Add a summary text to full text module

[david-thrower]

If we add a word2vec or tfidf input embedding to a Cerebros module and make 500 separate softmax outputs, 1 for each of 500 sequential words, then an argmax then a vocabulary lookup, we could train this on a small sample of pairs of summarized writing instructions and desired full text writing task results. This may make this do the same type of task as ChatGPT.

Also, doing the inverse operation may make a good summarization engine.