Python environment¶

When you connect to a terminal over SSH or EC2 Instance Connect, the Python environment at /opt/ml/venv is already active, so python, pip and jupyter all use it. In Session Manager, run sudo -iu ubuntu first.

What is installed¶

Python 3.12
scikit-learn, pandas, NumPy, SciPy and joblib
matplotlib and seaborn, for charts
openpyxl and xlrd, for Excel files
JupyterLab

The exact version of every package in this environment is listed in /opt/ml/requirements.lock.txt.

It is yours to change

Install, upgrade or remove packages with pip as you like. The Studio runs from its own separate environment, so nothing you change here can break it.

Use a model you downloaded¶

The .joblib file downloads to your own computer. To use it on the instance, copy it back, either with scp:

scp -i <key>.pem <model>.joblib ubuntu@<public-ip>:~

or by uploading it through JupyterLab's file browser. Then follow Score a file in Python.

The Studio's service¶

The Studio runs as a system service named sushi-ml-studio, and starts by itself when the instance boots.

sudo systemctl status sushi-ml-studio      # is it running?
sudo systemctl restart sushi-ml-studio     # restart it
sudo journalctl -u sushi-ml-studio -f      # follow its log

The log also records failed sign-ins and passwords changed from the Studio, with the address they came from. Through a tunnel, the address shows as unknown.