Median model alone leading to low scores when submitted - probably due to nan replacement

[baogorek]

I'm learning a lot going through your code, and I was hoping to submit the median-based regression model alone to get a feel of how that works pre-keras. But last time I submitted it, I got a score of 190. Probably the reason is because, before submitting, I noticed that there were nans in test_all, so I filled them in with 0s. Now that I'm looking at the example values, I see patterns like nan, nan, 63.45, nan, so I probably shouldn't have done that! Will continue going through the later files to learn more, but if you're able to point me in the right direction, I'd appreciate it.

[jfpuget]

I think I filled all Nan with 0.