caffe继续训练/pretrain_model
作者:互联网
snapshot是后续接着这次的进行训练,梯度学习率什么的参数都是直接copy之后接着训练,比如resnet_50_iter_150000.solverstate
weights只是拷贝过来之前的Model,从第0次迭代开始,比如resnet_50_iter_150000.caffemodel
参考链接:
https://blog.csdn.net/sinat_26917383/article/details/54141697
https://blog.csdn.net/laoxuan2011/article/details/51594155
标签:150000,resnet,pretrain,blog,csdn,details,https,caffe,model 来源: https://blog.csdn.net/qinglv1/article/details/90711698