Environment Setup
- Install the development toolkit Ascend-cann-toolkit_{version}_linux-{arch}.run or the deep learning engine package Ascend-cann-nnae_{version}_linux-{arch}.run, as well as the framework package. For details, see Environment Setup.
- Install the dependencies listed in Table 1.
Table 1 Dependency information Dependency
Version
gcc
8.4 or later
You can run the cc --version or c++ --version command to check the GCC version in use.
g++
zip
Any version
unzip
libtool
automake
Python
3.7.5
TensorFlow
1.15.0
tensorflow-serving-api
1.15.0
future
Any version
bazel
0.24.1 or later
CMake
3.14.0 or later
swig
- If the operating system architecture is aarch64, the software version must be 3.0.12 or later.
- If the operating system architecture is x86_64, the software version must be 4.0.1 or later.
- The GCC and G++ versions must be the same. Otherwise, an error may occur during TF Serving source code compilation.
- For details about how to use bazel for compilation, see Installing bazel 0.24.1 from Source Code.
- For details about how to use CMake for compilation, see Installing CMake 3.14.0 from Source Code.
- If the swig software package fails to be installed, rectify the fault by referring to What Should I Do If the swig Dependency Package Cannot Be Installed?.
- The gcc, g++, zip, unzip, libtool, and automake software packages must be installed using apt or yum. The TensorFlow, tensorflow-serving-api, and future software packages must be installed using pip3.