利用ncnn2mem 转成二进制的之后加载（net.load_param_bin）就会报：layer -1 not exists or registered

[ChisenZhang]

darknet 模型，转成ncnn模型后直接调用没问题；利用ncnn2mem 转成二进制的之后加载（net.load_param_bin）就会报：layer -1 not exists or registered 用net.load_param加载就会报param is too old, please regenerate的错误？ 什么原因？

[ChisenZhang]

模型长这样

[net]

Training

batch=16

subdivisions=4

Testing batch=1 subdivisions=1

height=256 width=256 channels=3

min_crop=128 max_crop=256 burn_in=1000 learning_rate=0.0005 # 初始学习率 policy=poly power=4 max_batches=800000 # 最大bath momentum=0.9 # 动量 decay=0.0005 # 权重衰减正则项

angle=7 # 旋转角度 hue=.1 # 色调 saturation=.75 # 饱和度 exposure=.75 # 曝光量

aspect=.75

adam=1 B1=0.9 B2=0.999 eps=1e-7

[convolutional] filters=16 size=7 stride=4 pad=1 batch_normalize=1 activation=leaky

S1

[convolutional] filters=32 size=1 stride=1 pad=1 batch_normalize=1 activation=leaky

DWConv

[convolutional] filters=32 size=3 stride=2 groups=32 pad=1 batch_normalize=1 activation=leaky

[convolutional] filters=16 size=1 stride=1 pad=1 batch_normalize=1 activation=linear

S2

[convolutional] filters=32 size=3 stride=2 pad=1 batch_normalize=1 activation=leaky

DWConv

[convolutional] filters=64 size=1 stride=1 pad=1 batch_normalize=1 activation=leaky

[avgpool]

[convolutional] filters=2 size=1 stride=1 pad=1 activation=linear

[softmax] groups=1

[xiangweizeng]

官方的工具目前不支持自定义网络层， 需调整后方可支持，但目前本工具还未支持，后续考虑加入相关的功能支持。