Run on Remote Server¶

Runs EPM on dedicated infrastructure with more computational capacity than a local machine. Suited for large scenario batches and Monte Carlo runs that would be impractical to run locally.

Prerequisites¶

A World Bank computer, or VDI access with a YubiKey
EPM tested and working locally before running on the server

1. Connect¶

Go to privx.worldbank.org/auth/login
Sign in with your Microsoft credentials
In the Connections tab, select a host (e.g. Linux-xxx-.worldbank.org)

Once connected you have access to a Terminal and a Files tab for uploads/downloads.

2. Clone the repository¶

All data and code must be stored in /Data, not in /home.

cd /Data
mkdir your_project_folder
cd your_project_folder
git clone --branch your-branch-name --single-branch https://github.com/ESMAP-World-Bank-Group/EPM.git
cd EPM/epm

3. Keep code in sync¶

The server is for running simulations, not for development. Work locally, then sync:

# On your local machine
git add .
git commit -m "ready for server run"
git push origin your-branch-name

# On the server
cd /Data/your_project_folder/EPM
git pull origin your-branch-name

4. Run EPM¶

RAM usage

The cluster has 96 GB of total RAM. Individual runs can consume 10+ GB each. Monitor usage carefully when running in parallel or alongside other users.

Python (recommended)¶

First time only, add Miniconda to your PATH:

echo 'export PATH="/Data/miniconda3/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

Then run:

conda activate epm_env
cd /Data/your_project_folder/EPM/epm
python epm.py --folder_input your_data --config your_data/config.csv --scenarios --cpu 8 --output_zip

Use --output_zip to compress results into a single file (the server only allows downloading files, not folders).

GAMS (for low-level debugging only)¶

cd /Data/your_project_folder/EPM/epm
gams main.gms --FOLDER_INPUT your_data

5. Run in background¶

For long runs, use tmux so the job continues if you disconnect:

tmux new -s epmrun
# launch your command here
python epm.py --folder_input your_data --config your_data/config.csv --output_zip

Detach without stopping the job: Ctrl + B, then D

tmux attach -t epmrun      # reconnect to the session
tmux list-sessions          # see all active sessions

To check if your job is still running:

ps aux | grep epm.py

Find your username in the output; if the line is there, the job is running. To stop it:

kill -9 <PID>

Terminal command reminder

Action	Command
List files	`ls` or `ls -l`
Change directory	`cd folder_name`
Go up one level	`cd ..`
Print current path	`pwd`
Create folder	`mkdir folder_name`
Delete folder	`rm -rf folder_name`
Monitor CPU/RAM	`top` or `free -h`
Find heavy processes	`ps aux --sort=-%mem \\| head -10`