如何在服务/推理过程中设置云ml-engine的日志

Question

加载检查点后，我导出了模型，然后将saved_model.pb文件和所有变量复制到GCS bucket。

在推理/服务期间，我得到这个错误：

{
  "error": "Prediction failed: Error during model execution: AbortionError(code=StatusCode.FAILED_PRECONDITION, details=\"Attempting to use uninitialized value shadow/LSTMLayers/stack_bidirectional_rnn/cell_0/bidirectional_rnn/fw/basic_lstm_cell/bias\n\t [[Node: shadow/LSTMLayers/stack_bidirectional_rnn/cell_0/bidirectional_rnn/fw/basic_lstm_cell/bias/read = Identity[T=DT_FLOAT, _output_shapes=[[2048]], _device=\"/job:localhost/replica:0/task:0/device:CPU:0\"](shadow/LSTMLayers/stack_bidirectional_rnn/cell_0/bidirectional_rnn/fw/basic_lstm_cell/bias)]]\")"
}

请建议如何在云ml-engine中设置测试/推理时的日志。

Answer 1

如果您已经设置了GCP项目，则可以使用StackDriver Logging API将消息写入日志。然后，您可以从命令行(gcloud logging)或developer console访问日志消息。