我想在tf-slim
中使用内置的resnet
进行快速实验。我根据 github 中的README
做到了:
import tensorflow as tf
import tensorflow.contrib.slim as slim
resnet = tf.contrib.slim.nets.resnet_v1
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
x = tf.placeholder("float", shape=[None, 784])
y_ = tf.placeholder("float", shape=[None, 10])
pred = resnet.resnet_v1_50(x)
cross_entropy = -tf.reduce_sum(y_ * tf.log(pred))
但是得到了这样的错误:AttributeError: module 'tensorflow.contrib.slim' has no attribute 'nets'
.我已经安装了最新版本的tensorflow-0.12.0
.
如何解决此问题?
像这样导入:
from tensorflow.contrib.slim.nets import resnet_v1
slim = tf.contrib.slim