我一直在尝试恢复tensorflow中的模型,但是当我尝试导入元图时,我遇到了一些问题:
这是我导入元图的代码:
#Create a clean graph and import MetaGraphDef nodes
new_graph = tf.Graph()
with tf.Session(graph=new_graph) as sess:
# Import the previously exported metagraph
saver = tf.train.import_meta_graph('/tmp/saver-model.meta')
saver.restore(sess, tf.train.latest_checkpoint('./'))
Run Code Online (Sandbox Code Playgroud)
在我的Model类中,我已经指定了占位符和集合,如下所示:
"""Place Holders"""
self.input = tf.placeholder(tf.float32, [None, sl], name = 'input')
self.labels = tf.placeholder(tf.int64, [None], name = 'labels')
self.keep_prob = tf.placeholder("float", name= 'Drop_out_keep_prob')
tf.add_to_collection('vars', self.input)
tf.add_to_collection('vars', self.labels)
tf.add_to_collection('vars', self.keep_prob)
Run Code Online (Sandbox Code Playgroud)
我训练我的模型如下:
saver = tf.train.Saver(tf.global_variables())
# Session time
sess = tf.Session() # without context manager, close the session later.
writer = tf.summary.FileWriter("/tmp/model/log_tb", sess.graph) # …Run Code Online (Sandbox Code Playgroud) 我正在尝试使用
df.set_Value.
df 格式:
Count DTW
DateTime
2015-01-16 10 0
2015-01-17 28 0
Run Code Online (Sandbox Code Playgroud)
使用 df.setValue
dw.set_Value(idx, 'col', dtw) # idx and dtw are int values
TypeError: cannot insert DatetimeIndex with incompatible label
Run Code Online (Sandbox Code Playgroud)
我该如何解决这个错误,或者有什么具有可比效率的替代方法?