首先源碼參考
http://blog.csdn.net/u014365862/article/details/53868544
原文實現了隨機產生詩句或者給定詩句首個字符，訓練時也沒用到題目信息，現在我加以改進實現了用題目來得到詩句的代碼。
改進思想：詩句的題目在樣本中最多只有26個字，我們就填充空格標記到26的長度，然后再進行embedding，最后用（26x128）X128的矩陣映射到128維度，放在詩句起始標記[的最前面，這樣rnn就會多一個題目信息，長度增加1。
import collections import numpy as np import tensorflow as tf #-------------------------------數據預處理---------------------------# poetry_file ='poetry.txt' # 詩集 poetrys = [] titles=[] titletopoetry={} with open(poetry_file, "r", encoding='utf-8',) as f: for line in f: try: title, content = line.strip().split(':') content = content.replace(' ','') if '_' in content or '(' in content or '（' in content or '《' in content or '[' in content: continue if len(content) < 5 or len(content) > 79: continue content = '[' + content + ']' poetrys.append(content) titles.append(title) # titletopoetry[title]=content; except Exception as e: pass # 按詩的字數排序 poetrys = sorted(poetrys,key=lambda line: len(line)) print('唐詩總數: ', len(poetrys)) # 統計每個字出現次數 all_words = [] for poetry in poetrys: all_words += [word for word in poetry] for poetry in titles: all_words += [word for word in poetry] counter = collections.Counter(all_words) count_pairs = sorted(counter.items(), key=lambda x: -x[1]) words, _ = zip(*count_pairs) # 取前多少個常用字 words = words[:len(words)] + (' ',) # 每個字映射為一個數字ID word_num_map = dict(zip(words, range(len(words)))) # 把詩轉換為向量形式，參考TensorFlow練習1 to_num = lambda word: word_num_map.get(word, len(words)) poetrys_vector = [ list(map(to_num, poetry)) for poetry in poetrys] #[[314, 3199, 367, 1556, 26, 179, 680, 0, 3199, 41, 506, 40, 151, 4, 98, 1], #[339, 3, 133, 31, 302, 653, 512, 0, 37, 148, 294, 25, 54, 833, 3, 1, 965, 1315, 377, 1700, 562, 21, 37, 0, 2, 1253, 21, 36, 264, 877, 809, 1] #....] titles_vector = [ list(map(to_num, poetry)) for poetry in titles] maxlength = max(map(len,titles_vector)) # 每次取64首詩進行訓練 batch_size = 64 n_chunk = len(poetrys_vector) // batch_size x_batches = [] y_batches = [] xtitle=[] for i in range(n_chunk): start_index = i * batch_size end_index = start_index + batch_size tdata = np.full((batch_size,maxlength), word_num_map[' '], np.int32) tbatches=titles_vector[start_index:end_index] batches = poetrys_vector[start_index:end_index] for row in range(batch_size): tdata[row,:len(tbatches[row])] = tbatches[row] length = max(map(len,batches)) xdata = np.full((batch_size,length), word_num_map[' '], np.int32) for row in range(batch_size): xdata[row,:len(batches[row])] = batches[row] ydata = np.full((batch_size,length+1), word_num_map[' '], np.int32) ydata[:,0]=word_num_map['['] ydata[:,1:-1] = xdata[:,1:] """ xdata ydata [6,2,4,6,9] [2,4,6,9,9] [1,4,2,8,5] [4,2,8,5,5] """ x_batches.append(xdata) y_batches.append(ydata) xtitle.append(tdata) batch_size = 1 #生成詩句才用到，訓練時注釋掉 #---------------------------------------RNN--------------------------------------# tf.reset_default_graph() input_data = tf.placeholder(tf.int32, [batch_size, None]) input_tdata = tf.placeholder(tf.int32, [batch_size, None]) output_targets = tf.placeholder(tf.int32, [batch_size, None]) # 定義RNN def neural_network(model='lstm', rnn_size=128, num_layers=2): if model == 'rnn': cell_fun = tf.nn.rnn_cell.BasicRNNCell elif model == 'gru': cell_fun = tf.nn.rnn_cell.GRUCell elif model == 'lstm': cell_fun = tf.nn.rnn_cell.BasicLSTMCell cell = cell_fun(rnn_size, state_is_tuple=True) cell = tf.nn.rnn_cell.MultiRNNCell([cell] * num_layers, state_is_tuple=True) initial_state = cell.zero_state(batch_size, tf.float32) with tf.variable_scope('rnnlm'): softmax_w = tf.get_variable("softmax_w", [rnn_size, len(words)+1]) softmax_b = tf.get_variable("softmax_b", [len(words)+1]) softmax_wt = tf.get_variable("softmax_wt", [maxlength*rnn_size, rnn_size]) softmax_bt = tf.get_variable("softmax_bt", [rnn_size]) with tf.device("/gpu:0"): embedding = tf.get_variable("embedding", [len(words)+1, rnn_size]) inputs = tf.nn.embedding_lookup(embedding, input_data) tdatas=tf.nn.embedding_lookup(embedding, input_tdata) midresult=tf.reshape(tdatas,[-1,maxlength*rnn_size]) mid=tf.matmul(midresult,softmax_wt)+softmax_bt mid1=tf.reshape(mid,[batch_size,1,rnn_size]) newinputs=tf.concat([mid1,inputs],1) outputs, last_state = tf.nn.dynamic_rnn(cell, newinputs, initial_state=initial_state, scope='rnnlm') output = tf.reshape(outputs,[-1, rnn_size]) logits = tf.matmul(output, softmax_w) + softmax_b probs = tf.nn.softmax(logits) return logits, last_state, probs, cell, initial_state,newinputs,inputs #訓練 def train_neural_network(): logits, last_state, _, _, _,_,_ = neural_network() targets = tf.reshape(output_targets, [-1]) loss = tf.contrib.legacy_seq2seq.sequence_loss_by_example([logits], [targets], [tf.ones_like(targets, dtype=tf.float32)], len(words)) cost = tf.reduce_mean(loss) learning_rate = tf.Variable(0.0, trainable=False) tvars = tf.trainable_variables() grads, _ = tf.clip_by_global_norm(tf.gradients(cost, tvars), 5) optimizer = tf.train.AdamOptimizer(learning_rate) train_op = optimizer.apply_gradients(zip(grads, tvars)) with tf.Session() as sess: sess.run(tf.initialize_all_variables()) saver = tf.train.Saver(tf.all_variables()) for epoch in range(50): sess.run(tf.assign(learning_rate, 0.002 * (0.97 ** epoch))) n = 0 for batche in range(n_chunk): train_loss, _ , _ = sess.run([cost, last_state, train_op], feed_dict={input_data: x_batches[n], output_targets: y_batches[n],input_tdata:xtitle[n]}) n += 1 print(epoch, batche, train_loss) if epoch % 7 == 0: saver.save(sess, './my1/model.ckpt', global_step=epoch) train_neural_network()#訓練時用到，生成詩句時注釋掉 def gen_poetry(begin): def to_word(weights): t = np.cumsum(weights) s = np.sum(weights) sample = int(np.searchsorted(t, np.random.rand(1)*s)) return words[sample] _, last_state, probs, cell, initial_state,newinputs,inputs = neural_network() with tf.Session() as sess: sess.run(tf.initialize_all_variables()) saver = tf.train.Saver(tf.all_variables()) saver.restore(sess, './my1/model.ckpt-49') state_ = sess.run(cell.zero_state(1, tf.float32)) a=[a for a in begin] # print(map(word_num_map.get, a)) x = np.array([list(map(word_num_map.get, '['))]) # [probs_, state_] = sess.run([probs, last_state], feed_dict={input_data: x, initial_state: state_}) x1 = np.array([list(map(word_num_map.get, a))]) print(x.shape) y=np.full([1,maxlength],word_num_map[' ']) y[0,0:x1.shape[1]]=x1[0,:]; [probs_, state_] = sess.run([probs, last_state], feed_dict={input_data: x,input_tdata: y, initial_state: state_}) print(to_word(probs_[0])) word = to_word(probs_[1]) #word = words[np.argmax(probs_)] poem = "" while word != ']': poem += word x = np.zeros((1,1)) x[0,0] = word_num_map[word] m=inputs.eval(session=sess,feed_dict={input_data: x}) [probs_, state_] = sess.run([probs, last_state], feed_dict={newinputs:m,initial_state: state_}) word = to_word(probs_) #word = words[np.argmax(probs_)] return poem def gen_poetry_with_head(head): def to_word(weights): t = np.cumsum(weights) s = np.sum(weights) sample = int(np.searchsorted(t, np.random.rand(1)*s)) return words[sample] _, last_state, probs, cell, initial_state = neural_network() with tf.Session() as sess: sess.run(tf.initialize_all_variables()) saver = tf.train.Saver(tf.all_variables()) saver.restore(sess, './my/model.ckpt-49') state_ = sess.run(cell.zero_state(1, tf.float32)) poem = '' i = 0 for word in head: while word != '，' and word != '。': poem += word x = np.array([list(map(word_num_map.get, word))]) [probs_, state_] = sess.run([probs, last_state], feed_dict={input_data: x, initial_state: state_}) word = to_word(probs_) # time.sleep(1) if i % 2 == 0: poem += '，' else: poem += '。' i += 1 return poem #print(gen_poetry_with_head('一二三四')) str=input('輸入詩的題目：') print(gen_poetry(str))#生成詩句用到，訓練時注釋掉
資源文件在上面提到的的網址有，標綠的是自己修改的部分，下面是一些生成句子的示例：
輸入詩的題目：美女
鴛鴦不建玄，帝里初有無。久令長不動，坐是芳草音。官中花當飲，過鳥敲釵出。水蠹澄無溢，枝盤鴨珊珊。遙憐書帙服，簾外色聲疏。何止分離心，何人還不主。
輸入詩的題目：岳陽樓記
歲晚終如鏡，誰憐利味殊。已將張卜闕，無我更忘言。書生知更盡，際海本無囂。幾葉至十年，何當休奈蒸。魯臣怒抵象，願玩為公奏。少年為玉耳，簪艷在微中。
輸入詩的題目：歸鄉
西國應堪再，承閑騎紫微。豈知玄豹在，不解信仙郎。常設功殊系，登仁又作持。攀年招葯碗，作怕柏蟲閑。昨來刀骨厚，急拂青蓮低。唯有攜僧者，榮性不前尋。