lognet crashes when there are less than three values of lambda

[miguelcorrea]

I have noticed that sometimes, when glmnet has problems reaching convergence, lognet will crash with the error

File "/usr/lib/python3.6/site-packages/glmnet_py/lognet.py", line 257, in lognet t2 = scipy.log(alm[2]) IndexError: index 2 is out of bounds for axis 0 with size 2

This occurs only when there are less than three values of lambda (this does not require forcing glmnet to do so!). If we look at lognet.py, from line 255, we see that it is trying to retrieve a non-existing value of lambda, and therefore the crash.
if ulam[0] == 0.0: t1 = scipy.log(alm[1]) t2 = scipy.log(alm[2]) alm[0] = scipy.exp(2*t1 - t2)
I have tested the behaviour of the R version of glmnet with the same data, and it simply leaves the first value of the lambda sequence (here alm[0]) as it is, so I suppose that would be acceptable behaviour here.

[pancodia]

I encountered the same issue. My case is more extreme that I only have one value in ˋalmˋ

IndexError: index 1 is out of bounds for axis 0 with size 1

So I have two questions:

1. How does the algorithm choose the values for ˋlambdauˋ ?
2. Besides the solution mentioned by @miguelcorrea , can I just explicit provide a list of ˋlambdauˋ as a walk around of this issue?