https://github.com/Microsoft/CNTK
History
Tip revision: a3bd7819d19453534f33b8292c771db32147f5a4 authored by kaituoxu on 15 August 2018, 05:40:31 UTC
[BiVfsmnNode] avoid CPU GPU transfer but still use hard code minibatch size
Tip revision: a3bd781
File Mode Size
AllReduceDistGradAggregator.h -rw-r--r-- 29.8 KB
BlockMomentumDistributedLearner.h -rw-r--r-- 29.7 KB
BlockMomentumSGD.h -rw-r--r-- 14.2 KB
MatrixQuantizer.h -rw-r--r-- 2.6 KB
QuantizedDataParallelDistributedLearner.h -rw-r--r-- 4.1 KB
QuantizedDistributedCommunicator.h -rw-r--r-- 29.1 KB
V2AllReduceDistGradAggregator.h -rw-r--r-- 12.9 KB
V2BlockMomentumSGD.h -rw-r--r-- 16.6 KB

back to top