test-mordecai

locally installing and running mordecai from the windows command line

Instructions to make

Create an environment

Using Conda

conda create --name mordecai_env_3.8 python=3.8
conda activate mordecai_env_3.8

Or using pip in bash build from scratch

$ python -m venv mordecai_env
$ source mordecai_env/bin/activate # in windows cmd it's mordecai_env\Scripts\activate

then install packages

$ python -m pip install ipykernel
$ python -m pip install mordecai
$ python -m pip install spacy==2.3
$ python -m pip install numpy==1.21
$ python -m pip install h5py==3.8.0
$ python -m pip install futures
$ python -m pip install torch   
$ python -m pip install pip install ipykernel rioxarray

Packages for the gridding $conda install -c conda-forge cartopy $ python -m pip install shapely global_land_maskgeopandas

OR

$ python -m pip install -r requirements.txt

Then use ipykernel to open the environment in jupyter

$ ipython kernel install --user --name=mordecai_env_3.8
$ jupyter-lab

Then open juypyter lab and in the upper right hand select the kernel for mordecai_env

Install packages

In jupyter, run the file: Installing_mordecai.pynb to install the necessary packages and set up docker. Docker must be installed first (see https://docs.docker.com/engine/installation/).

Instructions to initialize on new computer

in windows cmd (run as admin)

cd (path to working directory)
conda env create -f mordecai_env_3.8.yml
conda activate mordecai_env_3.8
conda install -c anaconda ipykernel
ipython kernel install --user --name=envname
jupyter-lab

Then in open the file "installing_mordecai.ipynb". In the upper RH corner, select envname as the kernel, and run the installation for mordecai and download the spacy library. Also run the docker commands if needed

Without ipykernel, run from cmd line

Without ipykernel, the code will only run from the windows command line

Ctrl+C (twice to exit jupyter lab)

(path to conda env python.exe) 

Now you are running python from the cmd line: from mordecai import Geoparser geo = Geoparser() geo.geoparse("I travelled from Oxford to Ottawa") exit()

(path to conda env python.exe) test-geoparse.py

Instructions to run once already initialized

In Windows cmd (as admin)

conda activate mordecai_env_3.8

cd (path to working directory)

docker run -d -p 127.0.0.1:9200:9200 -v (path to wd)\geonames_index:/usr/share/elasticsearch/data elasticsearch:5.5.2 

Or

docker run -d -p 127.0.0.1:9200:9200 -v $(pwd)/geonames_index/:/usr/share/elasticsearch/data elasticsearch:5.5.2

And then one of the following commands: /usr/bin/python3 analyses/01_geoparse-mordecai.py # [path to python interpreter] [path to script]

(path to conda env python.exe) (name of .py script to run)

jupyter-lab # then select mordecai_env_3.8 for the kernel

What I'm using in rossinante

• Note to check if the docker is already running, use docker ps -a ; and to remove docker rm "random name"

  screen -S mordecai conda activate mordecai_env_3.8 cd (path to working directory) docker run -d -p 127.0.0.1:9200:9200 -v $(pwd)/geonames_index/:/usr/share/elasticsearch/data elasticsearch:5.5.2

INSTEAD, try running this to limit the number of cpus that can be used: docker run -d -p 127.0.0.1:9200:9200 -cpus="10.0" -v $(pwd)/geonames_index/:/usr/share/elasticsearch/data elasticsearch:5.5.2

For script in command line /usr/bin/python3 analyses/01_geoparse-mordecai.py # [path to python interpreter] [path to script]

For working in jupyter lab jupyter-lab # then select mordecai_env_3.8 for the kernel