# Environment setup on the SciNet Teach Cluster

Portal: <https://ondemand-teach.scinet.utoronto.ca/pun/sys/dashboard>

Open **Interactive Apps → JupyterLab**, request one core and 4 GB of memory,
then open a terminal in JupyterLab and run:

```bash
module load python/3.11.5 scipy-stack/2026a
python3 -m venv --system-site-packages ~/mlworkshop-env
source ~/mlworkshop-env/bin/activate
pip install --no-index --upgrade pip
pip install --no-index scikit-learn
python -m ipykernel install --user --name mlworkshop --display-name "ML Workshop (Python 3.11)"
```

The SciNet scientific packages are supplied by a module. Make the notebook
kernel load that module whenever it starts:

```bash
KD=~/.local/share/jupyter/kernels/mlworkshop
printf '#!/bin/bash\nmodule load python/3.11.5 scipy-stack/2026a\nsource "$HOME/mlworkshop-env/bin/activate"\nexec python -m ipykernel_launcher "$@"\n' > $KD/kernel_wrapper.sh
chmod +x $KD/kernel_wrapper.sh
python3 -c "import json,os; p=os.path.expanduser('$KD/kernel.json'); k=json.load(open(p)); k['argv']=[os.path.expanduser('$KD/kernel_wrapper.sh'),'-f','{connection_file}']; json.dump(k,open(p,'w'),indent=1)"
```

Select **ML Workshop (Python 3.11)** as the notebook kernel. Check it with:

```python
import sys, numpy, pandas, matplotlib, sklearn
print(sys.executable)  # should contain mlworkshop-env
```

If an import fails, first check that the notebook is using this kernel.