安裝epel
$ sudo yum install epel-release
安裝依賴的系統(tǒng)庫
# $ sudo yum update
$ sudo yum install -y \
automake \
cmake3 \
gcc \
gcc-c++ \
git \
kernel-devel \
leveldb-devel \
lmdb-devel \
libtool \
protobuf-devel \
python-devel \
python-pip \
snappy-devel \
gflags-devel \
glog-devel
安裝依賴的 Python 庫
$ sudo pip install --upgrade pip
$ sudo pip install \
flask \
future \
graphviz \
hypothesis \
jupyter \
matplotlib \
numpy \
protobuf \
pydot \
python-nvd3 \
pyyaml \
requests \
scikit-image \
scipy \
setuptools \
six \
tornado
編譯安裝 caffe2
$ git clone --recursive https://github.com/caffe2/caffe2
$ cd caffe2 && mkdir build
$ cd build && cmake3 ..
$ sudo make install
測試
首先使用下面的命令來檢查 caffe2 是否安裝成功
$ cd build
$ python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"
如果 caffe2 安裝成功筹煮，上面的命令應該打印 Success筏勒。
其次使用下面的命令來運行一個測試
$ cd build
$ python -m caffe2.python.operator_test.relu_op_test
$ python -m caffe2.python.operator_test.weighted_sample_test
$ python -m caffe2.python.examples.resnet50_trainer --train_data null --use_cpu true
問題
問題一
運行 “python -m caffe2.python.operator_test.relu_op_tes” 報錯
...
Traceback (most recent call last):
File "/usr/lib64/python2.7/runpy.py", line 162, in _run_module_as_main
"__main__", fname, loader, pkg_name)
File "/usr/lib64/python2.7/runpy.py", line 72, in _run_code
exec code in run_globals
File "/home/jhadmin/Downloads/caffe2/build/caffe2/python/operator_test/relu_op_test.py", line 22, in <module>
from hypothesis import given
File "/usr/lib/python2.7/site-packages/hypothesis/__init__.py", line 31, in <module>
from hypothesis.core import given, find, example, seed, reproduce_failure, \
File "/usr/lib/python2.7/site-packages/hypothesis/core.py", line 34, in <module>
from coverage.files import canonical_filename
ImportError: cannot import name canonical_filename
問題原因是 Python 的 coverage 版本問題移迫，升級 coverage 庫
sudo pip install coverage --upgrade
問題二
運行 “python -m caffe2.python.operator_test.relu_op_tes” 報錯
AttributeError: 'module' object has no attribute 'full'
問題原因是 numpy 版本問題，升級 numpy 庫
sudo pip install numpy --upgrade
