Fasttext: Purpose of -minn and -maxn for supervised training

[namp]

It is pretty clear what the purpose of the parameters minn and maxn is in unsupervised mode (e.g. skipgram) where vectors of sequences between minn and maxn characters are produced. However I have a difficulty in understanding the purpose of these parameters in supervised mode and especially how they are combined with wordNgrams>1

Judging by their default values in supervised mode which is 0 for both, I suspect they are only there for code compatibility purposes.

Thanks

[cpuhrsch]

Hello @namp,

Indeed the supervised training mode is not making used of subwords yet. Therefore these options have no effect.

Thanks,
Christian

[cpuhrsch]

Hello @namp,

This feature has been added as part of ebbd3bf.

Thanks,
Christian