qed-software-test-project's Introduction

Representativity predictor

Simple service to predict 'representattivity' of objects.

Initial Configuraiton

Service required docker with docker compose installed on host.

Before first run of service some additional actions must be performed:

1. Prepare image for jobs

Image must be present on host where service is running

$ docker build . -t mlflow-executor:latest

2. Prepare environment file

Fill .env file with required values

$ cp env_example .eenv
$ vi .env

MINIO_ROOT_USER and MINIO_ROOT_PASSWORD should be filled with proper values. Example:

MINIO_ROOT_USER=AKIAIOSFODNN7EXAMPLE
MINIO_ROOT_PASSWORD=wJalrXUtnFEMI/K7MDENG/bPxRfiCYE

Runing

To start service use following command:

$ docker compose up -d

Usage

API use csv files as data source format. Files hadn't contain headers. Symbol "," should be used s delimiter.

To creeate random test data following command can be used:

$ python -c 'import numpy as np;np.savetxt(<FILE_NAME>, np.random.rand(<OBJECT_COUNT>, <OBJECT_DIMENSION>), delimiter=",")'

Start model trainig process

$ curl -L -XPOST 'http://0.0.0.0:5000/model' -F file=@<TRAIN_DATA>.csv
{"model_id":"574a97e860044bd5a9b94f177c6ade7b"}

Check model training process status

$ curl -L 'http://0.0.0.0:5000/model'
# or $ curl -L 'http://0.0.0.0:5000/model/3acc540e0d354d8fb08fed8b40c7c69f'
{"id":"a8ea04a96eae4ea28fcddefd850702b4","progress":0.1,"start_time":"2022-05-05T17:48:50.315000","status":"RUNNING"}

Predict representativity of object

$ curl -L -XPUT 'http://0.0.0.0:5000/model/predict' -F file=@<EVALUATE_DATA>.csv -o result.csv
# or $ curl -L -XPUT 'http://0.0.0.0:5000/model/3acc540e0d354d8fb08fed8b40c7c69f/predict' -F file=@<EVALUATE_DATA>.csv -o result.csv

Recommend Projects