I0307 11:58:16.956048 2795 net.cpp:226] Convolution2 needs backward computation.
I0307 11:58:16.956056 2795 net.cpp:226] Convolution1_ReLU1_0_split needs backward computation.
I0307 11:58:16.956068 2795 net.cpp:226] ReLU1 needs backward computation.
I0307 11:58:16.956076 2795 net.cpp:226] Scale1 needs backward computation.
I0307 11:58:16.956116 2795 net.cpp:226] BatchNorm1 needs backward computation.
I0307 11:58:16.956130 2795 net.cpp:226] Convolution1 needs backward computation.
I0307 11:58:16.956140 2795 net.cpp:228] Data2_Data1_1_split does not need backward computation.
I0307 11:58:16.956151 2795 net.cpp:228] Data1 does not need backward computation.
I0307 11:58:16.956158 2795 net.cpp:270] This network produces output Accuracy1
I0307 11:58:16.956169 2795 net.cpp:270] This network produces output SoftmaxWithLoss1
I0307 11:58:16.956445 2795 net.cpp:283] Network initialization done.
I0307 11:58:16.958571 2795 solver.cpp:60] Solver scaffolding done.
I0307 11:58:16.959854 2795 caffe.cpp:251] Starting Optimization
I0307 11:58:16.959905 2795 solver.cpp:279] Solving resnet_cifar10
I0307 11:58:16.959976 2795 solver.cpp:280] Learning Rate Policy: multistep
I0307 11:58:16.963498 2795 solver.cpp:337] Iteration 0, Testing net (#0)
I0307 12:04:58.970382 2795 solver.cpp:404] Test net output #0: Accuracy1 = 0
I0307 11:58:16.956056 2795 net.cpp:226] Convolution1_ReLU1_0_split needs backward computation.
I0307 11:58:16.956068 2795 net.cpp:226] ReLU1 needs backward computation.
I0307 11:58:16.956076 2795 net.cpp:226] Scale1 needs backward computation.
I0307 11:58:16.956116 2795 net.cpp:226] BatchNorm1 needs backward computation.
I0307 11:58:16.956130 2795 net.cpp:226] Convolution1 needs backward computation.
I0307 11:58:16.956140 2795 net.cpp:228] Data2_Data1_1_split does not need backward computation.
I0307 11:58:16.956151 2795 net.cpp:228] Data1 does not need backward computation.
I0307 11:58:16.956158 2795 net.cpp:270] This network produces output Accuracy1
I0307 11:58:16.956169 2795 net.cpp:270] This network produces output SoftmaxWithLoss1
I0307 11:58:16.956445 2795 net.cpp:283] Network initialization done.
I0307 11:58:16.958571 2795 solver.cpp:60] Solver scaffolding done.
I0307 11:58:16.959854 2795 caffe.cpp:251] Starting Optimization
I0307 11:58:16.959905 2795 solver.cpp:279] Solving resnet_cifar10
I0307 11:58:16.959976 2795 solver.cpp:280] Learning Rate Policy: multistep
I0307 11:58:16.963498 2795 solver.cpp:337] Iteration 0, Testing net (#0)
I0307 12:04:58.970382 2795 solver.cpp:404] Test net output #0: Accuracy1 = 0
I0307 12:04:59.017705 2795 solver.cpp:404] Test net output #1: SoftmaxWithLoss1 = -nan (* 1 = -nan loss)
killed
原因:内存不够