Use of utils.py

Spot the Trend
1

Hi,
Since it’s mandatory to use the classes and functions provided in utils.py, and I want to apply a preprocessing step on the data before storing in DataObject class.

Should I put the preprocessing function inside the utils.py file and call it inside utils.py when I need it. Or should I put preprocessing function in the final notebook?

Also, can we modify the utils.py file by adding new functions or modifying existing functions or classes like the DataObject class?

thanks

2

Hi @modd42 ,

Using the classes/functions in utils.py is not strictly mandatory; they are provided to assist participants in processing data and ensuring their predictions and submission files are in the correct format. However, adhering to the required submission format is mandatory. Your submission must be a JSON file with the structure specified in the Starter Notebook. To be correctly evaluated for the Predictive Leaderboard, your submission must be in the proper format and structure. Additionally, if you are asked to submit your Notebook as one of the top Predictive Leaderboard participants, your model’s inference function must save the predicted intervals in the same correct JSON format when run on the holdout dataset. You are allowed to modify the classes/functions in utils.py and add new ones as long as your code outputs a correctly formatted JSON submission file.

Best of luck Spotting the Trend!
ThinkOnward Team

3

I thought it was mandatory because of this paragraph in the notebook:

" Code for classes and functions are available in the utils.py file. For your solution to be accepted, you should use these classes and function. In particular, you should pay attention to the generation of submissions."

4

@modd42,

As mentioned in our first response, it is not mandatory as long as your code outputs correctly formatted JSON submission files. However, we strongly recommend using them to avoid any errors or misevaluation of your submissions (specifically when run on the holdout dataset). If your code does not output a correctly formatted and structured JSON file when run on the holdout dataset (as well as on the training dataset when checking your submission’s reproducibility) resulting in an error, your submission will be discarded.

Best of luck Spotting the Trend!
ThinkOnward Team

5

Thank you for clarifying this