Using ClearML interactive sessions

ClearML interactive sessions allow you to launch a JupyterLab, VS Code, and SSH session to execute code on a remote GPU. For more information, see the documentation here.

Installation

If you have any issues using ClearML Sessions, check the prerequisites in the documentation. If you have already set up ClearML for silnlp (and have an ssh client installed, check by running ssh), there should be no additional steps before installation.

To install, run:

pip install clearml-session

VPN

To connect to GPUs in the Dallas server, you will need a VPN connection.

Usage

Launch a sesison with silnlp's published docker image to have a pre-configured environment to work in:

clearml-session --docker ghcr.io/sillsdev/silnlp:latest

Once the session has been set up, you will be given links to the JupyterLab and VS Code sessions, as well as a port number and password for SSH.

Running silnlp scripts

1. Get latest code
   • Run git pull in a terminal to get the most recent version of silnlp
   • To work on a specific branch, run git fetch origin and git checkout branch_name
2. Install necessary libraries

   curl -sSL https://install.python-poetry.org | python - --version 1.7.1
   export PATH="/root/.local/bin:$PATH"
   poetry export -f requirements.txt --output requirements.txt
   pip install -r requirements.txt
   

3. Run scripts with --clearml-queue local to have them logged on ClearML without sending the job to another GPU.