我正在使用keras子类化api实现一个tensorflow模型，如上一篇文章中所述：
Tensorflow Model Subclassing Mutli-Input
我遇到了一个错误，该错误使我的整个系统都无法跟踪，但无法缩小发生此错误的位置，我在代码中的任何地方都没有使用@ tf.function。我正在寻找有关缩小错误位置或解决方法的支持。
确切的错误是：
警告：在标志解析之前记录到stderr。 W0127 15：31：00.867754 4815351232 def_function.py:586]最后6个中的5个 在0x157208d08>处调用.distributed_function触发了tf.function跟踪。跟踪很昂贵，并且 可能由于传递python对象而导致跟踪过多 而不是张量。另外，tf.function有 Experiment_relax_shapes =放宽参数形状的True选项 这样可以避免不必要的追溯。请参阅 https://www.tensorflow.org/tutorials/customization/performance#python_or_tensor_args 和https://www.tensorflow.org/api_docs/python/tf/function了解更多 详细信息。
进一步查看它是否禁用急切执行，则会抛出此警告，可能是导致该错误的原因：
2020-01-27 16：11：42.781884：W tensorflow / c / c_api.cc：326]操作 '{名称：'basic_parsing_model / time_distributed / bidirectional / backward_gru / while' id：542 op设备：{} def：{{{node basic_parsing_model / time_distributed / bidirectional / backward_gru / while}} =虽然[T = [DT_INT32，DT_INT32，DT_INT32，DT_VARIANT，DT_FLOAT，...，DT_VARIANT，DT_VARIANT，DT_VARIANT，DT_VARIANT，DT_VARIANT]， _lower_using_switch_merge = true，_num_original_outputs = 52，body = basic_parsing_model_time_distributed_bidirectional_backward_gru_while_body_708 []， cond = basic_parsing_model_time_distributed_bidirectional_backward_gru_while_cond_707 []， output_shapes = [[]，[]，[]，[]，[?, 32]，...，[]，[]，[]，[]，[]]， parallel_iterations = 32]（basic_parsing_model / time_distributed / bidirectional / backward_gru / while / loop_counter，basic_parsing_model / time_distributed / bidirectional / backward_gru / while / maximum_iterations， basic_parsing_model / time_distributed / bidirectional / backward_gru / time， basic_parsing_model / time_distributed / bidirectional / backward_gru / TensorArrayV2_1， basic_parsing_model / time_distributed / bidirectional / backward_gru / zeros， basic_parsing_model / time_distributed / bidirectional / backward_gru / strided_slice_1， basic_parsing_model / time_distributed / bidirectional / backward_gru / TensorArrayUnstack / TensorListFromTensor， basic_parsing_model / time_distributed / backward_gru / bias， basic_parsing_model / time_distributed / backward_gru / kernel， basic_parsing_model / time_distributed / backward_gru / recurrent_kernel， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_1， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_2， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_3， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_4， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_5， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_6， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_7， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_8， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_9， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_10， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_11， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_12， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_13， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_14， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_15， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_16， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_17， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_18， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_19， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_20， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_21， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_22， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_23， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_24， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_25， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_26， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_27， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_28， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_29， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_30， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_31， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_32， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_33， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_34， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_35， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_36， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_37， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_38， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_39， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_40， basic_parsing_model / time_distributed / bidirectional / backward_gru / while / EmptyTensorList_41）}}' 在会话运行后通过设置属性进行更改。这个 突变将无效，并且将来会触发错误。 在运行节点后不要修改节点或创建新会话。 2020-01-27 16：11：46.854955：W tensorflow / core / kernels / data / generator_dataset_op.cc：103]错误 最终确定GeneratorDataset迭代器时发生：已取消： 操作被取消