这些是一些张量和类似的运算符:
x = tf.placeholder(....., name='input')
y = tf.placeholder(....., name='label')
...
fc_droop = tf.nn.dropout(....., name='fc1')
...
y_pred = tf.nn.softmax(....., name='predict')
...
train_step = tf.train.AdamOptimizer(lr).minimize(loss, name='train_step')
训练完成后,我将其另存为模型。
但是当我还原它并通过tf.get_tensor_by_name('xx:0')
获取图节点时,“输入”,“标签”,“预测”都可以,但是“ fc1”,“ train_step”得到了错误:
the name 'xx' refers to a tensor which does not exist
请问是什么问题