Working in the Conda environment¶

Everything runs inside the rlbootcamp environment. This page is the muscle-memory you need — it is short on purpose.

Activate / deactivate¶

conda activate rlbootcamp     # start working
conda deactivate              # leave the environment

Your prompt shows (rlbootcamp) when it is active.

If you remember one thing

"It says the module isn't installed, but I installed it" is almost always a deactivated environment. Check the prompt first.

Run a single command without activating¶

Handy in scripts, IDEs, cron jobs, or when you simply forget:

conda run -n rlbootcamp python scripts/smoke_test.py
conda run -n rlbootcamp tensorboard --logdir runs/

Confirm which Python you're using

conda run -n rlbootcamp python -c "import sys; print(sys.executable)"

The path must contain envs/rlbootcamp. If it does not, you are running a different interpreter and that explains your import errors.

Keep the environment up to date¶

When environment.yml changes, refresh rather than recreate:

git pull
conda env update -f environment.yml --prune

--prune removes packages that were dropped from the file.

Add a package for an experiment¶

conda activate rlbootcamp
pip install <package>

Mixing pip and conda inside one environment is fine as long as you install conda packages first and pip packages second — which is what environment.yml already does. Avoid conda install after heavy pip installs; the solver can downgrade things underneath pip's feet.

Use it in Jupyter¶

JupyterLab and ipykernel ship with the environment. Register it as a kernel once, so notebooks can find it:

conda activate rlbootcamp
python -m ipykernel install --user --name rlbootcamp --display-name "Python (rlbootcamp)"
jupyter lab

Then pick the Python (rlbootcamp) kernel in the notebook's kernel menu.

The notebook kernel is a separate choice

Activating the environment in your terminal does not change which kernel an already-open notebook uses. If imports fail inside a notebook but work in the terminal, you are on the wrong kernel. Check the top-right of the notebook.

Use it in VS Code / PyCharm¶

The interpreter and the notebook kernel are separate settings in both IDEs, which is where most of the confusion comes from:

Start over (last resort)¶

Recreating is cheap and fixes a surprising number of problems:

conda deactivate
conda env remove -n rlbootcamp
conda env create -f environment.yml

Next: Troubleshooting →