ImportError: cannot import name 'dense_features' from 'tensorflow.python.feature

前提・実現したいこと

ディープラーニングの教材でリカレントニューラルネットワークの勉強をしてたところ，tf.contrib.rnn.BasicRNNCellを使用した際に出たエラーで

発生している問題・エラーメッセージ

ImportError: cannot import name 'dense_features' from 'tensorflow.python.feature_column' (C:\Users\kopur\anaconda3\lib\site-packages\tensorflow\python\feature_column\__init__.py)

と出ました．対処法がわからず，困っています．

該当のソースコード

python
1import tensorflow as tf
2import numpy as np
3from sklearn.utils import shuffle
4from sklearn.model_selection import train_test_split
5
6def inference(x, n_batch, maxlen=None, n_hidden=None, n_out=None):
7        def weight_variable(shape):
8            initial = tf.truncated_normal(shape, stddev=0.01)
9            return tf.Variable(initial)
10    
11        def bias_variable(shape):
12            initial = tf.zeros(shape, dtype=tf.float32)
13            return tf.Variable(initial)
14    
15        cell = tf.contrib.rnn.BasicRNNCell(n_hidden)
16        initial_state = cell.zero_state(n_batch, tf.float32)
17    
18        state = initial_state
19        outputs = []
20        with tf.variable_scope('RNN'):
21            for t in range(maxlen):
22                if t >0:
23                    tf.get_variable_scope().reuse_variables()
24                (cell_output, state) = cell(x[:,t,:],state)
25                outputs.append(cell_output)
26    
27        output = outputs[-1]
28    
29        V = weight_variable([n_hidden,n_out])
30        c = bias_variable([n_out])
31        y = tf.matmul(output, V) + c  #線形活性
32    
33        return y
34    
35def sin(x, T=100):
36    return np.sin(2.0 * np.pi * x/T)
37
38def toy_problem(T=100, ampl=0.05):
39    x = np.arange(0, 2 * T + 1)
40    noise = ampl * np.random.uniform(low=-1.0, high=1.0, size=len(x))
41    return sin(x) + noise
42
43
44T = 100
45f = toy_problem(T)
46
47length_of_sequences = 2 * T #全時系列の長さ
48maxlen = 25 #一つの時系列データの長さ
49
50data=[]
51target=[]
52
53for i in range(0, length_of_sequences - maxlen + 1):
54    data.append(f[i: i + maxlen])  #i<= <i + maxlen　という意味
55    target.append(f[i + maxlen])
56
57X = np.array(data).reshape(len(data), maxlen, 1)
58Y = np.array(target).reshape(len(data), 1)
59
60N_train = int(len(data) * 0.9)
61N_validation = len(data) - N_train
62X_train, X_validation, Y_train, Y_validation =\
63    train_test_split(X, Y, test_size=N_validation)
64
65
66n_in = len(X[0][0])
67n_hidden = 1
68n_out = len(Y[0])
69
70x = tf.compat.v1.placeholder(tf.float32, shape=[None, maxlen, n_in])
71t = tf.compat.v1.placeholder(tf.float32, shape=[None, n_out])
72n_batch = tf.compat.v1.placeholder(tf.int32)
73
74y = inference(x, n_batch, maxlen=maxlen, n_hidden=n_hidden, n_out=n_out)
75
76