Stuck on "Building preproccessing pipelines"


Will not go past this using these settings:

and this:

Have tried a few things, output logs give no clues.

Welcome @Miner! Nice to have another new face round here :slight_smile:

I’m sure one of the PL guys will help out with a definitive answer, but one thing I know is that the train/val/test split relies on having a minimum number of samples (see forum here) However, if you don’t have enough I’d expect an error rather than constant waiting, but maybe worth checking you have enough.

And if you can share a bit more about your model dataset, setup etc. that could be helpful to them.

You can also run PerceptiLabs with the option -v=3 which will give more diagnostic info in the console… just in case that helps.

@robertl Suggestion - greater granularity in status messages possible? e.g. Step #x of N, and/or if there are steps that iterate on data, “Processing row/… of M?”

Hi @Miner,
Welcome to the forum!
I believe we spoke through intercom where we figured out that the Kernel had not started properly and therefore you got stuck in the eternal spinner hell (also known as “Building preprocessing pipelines”).

For anyone else passing by here, if your Data Wizard has “Select” in the bottom two dropboxes instead of a datatype such as “Image”, “Categorical”, etc, then that means that the Kernel has some issues. (We will add something in the frontend that tells you there is an error instead of continuing silently)
Easiest way to debug this is to run the tool as “perceptilabs -v=3” and look at the messages in the terminal or send them here in the forum :slight_smile:

Speaking of, if you didn’t get it to run, @Miner, we can pick it up here instead.

@JulianSMoore, the greater granularity of the message being worked on as we speak :wink:

1 Like