Closed
Description
Option 1
Pack installation procedures of core Cortex dependencies into a script users can easily run when building custom Docker images
Example
FROM cortexlabs/python-predictor-cpu:0.26
# run a bunch of stuff for the creation of the custom image
# in the process, the python packages get removed (i.e. due to upgrading the Python runtime)
# reinstall core Cortex dependencies
RUN /usr/local/cortex/install-core-dependencies.sh
Option 2 (could be in place of or in addition to Option 1)
At runtime, detect that the cortex_internal package has been uninstalled, and reinstall it
Additional tasks
- Add docs for how to change the python version:
FROM quay.io/cortexlabs/python-predictor-gpu-slim:0.26.0
# update python runtime version
RUN conda install conda-forge::python=3.7
# and reinstall core cortex runtime dependencies
RUN pip --no-cache-dir install \
-r /src/cortex/serve/serve.requirements.txt \
/src/cortex/serve/