search

pyannote / pyannote-metrics

A toolkit for reproducible evaluation, diagnostic, and error analysis of speaker diarization systems
http://pyannote.github.io/pyannote-metrics
MIT License
184 stars 32 forks source link

Python 3.5 results are not same to Python 2.7 #10

Closed NeoBoy closed 7 years ago

NeoBoy commented 7 years ago

image I ran the example notebook and results for Cluster purity and coverage are not same, a warning (utils.py:160: UserWarning: 'uem' was approximated by the union of 'reference' and 'hypothesis' extents. "'uem' was approximated by the union of 'reference' ") is generated

hbredin commented 7 years ago

A screenshot is not very helpful. Can you please provide me with a simple Python script that reproduce the error (expected output, actual output).

NeoBoy commented 7 years ago

Sorry for that, I am running the example file by you. I just changed the print statements which can't directly run in Python 3 by adding ( )

NeoBoy commented 7 years ago

No other change in the code

hbredin commented 7 years ago

These notebooks have not been updated for a while. Can you check with the latest version I just released?

Also, do you confirm you ran both Python 2.7 and Python 3.5 and did not get the same results between your two runs?

NeoBoy commented 7 years ago

I ran the code on python 3.5 and compared the results with the ones already in the notebook which clearly was python 2 (evident from print statement used as a keyword instead of a function). My Python 2.7 installation was giving import issue on the notebook so it didn't run

NeoBoy commented 7 years ago

Seen the current notebook, the results are now same to what I am getting in Python 3.5 apart from the userWarning