There are warnings in the np.log calculations in ldacgsmulti:
/home/jammurdo/anaconda3/lib/python3.6/site-packages/vsm/mode/lldacgsmulti.py:435: RuntimeWarning: divide by zero encountered in log
log_kc = np.log(top_doc / top_doc.sum(0)[np.newaxis, :])
/home/jammurdo/anaconda3/lib/python3.6/site-packages/vsm/model/ldacgsmulti.py:461: RuntimeWarning: divide by zero encountered in log
mtrand_state[4])
/home/jammurdo/anaconda3/lib/python3.6/site-packages/vsm/model/ldacgsmulti.py:434: RuntimeWarning: divide by zero encountered in log
log_wk = np.log(gbl_word_top * inv_top_sums[np.newaxis, :])
Perhaps a consequence of using single-precision vectors, perhaps a consequence of an integer<->float miscast.
There are warnings in the np.log calculations in
ldacgsmulti
:Perhaps a consequence of using single-precision vectors, perhaps a consequence of an integer<->float miscast.