tensormsa
tensormsa copied to clipboard
early stopping for nlp logics
need to stop train when nlp networks (seq2seq, word2vec, doc2vec) reaches certain goal to prevent overfitting and save cpu powers