Question

我多次训练神经网络，然后尝试一次加载多个副本。但是得到

tensorflow.python.framework.errors_impl.NotFoundError: Key indics_conv/conv0_2_0/w/Adam not found in checkpoint
     [[Node: save/RestoreV2_34 = RestoreV2[dtypes=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save/Const_0_0, save/RestoreV2_34/tensor_names, save/RestoreV2_34/shape_and_slices)]]

即使我尝试从相同的文件加载。保存加载是在同一Python类中进行的。区别在于，在一个实例内部是以这种方式创建的

self.sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpuOptions))
self.initTfObjects()
tf.global_variables_initializer().run(session=self.sess) # Don't remember why this is made only for the main copy. But doing for the secondary too doesn't help

和第二个-

graph = tf.Graph()
self.sess = tf.Session(graph=graph, config=tf.ConfigProto(gpu_options=gpuOptions))
with graph.as_default():
    self.initTfObjects()

保存：

self.tfSaver = tf.train.Saver(max_to_keep=10000)
self.tfSaver.save(self.sess, fileName)

正在加载：

self.tfSaver.restore(self.sess, fileName)

我缺少某些方面吗？这以前对我有用。

TensorFlow 1.4.0，Python 3.5.2 x64，Windows 7

加载网络的辅助副本时出现NotFoundError

0 个答案: