Plugin for outlier detection's monitoring

Structure

/profiler-fe - frontend for plugin

/profiler - backend for plugin

Dockerfiles

profiler/Dockerfile

Dockerfile for backend part of profiler. This image can be used as part of Hydrosphere architecture or as independent project composed with profiler-fe.

Environment variables:

• INDEPENDENT_PROFILER_MODE - bool by default False
• AWS_ACCESS_KEY_ID - string
• AWS_SECRET_ACCESS_KEY - string
• POSTGRES_USER - string, default "root"
• POSTGRES_PASSWORD - string, default "root"
• POSTGRES_DB- string, default "profiler_plugin"
• POSTGRES_HOST - string, default "db"
• POSTGRES_PORT - number, default "5432"

Endpoints /static - endpoint for serving static frontend files(need for micro-frontend). By default service listens to 5000 port.

/docs - OpenApi

profiler-fe/Dockerfile

Dockerfile for frontend part of profiler.

DashboardModule - exposed as part of micro-frontend architecture. Can be used as independent project composed with profiler backend.

docker-compose.yml

Used for creating independent profiler application

How to

Run whole project(with monitoring-manager)

Start

• docker compose up -d - start application

Upload data

• open minio http://localhost:9001/ (login: minioadmin, pswd: minioadmin)
• create bucket adult
• to adult bucket add training bucket
• into training bucket upload /demo/dummy_model/train.csv
• to adult bucket add inference bucket
• • into inference bucket upload /demo/dummy_model/batch_1.csv

Register model

• POST http://localhost:8080/api/v1/model with body

{
    "name": "adult",
    "version": 1,
    "signature": {
        "inputs": [
            {
                "name": "age",
                "shape": [],
                "dtype": "DT_INT64",
                "profile": "NUMERICAL"
            },
            {
                "name": "workclass",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "fnlwgt",
                "shape": [],
                "dtype": "DT_INT64",
                "profile": "NUMERICAL"
            },
            {
                "name": "education",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "educational-num",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "marital-status",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "occupation",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "relationship",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "race",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "gender",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            },
            {
                "name": "capital-gain",
                "shape": [],
                "dtype": "DT_INT64",
                "profile": "NUMERICAL"
            },
            {
                "name": "hours-per-week",
                "shape": [],
                "dtype": "DT_INT64",
                "profile": "NUMERICAL"
            },
            {
                "name": "native-country",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            }
        ],
        "outputs": [
            {
                "name": "income",
                "shape": [],
                "dtype": "DT_STRING",
                "profile": "CATEGORICAL"
            }
        ]
    },
    "metadata": {},
    "trainingDataPrefix": "s3://adult/training/train.csv",
    "inferenceDataPrefix": "s3://adult/inference"
}

Open browser

• http://localhost:4200/models

Run independent project (Demo purposes)

• cd demo
• docker compose up -d
• open http://localhost/models
• From dummy_model model use files to upload model(with contract(contract.json) and training data(train.csv))
• Press Load data to upload inference data (from dummy_model batch_*.csv file)
• Click on any models row to go to dashboard