原文链接：https://github.com/BVLC/caffe/issues/2597

Hi,

 

Your GPU compute capacity is incorrect. Nano is sm=53, not sm=62.

Here is my installation steps for your reference:

 

1. Install dependencies

1. sudo apt-get update
2. sudo apt-get install software-properties-common
3. sudo add-apt-repository universe
4. sudo add-apt-repository multiverse

 

1. sudo apt-get install libboost-dev libboost-all-dev
2. sudo apt-get install libjpeg-turbo8-dev libjpeg8-dev libturbojpeg0-dev
3. sudo apt-get install liblapack-dev liblapack3 libopenblas-base libopenblas-dev
4. sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev libatlas-base-dev liblmdb-dev libblas-dev libatlas-base-dev libprotobuf-dev libleveldb-dev libsnappy-dev libhdf5-serial-dev protobuf-compiler

 

 

2. Prepare source

1. git clone https://github.com/NVIDIA/caffe.git $CAFFE_ROOT

 

>> Edit Makefile.config.example

1. +++ USE_CUDNN := 1

 

1. +++ OPENCV_VERSION := 3

 

1. +++ WITH_PYTHON_LAYER := 1

 

1. --- CUDA_ARCH := -gencode arch=compute_50,code=sm_50 \
2. --- -gencode arch=compute_52,code=sm_52 \
3. --- -gencode arch=compute_60,code=sm_60 \
4. --- -gencode arch=compute_61,code=sm_61 \
5. --- -gencode arch=compute_61,code=compute_61
6. +++ CUDA_ARCH := -gencode arch=compute_53,code=sm_53 \
7. +++ -gencode arch=compute_53,code=compute_53

 

 

3. Build

1. cp Makefile.config.example Makefile.config
2. make -j4
3. make pycaffe
4. export PYTHONPATH=$PWD/python:$PYTHONPATH

 

 

I can build caffe successfully with the steps above.

Thanks.

 

#10

Posted 04/24/2019 05:56 AM   

 

Issues:

 

Can't build with error ‘::gflags’ has not been declared 

https://github.com/BVLC/caffe/issues/2597

I also got this problem, and I installed gflags by myself before . And I solved this problem : in the file include/caffe/common.hpp //#ifndef GFLAGS_GFLAGS_H_ namespace gflags = google; //#endif // GFLAGS_GFLAGS_H_ comment ; namespace gflags = google; this code is important