Train models by publishing "model training" like for transforms?

Ask the Community
1

In Code workspace, in a Jupyter code notebook, I can publish a given dataset output as a “Transform output”, which makes my python code to execute like a normal transform, in the background, to produce the output dataset that I want.

Does the same exist for model training ?
Can I publish a model training as a “transform” and run that in the background instead of in my user’s code workspace instance ?

Also, as that’s the goal, if the data volume is too high and the build too slow, should I use code repository instead ?

2

Hey @VincentF, it is currently not possible to publish a model as a transform output in code workspaces. We are working on adding support for this, and it should be available in the next few weeks.

In terms of data volume being too high, its really up to how much compute resources you allocate to the code workspace.

3

Posting back, this is now supported! See the documentation for details

1 Like