I'm trying to implement LDA in either R or Python on a production system. I'd prefer Python because more of the team is comfortable with it, but either are an option.
In short, R's text2vec package implements WarpLDA (from this 2016 paper) and performs more stable clustering and much faster than either popular Python option (gensim and sklearn) which both implement Online Variational Bayers (from this 2010 paper).
Does anyone know of any stable, well-documented Python implementations of WarpLDA? I would love to be able to compare it to the R, so that I can get more of an apples-to-apples comparison.
Also, I'd be interested to know of any Python implementations of LightLDA or F+LDA, which are the two "state-of-the-art" algorithms that the aforementioned 2016 paper benchmarks itself against.