Distributed and multicore LDA print messages saying LDA is being run serially

[ghost]

There is some code in ldamodel.py that passes distributed=False when the user specified distributed=True. This causes the confusing error message that LDA is running serially on that node to be printed.

[ghost]

I have to say, these messages are quite confusing, especially since only one core is being used.

2015-07-01 00:33:08,859 : INFO : registering worker #1 at PYRO:gensim.lda_worker.63bd2@127.0.0.1:35365
2015-07-01 00:33:08,961 : INFO : initializing worker #1
2015-07-01 00:33:08,965 : INFO : using symmetric alpha at 0.0001
2015-07-01 00:33:08,966 : INFO : using serial LDA version on this node
2015-07-01 00:36:31,625 : INFO : registering worker #2 at PYRO:gensim.lda_worker.3ea807@127.0.0.1:40761
2015-07-01 00:36:31,725 : INFO : initializing worker #2
2015-07-01 00:36:31,728 : INFO : using symmetric alpha at 0.0001
2015-07-01 00:36:31,730 : INFO : using serial LDA version on this node
2015-07-01 00:39:54,802 : INFO : registering worker #3 at PYRO:gensim.lda_worker.52cba6@127.0.0.1:34256
2015-07-01 00:39:54,900 : INFO : initializing worker #3
2015-07-01 00:39:54,904 : INFO : using symmetric alpha at 0.0001
2015-07-01 00:39:54,905 : INFO : using serial LDA version on this node

[piskvorky]

Only the master node is distributed; all other nodes (workers) run serial computations.

The flow is simple: master prepares and sends out tasks to workers, workers do work and send results back to master, master aggregates them into the LDA model.