inconsistent Response curves

[ohad-monday]

Getting this weird behaviour on the Demo data - was running the demo end to end example 2 times (as is in 1st run and in the 2nd run with different splitting point - both in the same jupyter kernel)
split_point = data_size - 10 (instead of split_point = data_size - 13)
getting this for 1st run

and this for 2nd run

what could be the problem? could be something to do with the seed? ot something else im missing?
btw, when i restart the kernel and run the #2 option first, i get similar results to original #1 run

[michevan]

Hi ohad-monday! I'm actually not sure what the problem is here, is it that the response curves look slightly different between the two runs (especially channel 0)? If so, I don't think that is actually a major concern. This looks to me like about the sort of change you can expect by slightly varying your input data. I would expect a geo-level model to be a bit more stable, for what it's worth.