给我一份用tensorflow2.0写的关于词向量训练和预测，并作出词向量散点图的代码

以下是一个简单的用TensorFlow 2.0实现的词向量训练和预测的代码，同时作出词向量散点图：

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

# 定义数据集
corpus_raw = 'He is the king . The king is royal . She is the royal queen '
corpus_raw = corpus_raw.lower()

# 将数据集转换成单词列表
words = []
for word in corpus_raw.split():
    if word != '.':  # 去除句号
        words.append(word)

words = set(words)  # 去重
word2int = {}
int2word = {}

vocab_size = len(words)  # 词汇表大小
for i, word in enumerate(words):
    word2int[word] = i
    int2word[i] = word

# 构造训练数据
data = []
WINDOW_SIZE = 2  # 窗口大小

for i, word in enumerate(words):
    for j in range(max(i - WINDOW_SIZE, 0), min(i + WINDOW_SIZE, len(words))):
        if i != j:
            data.append((word, words[j]))

# 定义模型参数
EMBEDDING_DIM = 2  # 词向量维度
learning_rate = 0.1
epochs = 10000

# 定义模型
class Word2Vec(tf.keras.Model):
    def __init__(self, vocab_size, embedding_dim):
        super(Word2Vec, self).__init__()
        self.target_embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=1, name="w2v_embedding")
        self.context_embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=WINDOW_SIZE*2, name="w2v_context_embedding")
    
    def call(self, target, context):
        target_embedding = self.target_embedding(target)
        context_embedding = self.context_embedding(context)
        dot = tf.keras.layers.Dot(axes=2)([context_embedding, target_embedding])
        dot = tf.keras.layers.Reshape((1,))(dot)
        output = tf.keras.layers.Activation('sigmoid')(dot)
        return output

model = Word2Vec(vocab_size, EMBEDDING_DIM)
model.compile(loss='binary_crossentropy', optimizer=tf.keras.optimizers.Adam(learning_rate), metrics=['accuracy'])

# 训练模型
for epoch in range(epochs):
    x, y = [], []
    for w in data:
        x.append(word2int[w[0]])
        y.append(word2int[w[1]])
    loss, acc = model.train_on_batch([np.array(x), np.array(y)], np.ones(len(x)))
    if epoch % 1000 == 0:
        print("Epoch:", epoch, "Loss:", loss, "Accuracy:", acc)

# 可视化词向量
weights = model.get_layer('w2v_embedding').get_weights()[0]
for i, word in int2word.items():
    x, y = weights[i]
    plt.scatter(x, y)
    plt.annotate(word, xy=(x, y), xytext=(5, 2), textcoords='offset points', ha='right', va='bottom')
plt.show()

# 预测词向量
word = 'queen'
word_index = word2int[word]
target_word_vector = model.get_layer('w2v_embedding').get_weights()[0][word_index]
similar_words = []
for i, w in int2word.items():
    if w != word:
        vector = model.get_layer('w2v_embedding').get_weights()[0][i]
        similarity = np.dot(target_word_vector, vector) / (np.linalg.norm(target_word_vector) * np.linalg.norm(vector))
        similar_words.append((w, similarity))
similar_words = sorted(similar_words, key=lambda x: x[1], reverse=True)
print(similar_words[:5])

上述代码的输出结果包括每1000个epoch的训练损失、准确率，以及词向量散点图和预测词向量的相似词。