目录
我的显卡是RTX3080,30系列的显卡只支持CUDA11及以上版本,只有Ubuntu20才能安装CUDA11。由于显卡算力太高,即使安装CUDA9和Ubuntu16也无法匹配,会出现cudnn(4 vs 0)或者(8 vs 0)的问题。所以我把显卡换成了1080Ti,系统为Ubuntu16,CUDA9.0,cudnn7.0.5,Python2.7.然后再进行编译,可以训练。
1、下载源码
AffordanceNet代码github地址为GitHub - nqanh/affordance-net: AffordanceNet - Multiclass Instance Segmentation Framework - ICRA 2018将代码下载下载下来之后,将根目录命名为AffordanceNetRoot
cd AffordanceNetRoot
cd caffe-affordance-net
2、安装依赖
安装caffe的依赖库,如果不是在root权限下,则需要在下面的每个命令前加上sudo
apt install -y libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler
apt install -y --no-install-recommends libboost-all-dev
apt install -y libatlas-base-dev
apt install -y libgflags-dev libgoogle-glog-dev liblmdb-dev
3、编译caffe-affordance-net
这是根据caffe源码改的,与编译caffe流程基本一致,但仍然在此列出来。
- 复制makefile.config文件
-
cp Makefile.config.example Makefile.config
- 我是使用gpu加速,所以要去掉USE_CUDNN := 1的注释。
- 去掉OPENCV_VERSION := 3的注释,我的opencv版本为3.4.5.
- 注释掉下面代码段的前两行,
CUDA_ARCH := #-gencode arch=compute_20,code=sm_20 \
# -gencode arch=compute_20,code=sm_21 \
-gencode arch=compute_30,code=sm_30 \
-gencode arch=compute_35,code=sm_35 \
-gencode arch=compute_50,code=sm_50 \
-gencode arch=compute_52,code=sm_52 \
-gencode arch=compute_60,code=sm_60 \
-gencode arch=compute_61,code=sm_61 \
-gencode arch=compute_61,code=compute_61
- 去掉WITH_PYTHON_LAYER := 1的注释
- 更改以下代码。
PYTHON_LIBRARIES := boost_python python2.7
PYTHON_INCLUDE := /usr/include/python2.7 \
/usr/lib/python2.7/dist-packages/numpy/core/include
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial
make -j36 #如果电脑核数更多,可以使用更多核数
make pycaffe
添加环境变量
sudo gedit~/.bashrc
export PYTHONPATH=/home/qxn/AffordanceNetRoot/caffe-affordance-net/python:$PYTHONPATH #添加环境变量
#保存并关闭bashrc文件
source ~/.bashrc#使环境变量生效
验证是否成功,如图所示,即为成功。
4、编译lib
cd AffordanceNetRoot/lib
make
5、运行demo
python /tools/demo_img.py
6、训练
在训练过程中,出现问题
/home/qxn/AffordanceNetRootv1/tools/../lib/fast_rcnn/bbox_transform.py:49: RuntimeWarning: overflow encountered in exp
pred_h = np.exp(dh) * heights[:, np.newaxis]
/home/qxn/AffordanceNetRootv1/tools/../lib/fast_rcnn/bbox_transform.py:49: RuntimeWarning: overflow encountered in multiply
pred_h = np.exp(dh) * heights[:, np.newaxis]
/home/qxn/AffordanceNetRootv1/tools/../lib/fast_rcnn/bbox_transform.py:48: RuntimeWarning: overflow encountered in exp
pred_w = np.exp(dw) * widths[:, np.newaxis]
/home/qxn/AffordanceNetRootv1/tools/../lib/fast_rcnn/bbox_transform.py:48: RuntimeWarning: overflow encountered in multiply
pred_w = np.exp(dw) * widths[:, np.newaxis]
I1215 21:00:21.940760 19779 solver.cpp:229] Iteration 60, loss = nan
lib/setup.py的-arch=sm_61,因为1080Ti的算力为6.1,所以改为sm_61.