Accelerated by another 5% using FP16/32 Batch-norm for Tensor Cores.

AlexeyAB

2018-04-16 9bae70b22549b68f5cdeece8b6c3b3de00c22714

 cfg/darknet19_448.cfg

@@ -1,6 +1,8 @@
[net]
batch=128
subdivisions=4
#batch=128
#subdivisions=4
batch=1
subdivisions=1
height=448
width=448
max_crop=512