Question

我试图安装squeezeNet，这些是说明

SqueezeNet_v1.0/train_val.prototxt          #model architecture
SqueezeNet_v1.0/solver.prototxt             #additional training details (learning rate schedule, etc.)
SqueezeNet_v1.0/squeezenet_v1.0.caffemodel  #pretrained model parameters

当运行第一个命令时，我收到此错误，我不知道该怎么办？

decentmakeover@DecentMakeover:~/caffe-master$ ./build/tools/caffe 
train -solver SqueezeNet/SqueezeNet_v1.0/train_val.prototxt
[libprotobuf ERROR google/protobuf/text_format.cc:291] Error parsing 
text-format caffe.SolverParameter: 8:7: Message type 
"caffe.SolverParameter" has no field named "layer".
F0718 20:10:12.568349 10005 upgrade_proto.cpp:1101] Check failed: 
ReadProtoFromTextFile(param_file, param) Failed to parse 
SolverParameter file: SqueezeNet/SqueezeNet_v1.0/train_val.prototxt
*** Check failure stack trace: ***
    @     0x7fb3dac635cd  google::LogMessage::Fail()
    @     0x7fb3dac65433  google::LogMessage::SendToLog()
    @     0x7fb3dac6315b  google::LogMessage::Flush()
    @     0x7fb3dac65e1e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7fb3db033ef1  caffe::ReadSolverParamsFromTextFileOrDie()
    @           0x40a4d8  train()
    @           0x4072f8  main
    @     0x7fb3d92ec830  __libc_start_main
    @           0x407bc9  _start
    @              (nil)  (unknown)
Aborted (core dumped)

Answer 1

train_val.prototxt是拓扑描述。这仅适用于model参数，如caffe test命令。

对于caffe train命令，您使用solver选项（就像您一样），但您必须提供solver.prototxt文件的名称（可能是“solver.prototxt”） “）。

添加澄清

你应该有一个类似下面的解决方案文件。确保您在求解器文件和下面net:路径的路径准确无误。

net: "/home/prune/caffe/models/googlenet_v1/train_val.prototxt"
test_iter: 1000
test_interval: 4000
display: 50
average_loss: 50
base_lr: 0.01

Answer 2

您需要提供'solver.prototxt'作为caffe train的输入参数，而不是'train_val.prototxt

挤压网问题

2 个答案: