我已经通过以下文章对cartpole tensorflow模型进行了培训https://chuacheowhuan.github.io/DDQN/
在测试时，我遵循了这篇文章https://blog.metaflow.fr/tensorflow-how-to-freeze-a-model-and-serve-it-with-a-python-api-d4f3596b3adc和给出了我的x和y值如下：
"x=图形.get_tensor_by_name('prefix/s:0')---->初始状态y=图形.get_tensor_by_name('prefix/a:0')---->要执行的最终操作？？
# We launch a Session
with tf.Session(graph=graph) as sess:
# Note: we don't nee to initialize/restore anything
# There is no Variables in this graph, only hardcoded constants
y_out = sess.run(y, feed_dict={x: x})
我给出了x作为侧手杆的状态，y是侧手杆上的动作，但它是投掷错误，如下所述：tensorflow.python.framework.errors_impl.InvalidArgumentError:必须为占位符张量"prefix/a"提供一个dtype为int32、shape[？]的值[[｛节点前缀/a｝｝]]
张量为：前缀/r前缀/s_next前缀/完成前缀/型号_s_next_Q_valprefix/model_net/fully_connected/weights/Initializer/random_uuniform/shapeprefix/model_net/fully_connected/weights/Initializer/random_uuniform/minprefix/model_net/fully_connected/weights/Initializer/random_uuniform/maxprefix/model_net/fully_connected/weights/Initializer/随机统一prefix/model_net/fully_connected/weights/Initializer/random_uuniform/subprefix/model_net/fully_connected/weights/Initializer/random_uuniform/mulprefix/model_net/fully_connected/weights/Initializer/random_uniformprefix/model_net/fully_connected/weightsprefix/model_net/fully_connected/weights/Assignprefix/model_net/fully_connected/weights/readprefix/model_net/fully_connected/bias/Initializer/random_uuniform/shapeprefix/model_net/fully_connected/bias/Initializer/random_uuniform/minprefix/model_net/fully_connected/bias/Initializer/random_uuniform/maxprefix/model_net/fully_connected/bias/Initializer/random_uuniform/RandomUniformprefix/model_net/fully_connected/bias/Initializer/random_uuniform/subprefix/model_net/fully_connected/bias/Initializer/random_uuniform/mulprefix/model_net/fully_connected/bias/Initializer/random_uuniformprefix/model_net/fully_connected/biasprefix/model_net/fully_connected/bias/分配prefix/model_net/fully_connected/bias/readprefix/model_net/fully_connected/MatMulprefix/model_net/fully_connected/BiasAddprefix/model_net/fully_connected/Reluprefix/model_net/fully_connected_1/weights/Initializer/random_uuniform/shapeprefix/model_net/fully_connected_1/weights/Initializer/random_uuniform/minprefix/model_net/fully_connected_1/weights/Initializer/random_uuniform/maxprefix/model_net/fully_connected_1/weights/Initializer/random_uuniform/RandomUniformprefix/model_net/fully_connected_1/weights/Initializer/random_uuniform/subprefix/model_net/fully_connected_1/weights/Initializer/random_uuniform/mulprefix/model_net/fully_connected_1/weights/Initializer/random_uuniformprefix/model_net/fully_connected_1/weightsprefix/model_net/fully_connected_1/weights/Assignprefix/model_net/fully_connected_1/weights/readprefix/model_net/fully_connected_1/bias/Initializer/random_uuniform/shapeprefix/model_net/fully_connected_1/pias/Initializer/random_uuniform/minprefix/model_net/fully_connected_1/pias/Initializer/random_uuniform/maxprefix/model_net/fully_connected_1/bias/Initializer/random_uuniform/RandomUniformprefix/model_net/fully_connected_1/pias/Initializer/random_uuniform/subprefix/model_net/fully_connected_1/pias/Initializer/random_uuniform/mulprefix/model_net/fully_connected_1/bias/Initializer/random_uuniformprefix/model_net/fully_connected_1/biasprefix/model_net/fully_connected_1/bias/Assignprefix/model_net/fully_connected_1/bias/readprefix/model_net/fully_connected_1/MatMul前缀/model_net/fully_connected_1/BiasAddprefix/target_net/fully_connected/weights/Initializer/random_uuniform/shapeprefix/target_net/fully_connected/weights/Initializer/random_uuniform/minprefix/target_net/fully_connected/weights/Initializer/random_uuniform/maxprefix/target_net/fully_connected/weights/Initializer/random_uuniform/RandomUniformprefix/target_net/fully_connected/weights/Initializer/random_uuniform/subprefix/target_net/fully_connected/weights/Initializer/random_uuniform/mulprefix/target_net/fully_connected/weights/Initializer/random_uniformprefix/target_net/fully_connected/weightsprefix/target_net/fully_connected/weights/Assignprefix/target_net/fully_connected/wweights/readprefix/target_net/fully_connected/pias/Initializer/random_uuniform/shapeprefix/target_net/fully_connected/pias/Initializer/random_uuniform/minprefix/target_net/fully_connected/pias/Initializer/random_uuniform/maxprefix/target_net/fully_connected/pias/Initializer/random_uuniform/RandomUniformprefix/target_net/fully_connected/pias/Initializer/random_uuniform/subprefix/target_net/fully_connected/pias/Initializer/random_uuniform/mulprefix/target_net/fully_connected/pias/Initializer/random_uniformprefix/target_net/fully_connected/piasprefix/target_net/fully_connected/pias/Assignprefix/target_net/fully_connected/pias/readprefix/target_net/fully_connected/MatMulprefix/target_net/fully_connected/BiasAddprefix/target_net/fully_connected/Reluprefix/target_net/fully_connected_1/weights/Initializer/random_uuniform/shapeprefix/target_net/fully_connected_1/weights/Initializer/random_uuniform/minprefix/target_net/fully_connected_1/weights/Initializer/random_uuniform/maxprefix/target_net/fully_connected_1/weights/Initializer/random_uuniform/RandomUniformprefix/target_net/fully_connected_1/weights/Initializer/random_uuniform/subprefix/target_net/fully_connected_1/weights/Initializer/random_uuniform/mulprefix/target_net/fully_connected_1/weights/Initializer/random_uniformprefix/target_net/fully_connected_1/weightsprefix/target_net/fully_connected_1/weights/Assignprefix/target_net/fully_connected_1/weights/readprefix/target_net/fully_connected_1/pias/Initializer/random_uuniform/shapeprefix/target_net/fully_connected_1/pias/Initializer/random_uuniform/minprefix/target_net/fully_connected_1/pias/Initializer/random_uuniform/maxprefix/target_net/fully_connected_1/pias/Initializer/random_uuniform/RandomUniformprefix/target_net/fully_connected_1/pias/Initializer/random_uuniform/subprefix/target_net/fully_connected_1/pias/Initializer/random_uuniform/mulprefix/target_net/fully_connected_1/pias/Initializer/random_uuniformprefix/target_net/fully_connected_1/biasprefix/target_net/fully_connected_1/pias/Assignprefix/target_net/fully_connected_1/pias/readprefix/target_net/fully_connected_1/MatMulprefix/target_net/fully_connected_1/BiasAddprefix/td_target/ArgMax/维度prefix/td_target/ArgMaxprefix/td_target/Shapeprefix/td_target/straded_slice/stackprefix/td_target/straded_slice/stack_1prefix/td_target/straded_slice/stack_2前缀/td_target/straded_sliceprefix/td_target/range/startprefix/td_target/range/deltaprefix/td_target/rangeprefix/td_target/stackprefix/td_target/GatherNdprefix/td_target/Reshape/shapeprefix/td_target/Reshapeprefix/td_target/Castprefix/td_target/sub/x前缀/td_target/subprefix/td_target/Mul/xprefix/td_target/Mul前缀/td_target/mul_1prefix/td_target/addprefix/td_target/StopGradient前缀/预测_Q_val/形状prefix/predicted_Q_val/stridded_slice/stackprefix/predicted_Q_val/stridded_slice/stack_1prefix/predicted_Q_val/stridded_slice/stack_2前缀/predicted_Q_val/stridded_sliceprefix/predicted_Q_val/range/startprefix/predicted_Q_val/range/delta前缀/预测_Q_val/范围prefix/predicted_Q_val/stackprefix/predicted_Q_val/GatherNdprefix/predicted_Q_val/Reshape/shape前缀/预测_Q_val/重塑prefix/loss/huber_loss/Subprefix/loss/huber_loss/Absprefix/loss/huber_loss/Minimum/yprefix/loss/huber_loss/Minimumprefix/loss/huber_loss/Sub_1prefix/loss/huber_loss/Constprefix/loss/huber_loss/Mulprefix/loss/huber_loss/Mul_1prefix/loss/huber_loss/Mul_2/xprefix/loss/huber_loss/Mul_2prefix/loss/huber_loss/Addprefix/loss/huber_loss/assert_broadcastle/weightsprefix/loss/huber_loss/assert_broadcastle/weights/shapeprefix/loss/huber_loss/assert_broadcastle/weights/rankprefix/loss/huber_loss/assert_broadcastle/values/shapeprefix/loss/huber_loss/assert_broadcastle/values/rankprefix/loss/huber_loss/assert_broadcastable/static_scaler_check_successprefix/loss/huber_loss/Cast/xprefix/loss/huber_loss/Mul_3prefix/loss/huber_loss/Const_1prefix/loss/huber_loss/Sumprefix/loss/huber_loss/num_present/Equal/yprefix/loss/huber_loss/num_present/Equalprefix/loss/huber_loss/num_present/zeros-likeprefix/loss/huber_loss/num_present/ones_like/Shapeprefix/loss/huber_loss/num_present/ones_like/Constprefix/loss/huber_loss/num_present/ones_likeprefix/loss/huber_loss/num_present/选择prefix/loss/huber_loss/num_present/broadcast_weights/assert_broadcast/weights/shapeprefix/loss/huber_loss/num_present/broadcast_weights/assert_broadcast/weights/rankprefix/loss/huber_loss/num_present/broadcast_weights/assert_broadcast/values/shapeprefix/loss/huber_loss/num_present/broadcast_weights/assert_broadcast/values/rankprefix/loss/huber_loss/num_present/broadcast_weights/assert_broadcastable/static_scaler_check_successprefix/loss/huber_loss/num_present/broadcast_weights/ones_like/Shapeprefix/loss/huber_loss/num_present/broadcast_weights/ones_like/Constprefix/loss/huber_loss/num_present/broadcast_weights/ones_likeprefix/loss/huber_loss/num_present/broadcast_weightsprefix/loss/huber_loss/num_present/Constprefix/loss/huber_loss/num_presentprefix/loss/huber_loss/Const_2prefix/loss/huber_loss/Sum_1prefix/loss/huber_loss/value前缀/优化器/梯度/形状前缀/优化器/梯度/grad_ys_0前缀/优化器/渐变/填充prefix/optimizer/gradients/loss/huber_loss/value_grad/Shapeprefix/optimizer/gradients/loss/huber_loss/value_grad/Shape_1prefix/optimizer/gradients/loss/huber_loss/value_grad/BroadcastGradientArgsprefix/optimizer/gradients/loss/huber_loss/value_grad/div_no_nanprefix/optimizer/gradients/loss/huber_loss/value_grad/Sumprefix/optimizer/gradients/loss/huber_loss/value_grad/Reshapeprefix/optimizer/gradients/loss/huber_loss/value_grad/Negprefix/optimizer/gradients/loss/huber_loss/value_grad/div_no_nan_1prefix/optimizer/gradients/loss/huber_loss/value_grad/div_no_nan2prefix/optimizer/gradients/loss/huber_loss/value_grad/mulprefix/optimizer/gradients/loss/huber_loss/value_grad/Sum_1prefix/optimizer/gradients/loss/huber_loss/value_grad/Reshape_1prefix/optimizer/gradients/loss/huber_loss/value_grad/tuple/group_depsprefix/optimizer/gradients/loss/huber_loss/value_grad/tuple/control_dependencyprefix/optimizer/gradients/loss/huber_loss/value_grad/tuple/control_dependency_1前缀/优化器/梯度/丢失/集线器丢失/Sum_1_grad/Reshape/shape前缀/优化器/梯度/丢失/集线器丢失/Sum_1_grad/Reshape前缀/优化器/梯度/丢失/集线器丢失/Sum_1_grad/Constprefix/optimizer/gradients/loss/huber_loss/Sum_1_grad/Tile前缀/优化器/梯度/丢失/集线器丢失/Sum_grad/Reshape/shapeprefix/optimizer/gradients/loss/huber_loss/Sum_grad/Reshapeprefix/optimizer/gradients/loss/huber_loss/Sum_grad/Constprefix/optimizer/gradients/loss/huber_loss/Sum_grad/Tileprefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/BroadcastGradientArgs/s0prefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/BroadcastGradientArgs/s1prefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/BroadcastGradientArgsprefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/Mul前缀/优化器/梯度/丢失/集线器丢失/Mul_3_grad/Mul_1prefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/Sum/reduction_indexprefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/Sumprefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/Reshape/shapeprefix/optimizer/gradients/loss/huber_loss/Mul_3_grad/Reshape前缀/优化器/梯度/丢失/集线器丢失/Mul_3_grad/tuple/group_deps前缀/优化器/梯度/丢失/集线器丢失/Mul_3_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Mul_3_grad/tuple/control_dependency_1prefix/optimizer/gradients/loss/huber_loss/Add_grad/tuple/group_depsprefix/optimizer/gradients/loss/huber_loss/Add_grad/tuple/control_dependencyprefix/optimizer/gradients/loss/huber_loss/Add_grad/tuple/control_dependency_1prefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/BroadcastGradientArgs/s0prefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/BroadcastGradientArgs/s1prefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/BroadcastGradientArgsprefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/Mulprefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/Sum/reduction_indexprefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/Sumprefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/Reshape/shapeprefix/optimizer/gradients/loss/huber_loss/Mul_1_grad/Reshape前缀/优化器/梯度/丢失/集线器丢失/Mul_1_grad/Mul_1前缀/优化器/梯度/丢失/集线器丢失/Mul_1_grad/tuple/group_deps前缀/优化器/梯度/丢失/集线器丢失/Mul_1_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Mul_1_grad/tuple/control_dependency_1prefix/optimizer/gradients/loss/huber_loss/Mul_2_grad/Mulprefix/optimizer/gradients/loss/huber_loss/Mul_2_grad/Sum/reduction_indexprefix/optimizer/gradients/loss/huber_loss/Mul_2_grad/Sum前缀/优化器/梯度/丢失/集线器丢失/Mul_2_grad/Reshape/shapeprefix/optimizer/gradients/loss/huber_loss/Mul_2_grad/Reshape前缀/优化器/梯度/丢失/集线器丢失/Mul_2_grad/Mul_1前缀/优化器/梯度/丢失/集线器丢失/Mul_2_grad/tuple/group_deps前缀/优化器/梯度/丢失/集线器丢失/Mul_2_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Mul_2_grad/tuple/control_dependency_1prefix/optimizer/gradients/loss/huber_loss/Mul_grad/Mul前缀/优化器/梯度/丢失/集线器丢失/Mul_grad/Mul_1前缀/优化器/梯度/丢失/集线器丢失/Mul_grad/tuple/group_deps前缀/优化器/梯度/丢失/集线器丢失/Mul_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Mul_grad/tuple/control_dependency_1前缀/优化器/梯度/丢失/集线器丢失/Sub_1_grad/Neg前缀/优化器/梯度/丢失/集线器丢失/Sub_1_grad/tuple/group_deps前缀/优化器/梯度/丢失/集线器丢失/Sub_1_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Sub_1_grad/tuple/control_dependency_1前缀/优化器/梯度/AddNprefix/optimizer/gradients/loss/huber_loss/Minimum_grade/Shapeprefix/optimizer/gradients/loss/huber_loss/Minimum_grad/Shape_1前缀/优化器/梯度/丢失/集线器丢失/Minimum_grad/Shape_2prefix/optimizer/gradients/loss/huber_loss/Minimum_grad/zeros/Constprefix/optimizer/gradients/loss/huber_loss/Minimum_grad/zerosprefix/optimizer/gradients/loss/huber_loss/Minimum_grad/LessEqualprefix/优化器/梯度/丢失/集线器丢失/Minimum_grade/BroadcastGradientArgsprefix/optimizer/gradients/loss/huber_loss/Minimum_grad/Selectprefix/optimizer/gradients/loss/huber_loss/Minimum_grad/Sumprefix/optimizer/gradients/loss/huber_loss/Minimum_grade/Reshapeprefix/optimizer/gradients/loss/huber_loss/Minimum_grad/Select_1前缀/优化器/梯度/丢失/集线器丢失/Minimum_grad/Sum_1prefix/optimizer/gradients/loss/huber_loss/Minimum_grad/Reshape_1前缀/优化器/梯度/丢失/集线器丢失/Minimum_grad/tuple/group_deps前缀/优化器/梯度/丢失/集线器丢失/Minimum_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Minimum_grad/tuple/control_dependency_1前缀/优化器/梯度/AddN_1前缀/优化器/梯度/丢失/集线器丢失/Abs_grad/Sign前缀/优化器/梯度/丢失/集线器丢失/Abs_grad/mulprefix/optimizer/gradients/loss/huber_loss/Sub_grad/Negprefix/optimizer/gradients/loss/huber_loss/Sub_grad/tuple/group_depsprefix/optimizer/gradients/loss/huber_loss/Sub_grad/tuple/control_dependency前缀/优化器/梯度/丢失/集线器丢失/Sub_grad/tuple/control_dependency_1prefix/optimizer/gradients/predicted_Q_val/Reshape_grad/Shapeprefix/optimizer/gradients/predicted_Q_val/Reshape_grad/Reshapeprefix/optimizer/gradients/predicted_Q_val/GatherNd_grad/Shape前缀/优化器/梯度/预测的Q_val/GatherNd_grad/SscatterNdprefix/optimizer/gradients/model_net/fully_connected_1/BiasAdd_grad/BiasAndGrad前缀/优化器/梯度/model_net/fully_connected_1/BiasAdd_grad/tuple/group_depsprefix/optimizer/gradients/model_net/fully_connected_1/BiasAdd_grad/tuple/control_dependenceprefix/optimizer/gradients/model_net/fully_connected_1/BiasAdd_grad/tuple/control_dependency_1prefix/optimizer/gradients/model_net/fully_connected_1/MatMul_grad/MatMulprefix/optimizer/gradients/model_net/fully_connected_1/MatMul_grad/MatMul_1prefix/optimizer/gradients/model_net/fully_connected_1/MatMul_grad/tuple/group_depsprefix/optimizer/gradients/model_net/fully_connected_1/MatMul_grad/tuple/control_dependenceprefix/optimizer/gradients/model_net/fully_connected_1/MatMul_grad/tuple/control_dependency_1prefix/optimizer/gradients/model_net/fully_connected/Relu_grad/ReluGradprefix/optimizer/gradients/model_net/fully_connected/BiasAdd_grad/BiasAddressGradprefix/optimizer/gradients/model_net/fully_connected/BiasAdd_grad/tuple/group_depsprefix/optimizer/gradients/model_net/fully_connected/BiasAdd_grad/tuple/control_dependencyprefix/optimizer/gradients/model_net/fully_connected/BiasAdd_grad/tuple/control_dependency_1prefix/optimizer/gradients/model_net/fully_connected/MatMul_grad/MatMul前缀/优化器/梯度/模型集/完整连接/MatMul_grad/MatMul_1prefix/optimizer/gradients/model_net/fully_connected/MatMul_grad/tuple/group_depsprefix/optimizer/gradients/model_net/fully_connected/MatMul_grad/tuple/control_dependencyprefix/optimizer/gradients/model_net/fully_connected/MatMul_grad/tuple/control_dependency_1prefix/优化器/GradientDescent/learning_rateprefix/optimizer/GradientDescent/update_model_net/fully_connected/weights/ApplyGradientDescendprefix/optimizer/GradientDescent/update_model_net/fully_connected/bias/ApplyGradientDescendprefix/optimizer/GradientDescent/update_model_net/fully_connected_1/weights/ApplyGradientDescenseprefix/optimizer/GradientDescent/update_model_net/fully_connected_1/bias/ApplyGradientDescenseprefix/优化器/GradientDescentprefix/hard_replace/分配prefix/hard_replace/Assign_1prefix/hard_replace/Assign_2prefix/hard_replace/Assign_3prefix/soft_replace/mul/x前缀/soft_replace/mul前缀/soft_replace/mul_1/x前缀/soft_replace/mul_1prefix/soft_replace/addprefix/soft_replace/分配前缀/soft_replace/mul_2/x前缀/soft_replace/mul_2前缀/soft_replace/mul_3/x前缀/soft_replace/mul_3前缀/soft_replace/add_1prefix/soft_replace/Assign_1前缀/soft_replace/mul_4/x前缀/soft_replace/mul_4prefix/soft_replace/mul_5/x前缀/soft_replace/mul_5前缀/soft_replace/add_2prefix/soft_replace/Assign_2前缀/soft_replace/mul_6/x前缀/soft_replace/mul_6前缀/soft_replace/mul_7/x前缀/soft_replace/mul_7前缀/soft_replace/add_3prefix/soft_replace/Assign_3前缀/init前缀/分配前缀/分配_1前缀/分配_2前缀/分配_3prefix/save/filenname/inputprefix/save/filenameprefix/save/Constprefix/save/SaveV2/tensor_namesprefix/save/SaveV2/shape_and_sciesprefix/save/SaveV2前缀/保存/控制相关性prefix/save/RestoreV2/tensor_namesprefix/save/RestoreV2/shape_and_schiesprefix/save/RestoreV2前缀/保存/分配前缀/保存/分配_1prefix/save/Assign_2prefix/save/Assign_3prefix/save/Assign_4prefix/save/Assign_5prefix/save/Assign_6prefix/save/Assign_7prefix/save/restore_all''
我认为问题在于您定义占位符的方式。您可以参考相关页面：https://www.tensorflow.org/api_docs/python/tf/compat/v1/placeholder