利用循环神经网络RNN生成唐诗代码和数据集（数据集可少一点）

以下是一个简单的利用循环神经网络生成唐诗的代码示例：

import numpy as np
import tensorflow as tf

# 加载唐诗数据集
data = open('tangshi.txt', 'r').read()
chars = list(set(data))
data_size, vocab_size = len(data), len(chars)
char_to_ix = { ch:i for i,ch in enumerate(chars) }
ix_to_char = { i:ch for i,ch in enumerate(chars) }

# 构建模型
hidden_size = 100
seq_length = 25
learning_rate = 1e-1

inputs = tf.placeholder(shape=[None, seq_length, vocab_size], dtype=tf.float32)
targets = tf.placeholder(shape=[None, vocab_size], dtype=tf.float32)

cell = tf.nn.rnn_cell.BasicRNNCell(hidden_size)
initial_state = cell.zero_state(batch_size=tf.shape(inputs)[0], dtype=tf.float32)

outputs, states = tf.nn.dynamic_rnn(cell, inputs, initial_state=initial_state, dtype=tf.float32)

weights = tf.Variable(tf.random_normal([hidden_size, vocab_size]))
biases = tf.Variable(tf.random_normal([vocab_size]))
logits = tf.matmul(outputs[:, -1, :], weights) + biases
predictions = tf.nn.softmax(logits)

loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=targets, logits=logits))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(loss)

# 训练模型
num_epochs = 50
batch_size = 128

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())

    for epoch in range(num_epochs):
        num_batches = int(data_size / batch_size)
        total_loss = 0

        for i in range(num_batches):
            batch_inputs = []
            batch_targets = []
            for j in range(batch_size):
                idx = i * batch_size + j
                input_seq = [char_to_ix[ch] for ch in data[idx:idx+seq_length]]
                target_seq = [char_to_ix[data[idx+seq_length]]]
                batch_inputs.append(np.eye(vocab_size)[input_seq])
                batch_targets.append(np.eye(vocab_size)[target_seq])
            
            _, loss_val = sess.run([optimizer, loss], feed_dict={inputs: batch_inputs, targets: batch_targets})
            total_loss += loss_val
        
        avg_loss = total_loss / num_batches
        print("Epoch: {}, Loss: {}".format(epoch+1, avg_loss))

    # 生成唐诗
    seed_text = '春风得意马蹄疾'
    generated_text = seed_text
    for i in range(100):
        input_seq = [char_to_ix[ch] for ch in seed_text]
        input_seq = np.eye(vocab_size)[input_seq]
        input_seq = np.expand_dims(input_seq, axis=0)

        preds = sess.run(predictions, feed_dict={inputs: input_seq})
        next_char = ix_to_char[np.argmax(preds)]
        generated_text += next_char
        seed_text = seed_text[1:] + next_char

    print(generated_text)

由于唐诗数据集比较大，这里只给出了一个简单的示例代码。如果需要更完整的数据集，可以在网上搜索并下载。需要注意的是，在构建模型时需要根据数据集的实际情况调整参数，以达到最佳效果。