AttributeError：对象没有属性“_output_tensor_cache”答案

【问题标题】：AttributeError: object has no attribute '_output_tensor_cache'AttributeError：对象没有属性“_output_tensor_cache”
【发布时间】：2020-04-13 22:58:45
【问题描述】：

我正在尝试基于rotten tomatoes dataset 进行情绪分析分类器。我收到此错误：

AttributeError: 'RNN' 对象没有属性 '_output_tensor_cache'

不幸的是，没有任何帮助。我什至不知道在哪里看。

import pandas as pd
from sklearn.model_selection import train_test_split
from tensorflow import keras
from keras import Model
from tensorflow.keras.layers import LSTM, Embedding, Dense
import os
os.chdir('/home/nicolas/Documents/datasets')

df = pd.read_csv('rotten_tomatoes_reviews.csv', nrows=50_000)

df = df.loc[df.Review.str.len() >= 3]

array = df.Review.values
target = df.Freshness.values

tokenizer = keras.preprocessing.text.Tokenizer(num_words=3_000)

tokenizer.fit_on_texts(array)
vector = tokenizer.texts_to_sequences(array)

padded = keras.preprocessing.sequence.pad_sequences(vector, maxlen=40)

X_train, X_test, y_train, y_test = train_test_split(padded, target, test_size=2e-1)


class RNN(Model):
    def __init__(self):
        super(RNN, self).__init__()
        self.rnn1 = LSTM(8, return_sequences=True, return_state=True)
        self.rnn2 = LSTM(8)
        self.emb1 = Embedding(input_dim=3_000, output_dim=50, input_length=40)
        self.flc1 = Dense(2)

    def __call__(self, inputs, training=None, mask=None):
        x = self.emb1(inputs)
        x = self.rnn1(x)
        x = self.rnn2(x)
        out = self.flc1(x)
        print(out.shape)
        return out


def main():
    model = RNN()

    model.compile(optimizer=keras.optimizers.Adam(0.001),
                  loss=keras.losses.BinaryCrossentropy(from_logits=True),
                  metrics=['accuracy'])

    model.fit(X_train, y_train, batch_size=16, epochs=10,
              validation_data=[X_test, y_test], verbose=1)

    scores = model.evaluate(X_test, y_test, batch_size=16, verbose=1)
    print("Final test loss and accuracy :", scores)


if __name__ == '__main__':
    main()

【问题讨论】：

标签： python tensorflow machine-learning keras lstm

【解决方案1】：

那是因为你混合了原生 Keras 实现和 Keras 的 TensorFlow 实现（即tf.keras）：

from tensorflow import keras
from keras import Model     # Wrong! DON'T mix keras and tf.keras!
from tensorflow.keras.layers import LSTM, Embedding, Dense

你应该永远不要那样做。使用 from tensorflow.keras import Model 修复 Model 类的导入

【讨论】：