Would be a child of the cdl-jupyter image with pinned versions of some basic data science-/analysis-related packages. Essentially, it would include the data science packages we actually routinely use out of the behemoth >700MB stack shipped with Anaconda (e.g., no Spyder, and TensorFlow can be optional)
This is what cdl-jupyter was originally going to be, but ultimately a notebooks-only version seemed more useful since it would A) allow us to distribute an image with a near-arbitrary combination of packages and versions, and zero bloat due to unused default installs as well as B) simply serve as an intermediate build stage between cdl-python and the full cdl-datasci image, further reducing the final image's build time.
cdl-datasci would serve as a great jumping off point for creating a brand new environment to run analyses for a new project, as well as a perfect resource for quick hackathon-style events (e.g. MIND) where it's more important to have everyone download the same working, reasonably up-to-date packages and get going rather than agonize over which minor/micro version of sqlalchemy to choose.
Would be a child of the
cdl-jupyterimage with pinned versions of some basic data science-/analysis-related packages. Essentially, it would include the data science packages we actually routinely use out of the behemoth >700MB stack shipped with Anaconda (e.g., no Spyder, and TensorFlow can be optional)This is what
cdl-jupyterwas originally going to be, but ultimately a notebooks-only version seemed more useful since it would A) allow us to distribute an image with a near-arbitrary combination of packages and versions, and zero bloat due to unused default installs as well as B) simply serve as an intermediate build stage betweencdl-pythonand the fullcdl-datasciimage, further reducing the final image's build time.cdl-datasciwould serve as a great jumping off point for creating a brand new environment to run analyses for a new project, as well as a perfect resource for quick hackathon-style events (e.g. MIND) where it's more important to have everyone download the same working, reasonably up-to-date packages and get going rather than agonize over which minor/micro version ofsqlalchemyto choose.