Tensorflow实践报错汇总

1. Tensorflow版本问题而导致报错

如果项目代码基于tensorflow1.0，而自己机器上安装的是tensorflow2.0，则会出现兼容问题，可用以下方法解决。

#导入tensorflow
import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

2. AttributeError: module ‘tensorflow‘ has no attribute ‘compat’

如果解决版本问题时采用如下方式

sess = tf.compat.v1.InteractiveSession()

报错 AttributeError: module ‘tensorflow‘ has no attribute ‘compat’
解决办法同 1. Tensorflow版本问题而导致报错

3. tensorflow2.x中tf.contrib.layers.xavier_initializer()替换问题

tensorflow2.x中删去了tensorflow1.x中contrib，因此报错AttributeError: module 'tensorflow_core.compat.v1' has no attribute 'contrib'。解决办法如下：
将

tf.contrib.layers.xavier_initializer()

替换成

tf.keras.initializers.glorot_normal()

出错代码：

def conv_op(input_op, name, kh, kw, n_out, dh, dw, p):
    n_in = input_op.get_shape()[-1].value

    with tf.name_scope(name) as scope:
        kernel = tf.get_variable(scope + "w",
                                 shape=[kh, kw, n_in, n_out],
                                 dtype=tf.float32,
                                 #下行报错                                 
                                 initializer=tf.keras.initializers.glorot_normal())
        conv = tf.nn.conv2d(input_op, kernel, (1, dh, dw, 1), padding='SAME')
        bias_init_val = tf.constant(0.0, shape=[n_out], dtype=tf.float32)
        biases = tf.Variable(bias_init_val, trainable=True, name='b')
        z = tf.nn.bias_add(conv, biases)
        activation = tf.nn.relu(z, name=scope)
        p += [kernel, biases]
        return activation

4. tf2.x中使用slim报错

参考链接https://blog.csdn.net/lingchuxiao/article/details/109489809

Traceback (most recent call last):
  File "/Users/jyc/PycharmProjects/研究生新生教程作业/Tensorflow/GoogleNet.py", line 19, in <module>
    slim = tf.contrib.slim
AttributeError: module 'tensorflow_core.compat.v1' has no attribute 'contrib'

原因：

#下行报错
slim = tf.contrib.slim

解决办法：tensorflow2.0中要单独安装tf-slim库
终端运行pip install --upgrade tf_slim

# slim = tf.contrib.slim
import tensorflow.compat.v1 as tf
import tf_slim as slim

tf.disable_v2_behavior()