peaceorwell/caffe-pruned

caffe-pruned's Issues

the refuned model didn't work at step three

your work is pretty good.But I found the refuned model didn't work at step three with a low accuracy.Can u give me some suggesions?

alexnet没有加速效果

您好，首先感谢您的付出，在这里跟您讨论一个问题，程序src/caffe/layers/inner_product_layer.cu中Forward_gpu代码中if (M_ == 1)，表示只支持n=1(nchw)，是吧，但是为什么prototxt文件中的 batch_size不能为1，否则会报错F0519 11:18:17.604454 455 blob.cpp:147] Check failed: csrval_；另外，我也做了alexnet实验，精调之后模型确实能压缩，但是前向运算测试并没有加速效果，请问为什么，我根据代码来看应该也和第一个问题有关吧。

AttributeError

Hi，when I run the python fixModel.py ，I have encounter the following problem,
Traceback (most recent call last):
File "./test_alexnet/fixModel.py", line 49, in
v[0].mask[h_idx][w_idx]=0.0
AttributeError: 'Blob' object has no attribute 'mask'
Can you tell me how to solver this problem? thanks.

reduce graphic memory

how much can reduce graphic memory?thanks
@pluskid

How to solve these warnings and errors about cusparse_descr_, capacity_, sparse_, dirA, fLS::FLAGS_step when making the caffe-pruned without any modification?

warning 1:

./include/caffe/common.hpp: In constructor 'caffe::Caffe::Caffe()':
./include/caffe/common.hpp:173: warning: 'caffe::Caffe::cusparse_handle_' will be initialized after
./include/caffe/common.hpp:172: warning:   'cusparseMatDescr* caffe::Caffe::cusparse_descr_'
src/caffe/common.cpp:107: warning:   when initialized here

warning 2:

./include/caffe/blob.hpp: In constructor 'caffe::Blob<Dtype>::Blob(const std::vector<int, std::allocator<int> >&) [with Dtype = float]':
src/caffe/blob.cpp:979:   instantiated from here
./include/caffe/blob.hpp:347: warning: 'caffe::Blob<float>::capacity_' will be initialized after
./include/caffe/blob.hpp:345: warning:   'bool caffe::Blob<float>::sparse_'
src/caffe/blob.cpp:86: warning:   when initialized here
./include/caffe/blob.hpp: In constructor 'caffe::Blob<Dtype>::Blob(const std::vector<int, std::allocator<int> >&) [with Dtype = double]':
src/caffe/blob.cpp:979:   instantiated from here
./include/caffe/blob.hpp:347: warning: 'caffe::Blob<double>::capacity_' will be initialized after
./include/caffe/blob.hpp:345: warning:   'bool caffe::Blob<double>::sparse_'
src/caffe/blob.cpp:86: warning:   when initialized here
./include/caffe/blob.hpp: In constructor 'caffe::Blob<Dtype>::Blob(const std::vector<int, std::allocator<int> >&) [with Dtype = int]':
src/caffe/blob.cpp:980:   instantiated from here
./include/caffe/blob.hpp:347: warning: 'caffe::Blob<int>::capacity_' will be initialized after
./include/caffe/blob.hpp:345: warning:   'bool caffe::Blob<int>::sparse_'
src/caffe/blob.cpp:86: warning:   when initialized here
./include/caffe/blob.hpp: In constructor 'caffe::Blob<Dtype>::Blob(const std::vector<int, std::allocator<int> >&) [with Dtype = unsigned int]':
src/caffe/blob.cpp:981:   instantiated from here
./include/caffe/blob.hpp:347: warning: 'caffe::Blob<unsigned int>::capacity_' will be initialized after
./include/caffe/blob.hpp:345: warning:   'bool caffe::Blob<unsigned int>::sparse_'
src/caffe/blob.cpp:86: warning:   when initialized here

warning 3:

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

src/caffe/util/math_functions.cu(63): warning: variable "dirA" was declared but never referenced

error

.build_release/lib/libcaffe.so: undefined reference to `fLS::FLAGS_step'
collect2: ld returned 1 exit status
make: *** [.build_release/tools/net_speed_benchmark.bin] Error 1
make: *** Waiting for unfinished jobs....
.build_release/lib/libcaffe.so: undefined reference to `fLS::FLAGS_step'
collect2: ld returned 1 exit status
.build_release/lib/libcaffe.so: undefined reference to `fLS::FLAGS_step'
collect2: ld returned 1 exit status
make: *** [.build_release/tools/train_net.bin] Error 1
make: *** [.build_release/tools/test_net.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `fLS::FLAGS_step'
collect2: ld returned 1 exit status
make: *** [.build_release/tools/finetune_net.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `fLS::FLAGS_step'
collect2: ld returned 1 exit status
make: *** [.build_release/tools/device_query.bin] Error 1

make test error undefined reference to 'fLS::FLAGS_step'

error:
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_upgrade_proto.testbin' failed
make: *** [.build_release/test/test_upgrade_proto.testbin] Error 1
make: *** 正在等待未完成的任务....
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_split_layer.testbin' failed
make: *** [.build_release/test/test_split_layer.testbin] Error 1
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_flatten_layer.testbin' failed
make: *** [.build_release/test/test_flatten_layer.testbin] Error 1
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_filler.testbin' failed
make: *** [.build_release/test/test_filler.testbin] Error 1
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_reshape_layer.testbin' failed
make: *** [.build_release/test/test_reshape_layer.testbin] Error 1
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_neuron_layer.testbin' failed
make: *** [.build_release/test/test_neuron_layer.testbin] Error 1
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_filter_layer.testbin' failed
make: *** [.build_release/test/test_filter_layer.testbin] Error 1
.build_release/lib/libcaffe.so：对‘fLS::FLAGS_step[abi:cxx11]’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:606: recipe for target '.build_release/test/test_solver_factory.testbin' failed
make: *** [.build_release/test/test_solver_factory.testbin] Error 1

How to solve it?

readModel.py

您好，用readModel.py文件无法读取retune的csr矩阵出来，请问您有什么好的解决办法吗？