Saving language models

[larsgrobe]

Dear all,

I am wondering if noone ever wanted to persistently save a model for later reuse. Currently, the training functions (lda, nmf, elda) do not return the gensim-model itself, so one cannot directly use the gensim-functions to save and load models.

I could imagine two ways to allow for persistent models:

1. By making the gensim-model a member of TopicModel, and implementing .save() and .load() member function.
2. By adding a file-name parameter to the training functions. The functions could then test if the given file exists - if yes, it would be loaded and training by-passed, otherwise the file would be written after training.

Best, Lars.