报错了 AttributeError: 'HashedCategoricalColumn' object has no attribute 'vocab_size'

非常抱歉，上述代码中存在错误。对于`HashedCategoricalColumn`没有`vocab_size`属性。我会为您提供更新后的代码。

import tensorflow as tf
from tensorflow.keras.layers import Dense, Embedding, Concatenate
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
from sklearn.model_selection import train_test_split

# 定义数据特征
feature_columns = []
embedding_dims = {}

# userId特征
user_id = tf.feature_column.categorical_column_with_hash_bucket('userId', hash_bucket_size=1000)
user_id_embedding = tf.feature_column.embedding_column(user_id, dimension=10)
feature_columns.append(user_id_embedding)
embedding_dims['userId'] = 10

# movieId特征
movie_id = tf.feature_column.categorical_column_with_hash_bucket('movieId', hash_bucket_size=10000)
movie_id_embedding = tf.feature_column.embedding_column(movie_id, dimension=10)
feature_columns.append(movie_id_embedding)
embedding_dims['movieId'] = 10

# rating特征
rating = tf.feature_column.numeric_column('rating')
feature_columns.append(rating)

# timestamp特征（可选）
timestamp = tf.feature_column.numeric_column('timestamp')
feature_columns.append(timestamp)

# 读取数据
data = pd.read_csv('your_data.csv')  # 替换为您的数据文件路径

# 划分训练集和测试集
train_data, test_data = train_test_split(data, test_size=0.2, random_state=42)

# 创建输入管道
def input_fn(df, num_epochs, shuffle=True, batch_size=32):
    df = df.copy()
    labels = df.pop('rating')
    dataset = tf.data.Dataset.from_tensor_slices((dict(df), labels))
    if shuffle:
        dataset = dataset.shuffle(buffer_size=len(df))
    dataset = dataset.batch(batch_size)
    dataset = dataset.repeat(num_epochs)
    return dataset

# 创建Wide & Deep模型
def create_wide_deep_model(feature_columns, embedding_dims):
    inputs = {}
    embeddings = []
    
    for feature_col in feature_columns:
        feature_name = feature_col.categorical_column.key
        input_tensor = tf.keras.Input(shape=(1,), name=feature_name)
        inputs[feature_name] = input_tensor
        
        if feature_name in embedding_dims:
            embedding_dim = embedding_dims[feature_name]
            embedding = tf.keras.layers.Embedding(hash_bucket_size=feature_col.hash_bucket_size, 
                                                  input_dim=feature_col.num_buckets,
                                                  output_dim=embedding_dim)(input_tensor)
            embedding = tf.keras.layers.Flatten()(embedding)
            embeddings.append(embedding)
    
    wide = tf.keras.layers.concatenate(embeddings)
    wide_output = tf.keras.layers.Dense(1, activation='sigmoid', name='wide_output')(wide)
    
    deep = tf.keras.layers.Dense(64, activation='relu')(wide)
    deep = tf.keras.layers.Dense(64, activation='relu')(deep)
    deep_output = tf.keras.layers.Dense(1, activation='sigmoid', name='deep_output')(deep)
    
    model = tf.keras.Model(inputs=inputs, outputs=[wide_output, deep_output])
    
    return model

# 创建Wide & Deep模型
model = create_wide_deep_model(feature_columns, embedding_dims)

# 编译模型
model.compile(optimizer=Adam(learning_rate=0.001),
              loss={'wide_output': 'binary_crossentropy', 'deep_output': 'mean_squared_error'},
              metrics={'wide_output': 'accuracy', 'deep_output': 'mae'})

# 训练模型
model.fit(input_fn(train_data, num_epochs=10),
          steps_per_epoch=len(train_data) // 32,
          validation_data=input_fn(test_data, num_epochs=1, shuffle=False),
          validation_steps=len(test_data) // 32,
          epochs=10)

这次我更正了`Embedding`层的输入维度和哈希桶的大小。请尝试运行更新后的代码，希望能解决问题。如果还有其他疑问，请随时提问。