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

AlexeyAB

2018-04-16 9bae70b22549b68f5cdeece8b6c3b3de00c22714

1
2
3
4
5
6
7
classes= 80
train  = data/coco/trainvalno5k.txt
valid  = data/coco_testdev
#valid = data/coco_val_5k.list
names = data/coco.names
backup = backup/
eval=coco