Iris Dataset ML pipeline

Pipeline contains the following stages:

• data preparation (data clean, data split)
• train
• eval
• inference

Build each stage

data preparation: docker build -t iris-prep --target prep -f ./prep_train_eval/Dockerfile

train: docker build -t iris-train --target train -f ./prep_train_eval/Dockerfile

eval: docker build -t iris-eval --target eval -f ./prep_train_eval/Dockerfile

inference: docker build -t iris-inference --target inference -f ./inference/Dockerfile

Run pipeline

First time: docker-compose -f iris-pipeline.yaml up --build (v1.x) docker compose -f iris-pipeline.yaml up --build (v2.x)

Not first time/if build is not required: docker-compose -f iris-pipeline.yaml up (v1.x) docker compose -f iris-pipeline.yaml up (v2.x)

Inference sample requests

curl -X POST localhost:5000/predict -H 'Content-Type: application/json' -d '{"data": [4.0,3.3,1.7,0.5]}' {"class_id":0,"class_name":"Iris-setosa"}

curl -X POST localhost:5000/predict -H 'Content-Type: application/json' -d '{"data": [6.2,2.8,4.8,1.8]}' {"class_id":2,"class_name":"Iris-virginica"}

curl -X POST localhost:5000/predict -H 'Content-Type: application/json' -d '{"data": [7.2,2.8,4.8,1.8]}' {"class_id":1,"class_name":"Iris-versicolor"}

Requirements

• docker/docker compose 1.x, 2.x

Future ideas/tests

• scrips/modules unittests