lasso_missing model fails

[fonnesbeck]

The lasso_missing.py model does not produce expected output when run under current master:

$ python lasso_missing.py
Warning: Desired error not necessarily achieved due to precision loss.
         Current function value: 10000000000000000159028911097599180468360808563945281389781327557747838772170381060813469985856815104.000000
         Iterations: 0
         Function evaluations: 2
         Gradient evaluations: 1
Traceback (most recent call last):
  File "lasso_missing.py", line 43, in <module>
    start = pm.find_MAP()
  File "/Users/fonnescj/Repos/pymc3/pymc3/tuning/starting.py", line 153, in find_MAP
    specific_errors)
ValueError: Optimization error: max, logp or dlogp at max have non-finite values. Some values may be outside of distribution support. max: {'sib_mean_log_': array(-0.3665129244327545), 'siblings_imp_missing': array([], dtype=int64), 'p_disab_logodds_': array(0.0), 'disability_imp_missing': array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]), 'p_mother_logodds_': array(0.0), 'mother_imp_missing': array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0]), 's_log_': array(1.6094379425048828), 'beta': array([ 0.1,  0.1,  0.1,  0.1,  0.1,  0.1,  0.1])} logp: array(nan) dlogp: array([  8.98253870e+01,  -5.85000000e+01,  -9.50000000e+00,
         5.58700682e+04,   6.49721961e+02,   3.32593980e+02,
         7.25313956e+02,   1.26741992e+02,   3.48175559e+04,
         3.19725981e+02,   2.59157984e+02])Check that 1) you don't have hierarchical parameters, these will lead to points with infinite density. 2) your distribution logp's are properly specified. Specific issues: 
beta.logp bad: nan

[springcoil]

Ohh interesting. So I'm currently doing git bisect to see if I can diagnose this.
And this is the result I get.
[6cc0d0ab2daa7954cdecb9c2923f4479dbc18fea] Merge pull request #1685 from aseyboldt/fix-sparse

I wonder what happened in that commit.

[twiecki]

@aseyboldt cc

[junpenglao]

fixed in #2254