Open wildintellect opened 1 year ago
Going to do this before #7 so that the conda yml can be used to pin versions.
We added conda yml for every notebook however without mamba the install via conda is painfully slow. For now, no plans to finish this until new conda/mamba is available in SMSL.
Just noting the correct syntax for using a yaml file to define a new conda env is:
conda env create -n <name> --file env_dem.yml
Sagemaker studio lab has a built in shortcut for quickly creating new
conda
environments as seen in https://github.com/aws/studio-lab-examples/blob/main/geospatial-data-science/NOAA_Exploratory_Analysis/EDA_weather_climate.ipynband described in https://github.com/aws/studio-lab-examples/tree/main/custom-environments
We should adopt this approach to keep each example isolated with it's own dependencies.
Acceptance criteria
conda
yml defining both conda installed libraries and pip required libraries when needed