Serving: Serve Multiple Versions simultaneously Error

[brianschardt]

I am trying to serve multiple version of a model at the same time.
Mt goal is to get both these API endpoints working at the same time
http://localhost:8501/v1/models/xor/versions/3:predict
http://localhost:8501/v1/models/xor/versions/4:predict

right now only the higher number version will work so that is the one with version 4. How would I get both to work at the same time?

[chrisolston]

By default, only the latest version is served. To serve multiple versions simultaneously you'll have to change the "version policy" configuration. See https://github.com/tensorflow/serving/blob/master/tensorflow_serving/config/model_server_config.proto#L48

[cchung100m]

Hi @brianalois,

You can assign the version policy when starting the server.
Please check the issue I had encountered. #918

Hope helps,

[brianschardt]

Thanks that Solves the Issue just to have it in this answer as well, the config file that should look something like this.
you must add this to the specific model config:

model_version_policy: {
        all: {}
    }

So an example model.config would look like this

model_config_list: {
  config: {
    name:  "xor",
    base_path:  "/home/models/xor",
    model_platform: "tensorflow",
    model_version_policy: {
        all: {}
    }
  }
}