运行时错误:尝试在不构建函数的情况下捕获 EagerTensor

Question

我使用tf.data.datsetAPI 并使用残差网络。当我为 TensorBoard 运行代码以可视化我的嵌入时，我遇到了此错误，但是当我使用两层网络时，我没有这个问题。

def load_and_preprocess_from_path_label(path, label):
return load_and_preprocess_image(path), label
ds = tf.data.Dataset.from_tensor_slices((all_image_paths, all_image_labels))
with tf.Session() as sess:
# TODO (@omoindrot): remove the hard-coded 10000
# Obtain the test labels
image_label_ds = ds.map(load_and_preprocess_from_path_label)
ds = image_label_ds.shuffle(image_count)

---

RuntimeError                              Traceback (most recent call last)
<ipython-input-41-ead5d6a54baa> in <module>()
92         # TODO (@omoindrot): remove the hard-coded 10000
93         # Obtain the test labels
---> 94         image_label_ds = ds.map(load_and_preprocess_from_path_label)
95         ds = image_label_ds.shuffle(image_count)
96 
RuntimeError: Attempting to capture an EagerTensor without building a function.

Answer 1

该错误可能是由于Tensorflow版本造成的

运行以下代码对我有用：

import tensorflow.compat.v1 as tf

正确的功能：

tf.disable_v2_behavior()

Answer 2

运行以下代码对我有用：

from keras.models import Sequential
from keras.layers import LSTM, Dense, Dropout
from keras.callbacks import EarlyStopping
from keras import backend as K
import tensorflow as tf

tf.compat.v1.enable_eager_execution()

如果您还使用以下代码在创建模型后强制 LSTM 清除模型参数和图形，那就太好了。

K.clear_session()
tf.compat.v1.reset_default_graph()
#tf.compat.v1.disable_eager_execution()