Update ./jupyterlab/environment.yml with your dependencies
Update the ports in docker-compose.yml
Build and run the container:
docker compose up jupyterlab
Choose the conda env kernel (not IPython) in your notebooks.
Before training and pickling a model:
-
Rebuild the jupyterlab container to get the most recent python modules:
docker compose up --build --no-start binaries docker compose up --build --no-start forge docker compose up --build jupyterlab
-
Get the versions of numpy, pandas, scikit-learn (and any other serialized dependencies) from inside the notebook:
!pip list
-
Pin the versions found in ./jupyterlab/environment.yml so that pickled objects can be deserialized correctly later.
- numpy=1.23.1
- pandas=1.4.3
- scikit-learn==1.1.2