coldmooon / code-for-mpelu Goto Github PK

I wonder if this torch version of code can run in Pytorch? I use Pytorch. I don't know how to use the module you wrote in pytorch.

Use your code on caffe can be compiled successfully and successfully trained, but in the right straight visualization: "F0724 12:23:30.799641 24843 layer_factory.hpp:81] Check failed: registry.count(type) == 1 (0 vs. 1) Unknown layer type: M2PELU "

[----------] 12 tests from SGDSolverTest/0, where TypeParam = caffe::CPUDevice
[ RUN ] SGDSolverTest/0.TestLeastSquaresUpdateLROneHundredth
[libprotobuf ERROR google/protobuf/text_format.cc:245] Error parsing text-format caffe.SolverParameter: 1:113: Message type "caffe.SolverParameter" has no field named "layer_wise_reduce".
F0509 09:21:30.853242 5167 test_gradient_based_solver.cpp:56] Check failed: google::protobuf::TextFormat::ParseFromString(proto, &param)

The latest 'caffe-master' can be compiled completely. So the 'caffe-master' files has no problem.
Is there anyone who meets and solves this problem?

layer {
name: "m2pelu_conv1"
type: "M2PELU"
bottom: "conv1"
top: "conv1"
param {
lr_mult: 10
decay_mult: 0
}
param {
lr_mult: 10
decay_mult: 0
}
m2pelu_param {
alpha_filler {
type: "constant"
value: 1
}
beta_filler {
type: "constant"
value: 1
}
}
}

base_lr: 0.0001
lr_policy: "multistep"
gamma: 0.1
stepvalue: 100000
stepvalue: 600000
momentum: 0.9
weight_decay: 0.0005
max_iter: 930000

Sorry.
Check failed: false Unknown filler name: taylor.
How can I have the filter? Can you help me ?