projects / platform / upstream / caffe.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e294f6a) | patch
fix caffe train GPU initialization
authorJonathan L Long <jonlong@cs.berkeley.edu>
Mon, 15 Sep 2014 21:15:58 +0000 (14:15 -0700)
committerJonathan L Long <jonlong@cs.berkeley.edu>
Mon, 15 Sep 2014 21:15:58 +0000 (14:15 -0700)
commitbbd166e07313a6b30c68f3a7557e81316adb3d14
tree771033edfaab78659a0c495a2a972f7df4d3eba5tree | snapshot
parente294f6a6e4f5dfc0cc260301a1b04f86c042f7c0commit | diff
fix caffe train GPU initialization

Previously, the solver constructed nets before the caffe train tool read
the --gpu flag, which can cause errors due to LayerSetUp executing on
the wrong device (breaking cuDNN, for example).
src/caffe/solver.cpp diff | blob | history
tools/caffe.cpp diff | blob | history
Domain: Machine Learning / ML Framework;