根据预测，我加载的模型给了我一个 AttributeError

Question

总的来说，我对 Tensorflow 和机器学习还很陌生，但我知道我已经建立了一个小型模型。虽然，当我加载并使用model.predict 时，我得到一个属性错误：

import tensorflow as tf
import numpy as np

checkpoint_path = "training_1/cp.ckpt"
# Hyperparamters
vocab_size = 2000
embedding_dim = 16
max_length = 1
trunc_type = "post"
padding_type = "post"
oov_tok = "<OOV>"
training_size = 100

model = tf.keras.Sequential([
    tf.keras.layers.Embedding(
        vocab_size, embedding_dim, input_length=max_length),
    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(32)),
    tf.keras.layers.Dense(128, activation="relu"),
    tf.keras.layers.Dense(3, activation="softmax")
])

# Compile the model
model.compile(loss="sparse_categorical_crossentropy",
              optimizer="adam", metrics=["accuracy"])

model.load_weights(checkpoint_path)


test = ["Example of text here"]


prediction = model.predict(test)
print(prediction)

Traceback (most recent call last):
  File "./ModelTest.py", line 36, in <module>
    prediction = model.predict(test)
  File "/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py", line 1060, in predict
    x, check_steps=True, steps_name='steps', steps=steps)
  File "/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py", line 2651, in _standardize_user_data
    exception_prefix='input')
  File "/lib/python3.7/site-packages/tensorflow/python/keras/engine/training_utils.py", line 334, in standardize_input_data
    standardize_single_array(x, shape) for (x, shape) in zip(data, shapes)
  File "/lib/python3.7/site-packages/tensorflow/python/keras/engine/training_utils.py", line 334, in <listcomp>
    standardize_single_array(x, shape) for (x, shape) in zip(data, shapes)
  File "/lib/python3.7/site-packages/tensorflow/python/keras/engine/training_utils.py", line 265, in standardize_single_array
    if (x.shape is not None and len(x.shape) == 1 and
AttributeError: 'str' object has no attribute 'shape'

Answer 1

确保您提供的输入格式适合您构建的模型。在您的情况下，Embedding 层需要一个 2D 张量。数据应该是一个看起来像这样的 numpy 数组：[[0, 2, 64], [24, 6, 8]]。那里的每个数字代表一个单词，每个数字序列代表一个短语。整个张量代表一批序列。在我的示例中，这是一组 2 个序列，每个序列有 3 个单词。

您需要做的是使用您正在加载的模型的正确词汇表标记"Example of text here"。完成后，您将得到一个类似[[3, 8, 4, 6]] 的数组，其中每个数字对应于"Example of text here" 中的一个单词。如何正确标记它取决于它训练的数据是如何标记的，如果不知道你从哪里得到training_1/cp.ckpt，我们不知道这一点。