User-Guide

Overview

Model Management Service works with AIML Framework to manage the life cycle of trained AIML models, such as creating a model, storing the trained model, storing the trained model info. It exposes REST based API to work with models.

Steps to build and run Model Management Service Standalone

Prerequisites

  1. Install go

Steps

 git clone "https://gerrit.o-ran-sc.org/r/aiml-fw/awmf/modelmgmtservice.git"
cd modelmgmtservice
Update ENV variables in config.env
Execute below commands
export $(< ./config.env)
go get
go build -o mme_bin .
./mme_bin

Steps to run Model Management Service using AIMLFW deployment scripts

Follow the steps in this link: AIMLFW installation guide

APIs and samples

  1. Registering a model in Model Management Service Sample model-name value is “qos_301”

    curl  -i  -H "Content-Type: application/json"  \
        -X POST \
        -d '{"model-name":"qos_301", "rapp-id": "rapp_1", "meta-info" : {"accuracy":"90","model-type":"timeseries","feature-list":["pdcpBytesDl","pdcpBytesUl"]}}' \
        http://127.0.0.1:32006/registerModel

  2. Fetch trained model information from Model Management Service

    curl -X GET  http://127.0.0.1:32006/getModelInfo/qos_301

  3. Upload a trained AIML Model to Model Management Service

    curl -F "file=@<MODEL_ZIP_FILE_NAME>" http://127.0.0.1:32006/uploadModel/qos_301

  4. Download a trained model from Model Management Service

    curl -X GET http://127.0.0.1:32006/downloadModel/qos_301/model.zip --output model.zip