The tailor generation is not working

[RuiYang12345]

The tailor generation is not working

[Ghadjeres]

Hi @RuiYang12345 ,
Thanks a lot for your detailed issue!
The MuseScore plugin was indeed doing what you described. In fact I hardly checked this part of the code when rewriting it in Pytorch. Sorry for this.
Finally I solved this issue in #74 :
Now:

• The plugin works in MuseScore3
• It now takes into account fermatas (this influences generation!)
• It is now longer bound to generate in A minor / C major. Instead, it evaluates the most probable keys in the file that your provide.
• Part of the communication between the server and the client is done by writing on /tmp/deepbach.xml and /tmp/deepbach.mxl (this is not entirely satisfying but kind of works, you need to have unzip installed...).
• If you click compose with no selection, a novel chorale is generated from scratch. This is useful for instance if you want to start with a blank one-staff score.
• The number of iterations now depends on the length of the region to be recomposed.
• I also removed all the logging messages and the possibility to choose a server different from localhost:5000. You can easily change this in the .qml file by putting visible=true in some elements and manually setting the serverAddresss variable at the beginning of the file.

Remember that you can provide a --num_iterations argument to the musescore_flask_server.py if you think the default value takes too much time.
Hope it will work on your side. Also do not forget to reload the plugins in MuseScore3: just replacing the .qml file won't work.

And here is a small chorale composed with this plugin!
https://www.youtube.com/watch?v=X72YHyaTIUE&feature=youtu.be

[RuiYang12345]

Hi Ghadjeres,

It is working now, thank you so much!