pyhanlp is a natural language processing toolkit developed based on Java. Since my entire project is written based on Python, I stepped on a lot of pits when installing pynlp, and recorded it as a reference for other people in need.
1. Upgrade TensorFlow to the specified version
pip install tensorflow==version number -i https://pypi.doubanio.com/simple/ --trusted-host pypi.doubanio.com
pyhanlp requires TensorFlow version 2.0, so version 1.0 needs to be upgraded. Remember to use Douban mirror source will be much faster.
2. Configure conda mirror source
Stepped on a lot of pits in this step, enter the anaconda configuration file
vi /Users/limingyu/.condarc
The final successful configuration is as follows
channels:
- http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
- http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
ssl_verify: true
show_channel_urls: true
Note that http is not https, and /free and /main of the Tsinghua mirror source must be added, otherwise there will be some dependencies that are not available
3. Upgrade conda
If the following error occurs
RemoveError: 'requests' is a dependency of conda and cannot be removed
It is because the version of conda is too old. Some packages installed by pip have a little problem. The solution is also very simple, just upgrade conda
conda update conda
4. Install gcc with conda
conda install gcc
5. Install jpype1 with conda
conda install -c conda-forge jpype1
6. Finally you can install pyhanlp directly
pip install pyhanlp
Got it, finally verify it in Python
from pyhanlp import *