsoumith
2015-08-12Fixed calling reset() twice (saving time and making weights initialized as in...Martin Simonovsky
2015-08-02working R3 bindings for non-new modulesSoumith Chintala
2015-07-09small change in how bias applied in groupsSergey Zagoruyko
2015-06-19fixes memory overalloc bugSergey Zagoruyko
2015-05-06fixing serialization to not print annoying warningssoumith
2015-04-10using the new streams API (cudnn does not ovelap compute yet, weird)soumith
2015-04-02fixing bug pointed out by @nagadomisoumith
2015-03-27choosing reasonable defaults for buffer-limits in convolution module, with op...soumith
2015-01-02adding groups to SpatialConvolutionsoumith
2014-12-20lint fixes (80 columns)soumith
2014-12-20fixing potential contention bugsoumith
2014-12-19everything works with R2. all unit tests pass. Maxpooling has free zero-paddingSoumith Chintala
2014-11-26lint fixessoumith
2014-11-18adding non-batch modesoumith
2014-10-28for spatial convolution if self.gradInput is nil, then dont do gradinput (fir...soumith
2014-09-28adding SoftMax and SpatialSoftMax bindingsSoumith Chintala
2014-09-26adding serialization to unit tests and fixing descriptor checks. Fixes #4Soumith Chintala
2014-09-22keep compatibility with float networks and their getParametersSoumith Chintala
2014-09-21fixed #3Soumith Chintala
2014-09-21Multi-GPU supportSoumith Chintala
2014-09-21remove unused variableSoumith Chintala
2014-09-21fixes #1Soumith Chintala
2014-09-11first commitSoumith Chintala