我尝试在程序中加载模型，以便实时运行它。
以下是我尝试的方法：
saver =
tf.train.import_meta_graph(tf.train.latest_checkpoint(model_path)+".meta")graph = tf.get_default_graph()outputs = graph.get_operation_by_name('output')native = graph.get_operation_by_name('input')sess = tf.Session()sess.run(tf.global_variables_initializer()) sess.run(tf.local_variables_initializer()) if(tf.train.checkpoint_exists(tf.train.latest_checkpoint(model_path))): saver.restore(sess, tf.train.latest_checkpoint(model_path)) print(tf.train.latest_checkpoint(model_path) + "Session Loaded for Testing") 我尝试获取的输出是：
y_test_output= sess.run(outputs, feed_dict={native: x_test})我得到了以下错误：
Traceback (most recent call last): File "testing_reality.py", line 90, in <module> main() File "testing_reality.py", line 62, in main handle_client_connection(client_sock) File "testing_reality.py", line 45, in handle_client_connection y_train_pred= sess.run(outputs, feed_dict={native: x_test}) File "C:\Python35\lib\site-packages\tensorflow\python\client\session.py", line 929, in run run_metadata_ptr) File "C:\Python35\lib\site-packages\tensorflow\python\client\session.py", line 1095, in _run 'Cannot interpret feed_dict key as Tensor: ' + e.args[0])TypeError: Cannot interpret feed_dict key as Tensor: Can not convert a Operation into a Tensor.请告诉我在这个场景中我遗漏了什么。
使用Windows 10上的TensorFlow最新版本’1.12.0’。
回答：
feed_dict中的keys必须是占位符。错误显示您在feed_dict中使用了Operation作为键。如果input变量确实是placeholder，那么请按以下方式加载它：
native = graph.get_tensor_by_name("input:0")对于outputs也是如此：
outputs = graph.get_tensor_by_name("output:0")