Model Service use process

Use process

The process of using the ModelService is as follows. First, initialize the Xiaomi Cloud-ML client environment.

cloudml init

Then upload the model file to FDS or export the model directly to the FDS in the TensorFlow code.

<br />Finally, use the Cloud-ML command-line tool to submit your creation to the Model Service. The "-a" parameter can also be used to load multiple model versions simultaneously.

cloudml models create -n linear -v v1 -u fds://cloud-ml/linear -a ""

<br />Once the model is launched, you can directly view the model creation's status and log information.

cloudml models events linear v1

cloudml models logs linear v1 ```

Note that the current Model Service is publicly accessible.

Parameters introduction

  • -n is a mandatory parameter allowing users to choose the model name.
  • -v is a mandatory parameter allowing users to choose the model version. The name and version together indicate the Model Service instance.
  • -u is a mandatory parameter that indicates the path of the model file.
  • -a is an optional parameter that allows users to import any customized parameter when starting the Model Service.

Additional features

The Model Service also supports such functions as multiple copies, online upgrades and can continue to read later documents.