FAQs

Frequently Asked Questions

Can I train a model locally?

Yes, but you should still include the code in case we need to rerun your work.

We understand that training some models requires significant resources and computing time, which cannot easily fit within the weekly quota. Some models even require a network connection that is not available in the cloud environment.

For these models, we recommend training locally but including the training code in a separate file or commenting it out.

If you do not submit your training code, you might miss the opportunity to retrain with more data during the Out-of-Sample Phase, as we have done in past competitions.

SCIPY_ARRAY_API

As this flag introduces changes that could be changed in the future and was only intended as a temporary measure, you need to enable it manually.

To achieve this, insert this cell before the one that imports SciPy for the first time.

Python Notebook Cell

import os
os.environ["SCIPY_ARRAY_API"] = "1"

How can I get the output of a Run?

Files generated in the cloud environment (e.g., trained models) cannot be extracted or downloaded locally.

This ensures that data accessible only there cannot be exfiltrated.

I get a NameError on my constant

If you receive the following error:

NameError: name ‘MY_CONSTANT’ is not defined

This happened because your constant was commented out during the submission process. You can avoid this by doing one of the following:

• Use the # @crunch/keep:on and # @crunch/keep:off commands to wrap the declaration of your constant:

Python Notebook Cell

# @crunch/keep:on
MY_CONSTANT = true
# @crunch/keep:off

if MY_CONSTANT:
    ...

• Move your constants into a dedicated cell, then place # @crunch/keep:on at the very top:

Python Notebook Cell

# @crunch/keep:on
MY_CONSTANT = true

Python Notebook Cell

if MY_CONSTANT:
    ...

• Declare them in a class:

Python Notebook Cell

class Constants:
    MY_CONSTANT = true

if Constants.MY_CONSTANT:
    ...

For more details, see the Notebook Processor page.