When I ran the Chi2 divergence in HW2 Ex 7, I got the error:
ValueError: For each axis slice, the sum of the observed frequencies must agree with the sum of the expected frequencies to a relative tolerance of 1e-08, but the percent differences are:
0.25315591948140564
Seems related to this issue: https://github.com/scipy/scipy/issues/14298. Looks like some normalization of the inputs is needed, but I am not sure how since we are passing two token lists—should we instead use probs arrays? Or should we instead use the other three divergence measures? Thanks!
When I ran the Chi2 divergence in HW2 Ex 7, I got the error:
Seems related to this issue: https://github.com/scipy/scipy/issues/14298. Looks like some normalization of the inputs is needed, but I am not sure how since we are passing two token lists—should we instead use probs arrays? Or should we instead use the other three divergence measures? Thanks!