Thanks @jyotish. Following the above steps, the zip file submission now works.
Something I noticed from downloading my Colab submission was that there seems to be a reversal between claims and prices at the end of the predict.R file:
claims = predict_premium(trained_model, Xraw)
Not sure if this would have been a problem I introduced or if it related to the template linked to the submit.sh utility.
On another note, leaderboard seems to be based on my worst submission (id 110383) instead of the best or most recent ( 110384, 110389)
The important thing in the script is that we invoke predict_premium function during the weekly leaderboard evaluation and predict_expected_claim for RMSE leaderboard evaluation. We will fix the variables names.
I thought that was rather cruel when I saw it! If only all the others that are rapidly catching my RMSE score could oblige and do the same. It would take the heat off me to find a better submission!
We will look into this, but for now you should be able to install and use it in your model regardless by filling out the install.R file for a zip submission or the install_packages function for a submission through the notebooks.
