All LDA topic coefficients are 'nan'

[Lukas Toma]

Hi,

I followed this tutorial: https://markroxor.github.io/gensim/static/notebooks/lda_training_tips.html

With different data but exactly in the same format (or at least I cannot spot the difference)

top_topics = ldamodel.top_topics(corpus=corpus)

# Average topic coherence is the sum of topic coherences of all topics, divided by the number of topics.

avg_topic_coherence = sum([t[1] for t in top_topics]) / 5

print('Average topic coherence: %.4f.' % avg_topic_coherence)

from pprint import pprint

pprint(top_topics)

Average topic coherence: -10.6956.

[([(nan, u'imported'),
 (nan, u'soon'),
 (nan, u'privileged'),
 (nan, u'authorized'),
 (nan, u'specific'),
 (nan, u'uploaded'),
 (nan, u'twice'),
 (nan, u'touch'),
 (nan, u'teammate'),
 (nan, u'video'),
 (nan, u'product'),
 (nan, u'agency'),
 (nan, u'p'),
 (nan, u'standard'),
 (nan, u'responsibility'),

ANy ideas what I might be doing wrong?

Thanks

[Ivan Menshikh]

Hi Lukas,

this looks like a bug, can you create issue on github and add info about your gensim/numpy/scipy/python/os version?

[Brijesh Singh]

I am facing the same issue can you help me to solve, HDPModel and LSIModel is giving the proper result as in list of topic but LDAModel is giving with nan*

version-

gensim: 3.4.0

numpy: 1.14.3

scipy : 1.0.0

python : 3.6.4

[Radim Řehůřek]

Hi Brijesh,

can you try with an older version of gensim? Say 0.10.1.

I'd like to rule out any recent changes.

Thanks,

Radim

[Nate TeBlunthuis]

Greetings everyone,

I also appear to have this issue with LdaModel. I am using the same versions of gensim and numpy as the report, but python 3.6.3. For me the issue appears following a series of numpy runtime warnings on lines 446, 546, 103, 802. If I reduce the number of topics the problem goes away. I also tried gensim version 3.1 and did not have the problem (i think I noticed the problem after upgrading from 3.1. to 3.4). Hope that helps things narrow it down. 

My dataset is quite big so it would not be convenient to share. I can try to reproduce the issue on a smaller dataset if that would be helpful.

[Greg Chase]

I've also run into the same issue with the LDA model. It's been reproduced on Gensim 3.1.0 and 3.5, between 1 and 10 topics. 

Current software versions are below.

• Python 3.6.4
  
• Gensim 3.1.0
  
• NumPy 1.14.5
  
• SciPy 1.1.0
  

[markin...@gmail.com]

I solve it by upgrading numpy to 1.15.0 version.

пятница, 15 декабря 2017 г., 8:38:49 UTC+2 пользователь Lukas Toma написал: