In many machine learning projects you'd often find yourself using components that you have already built. If it's popular training algorithms like SVM, Neural Networks or any other scripts like data enrichment, preprocessing and more.
At cnvrg.io, we're making it easier for data scientists to store those components for reusability. Enabling teams to build a data base of solutions, algorithms, scripts and more.
Creating a Library
To create a new library, simply go to your Flows follow the instructions
- Click the add new task button, and fill-in the task details (command, environment, compute)
- Click the Publish as Library button and then click Save in the new popup
That's it, your task is now stored as a library and can be reused easily!
Now, when opening the new task dialog in Flows, you'll be able to see your tasks available for use:
- Who can see my libraries? Libraries inherit security permissions from the project they were created in.
- What kind of information is saved for every library? This library feature saves everything from command, parameters, compute and environment (docker image).
- Can I add documentation to my library? Sure! you can modify the documentation (right-pane) of the task on the right, so other data scientists will know more about the library.