generate sequence from LSTM in Keras: attempt from https://www.reddit.com/r/MachineLearning/comments/3dqdqr/keras_lstm_limitations/

genseq.py

complete_sentences = [["*-START-*"] for a in range(1000)]
sents = np.zeros((nb_samples, timesteps+1, len(vocab)))

for x in range(nb_samples):
    sents[i,0,word2index["*-START-*"]] = 1. # init the sequences

for t in range(timesteps):
    preds = self.model.predict(sents[:,0:t+1], verbose=0)
    # get the maximum predictions for this timestep for each sample
    next_word_indices = np.argmax(preds[:,t], axis=1)
    # fill in the input at the next timestep with the prediction from this timestep
    for i in range(nb_samples):
        sents[i, t+1, next_word_indices[i]] = 1. 
    next_words = [index2word[x] for x in next_word_indices]
    for i in range(len(next_words)):
        complete_sentences[i].append(next_words[i])