LDA.py

from gensim import corpora, models

# list_of_list_of_tokens = [["a","b","c"], ["d","e","f"]]
# ["a","b","c"] are the tokens of document 1, ["d","e","f"] are the tokens of document 2...
dictionary_LDA = corpora.Dictionary(list_of_list_of_tokens)
dictionary_LDA.filter_extremes(no_below=3)
corpus = [dictionary_LDA.doc2bow(list_of_tokens) for list_of_tokens in list_of_list_of_tokens]

num_topics = 20
%time lda_model = models.LdaModel(corpus, num_topics=num_topics, \
                                  id2word=dictionary_LDA, \
                                  passes=4, alpha=[0.01]*num_topics, \
                                  eta=[0.01]*len(dictionary_LDA.keys()))