Question

我已经通过Jupyter Notebook编写了一个非常简单的模型

n_input = 3  
X = tf.placeholder(tf.float32, [None, n_input], name="X")
decoder = tf.matmul(X, [[2.0,3.0],[2.0,3.0],[2.0,3.0]], name='decoder')

init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)

values = sess.run(decoder, feed_dict={X: df_norm[:10]})

graph = tf.get_default_graph()
tf.train.write_graph(graph, './model/','saved_model.pbtxt', as_text=False)

然后我用TensorflowSharp加载了它

using (var graph = new TFGraph())
            {
                var bytes = File.ReadAllBytes(@".\model\saved_model.pbtxt");
                graph.Import(bytes);
                var session = new TFSession(graph);


                var runner = session.GetRunner();
                runner.AddInput(graph["X"][0], new float[] { 176.75f, 7.95f, 40397.00f }); 

                runner.Fetch(graph["decoder"][0]);

                var output = runner.Run();

                // Fetch the results from output:
                TFTensor result = output[0];
            }

最后，我在下面遇到了异常：

TensorFlow.TFException HResult = 0x80131500메시지=您必须输入   dtype float和形状为[？，3]的占位符张量'X_1'的值
  [[{{node X_1}} = Placeholderdtype = DT_FLOAT，shape = [?, 3]，   _device =“ / job：localhost / replica：0 / task：0 / device：CPU：0”]]소스= TensorFlowSharp StackTrace：at   TensorFlow.TFStatus.CheckMaybeRaise（TFStatus entryStatus，布尔值   最后）在TensorFlow.TFSession.Run（TFOutput []输入，TFTensor []   inputValues，TFOutput []输出，TFOperation [] targetOpers，TFBuffer   runMetadata，TFBuffer runOptions，TFStatus状态）   TensorFlow.TFSession.Runner.Run（TFStatus状态）位于   tensorflowsharp_model_restore.Program.Main（String [] args）在   G：\ tensorflow \ tensorflowsharp \ tensorflowsharp_model_restore \ tensorflowsharp_model_restore \ Program.cs：line   29

您必须使用dtype float和shape [？，3]输入占位符张量“ X_1”的值

0 个答案: