如何构建用于Tensorflow服务的models.config

Question

I want to host multiple models on tensorflow serving . Tensorflow serving documents states that : Create a file containing an ASCII ModelServerConfig protocol buffer. But I am not able to build this models.config . Please help me with this

Answer 1

As mentioned in the documentation you are referring to, you need to create a plaintext file in the Google ProtoBuf structure. Then, you save this file, and you pass a model_config_file flag to the model server executable with the path to the file you have just created. If you are running docker, there is a section in GtHub README dedicated to that.