Strange output for TopTokensCoherenceScore

[DrozdikGleb]

I've created 18 topics for my model and used TopTokensCoherenceScore. When I call artm_model.score_tracker['TopTokensCoherenceScore'].average_coherence it gives me correct result for each collection pass, but when I call artm_model.score_tracker['TopTokensCoherenceScore'].coherence (or coherence[-1] for last collection pass) it gives me output in following format:

{'topic_1': [1.1617621183395386,
  1.1602195501327515,
  0.3456096351146698,
  0.9662702679634094,
  0.5320848226547241],
 'topic_2': [0.9413328170776367,
  0.3860290050506592,
  1.2057011127471924,
  1.1175670623779297,
  0.37220320105552673],
 'topic_3': [0.28798189759254456,
  1.1940522193908691,
  0.6581138372421265,
  0.42109599709510803,
  1.9474769830703735],
 'topic_4': [0.5959908366203308, 0.7645043134689331, 0.22957539558410645]}

I suppose correct format should be looks like this

{'topic_1': 1.1617621183395386,
 'topic_2':1.1602195501327515,
 'topic_3':0.3456096351146698,
 'topic_4':0.9662702679634094,
 'topic_5':0.5320848226547241,
 ...
 'topic_18':0.22957539558410645
 }

P.S. bigartm version - 0.10.1

[JeanPaulShapo]

Sorry for late response.
It seems that there was incorrect use of topic_name field for labeling coherence scores
Right now bugfix for this is merged into master (see #1019 )
Could you please check that library built from current master branch fixes format?