| commit | 123cb5dd07f16bdc4aa68c7250c7008c9a2204cf | [log] [tgz] |
|---|---|---|
| author | Aapo Kyrola <akyrola@fb.com> | Fri Oct 13 13:58:22 2017 -0700 |
| committer | Facebook Github Bot <facebook-github-bot@users.noreply.github.com> | Fri Oct 13 14:02:48 2017 -0700 |
| tree | 88298003015b3f5367e9a0c330e3bb08fa158fee | |
| parent | 4c3b02f31447037b009e80a0251c0ebd0ee95b7c [diff] |
use non-cudnn transpose for int tensors Summary: Turns out CuDNN's tensor transform only supports floats. Previous implementation pretended it would work with ints by casting to floats and indeed passed tests for some reason. But rgirdhar found a case where it returned nonsensical results. So rewire int-transposes to use non-cudnn version. Had to refactor a bit for that. Also added a test for the case. Reviewed By: asaadaldien Differential Revision: D6043284 fbshipit-source-id: cc3b14f9fbbdeff421b01da453a1d3c7c5ffd4ac
Caffe2 is a lightweight, modular, and scalable deep learning framework. Building on the original Caffe, Caffe2 is designed with expression, speed, and modularity in mind.
Caffe2 research award competition request for proposals
Please use Github issues (https://github.com/caffe2/caffe2/issues) to ask questions, report bugs, and request new features.
Please participate in our survey (https://www.surveymonkey.com/r/caffe2). We will send you information about new releases and special developer events/webinars.
Caffe2 is released under the Apache 2.0 license. See the NOTICE file for details.