在mac上跑resent的简化版简化成17个卷机层，7个eltwise层，这也太慢了,不过cpu上能跑了
I1003 17:26:27.097367 1958465536 solver.cpp:346] Iteration 0, Testing net (#0)
I1003 17:27:34.771304 1958465536 solver.cpp:413] Test net output #0: Accuracy1 = 0.108
I1003 17:27:34.771420 1958465536 solver.cpp:413] Test net output #1: SoftmaxWithLoss1 = 77.9041 (* 1 = 77.9041 loss)
I1003 17:27:39.329684 1958465536 solver.cpp:231] Iteration 0, loss = 2.30613
I1003 17:27:39.329733 1958465536 solver.cpp:247] Train net output #0: SoftmaxWithLoss1 = 2.30613 (* 1 = 2.30613 loss)
I1003 17:27:39.329764 1958465536 sgd_solver.cpp:106] Iteration 0, lr = 0.001
I1003 17:39:41.499366 1958465536 solver.cpp:231] Iteration 200, loss = 1.9399
I1003 17:39:41.499728 1958465536 solver.cpp:247] Train net output #0: SoftmaxWithLoss1 = 1.9399 (* 1 = 1.9399 loss)
I1003 17:39:41.499750 1958465536 sgd_solver.cpp:106] Iteration 200, lr = 0.001
I1003 17:51:43.544186 1958465536 solver.cpp:231] Iteration 400, loss = 1.55321
I1003 17:51:43.545187 1958465536 solver.cpp:247] Train net output #0: SoftmaxWithLoss1 = 1.55321 (* 1 = 1.55321 loss)
I1003 17:51:43.545213 1958465536 sgd_solver.cpp:106] Iteration 400, lr = 0.001
网络拓扑图为：
5090
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
