datacraft.py

import numpy as np
import streamlit as st

from utils.functions import (
    add_polynomial_features,
    generate_data,
    get_model_tips,
    get_model_url,
    plot_decision_boundary_and_metrics,
    train_model,
)

from utils.ui import (
    dataset_selector,
    footer,
    generate_snippet,
    polynomial_degree_selector,
    introduction,
    model_selector,
)

st.set_page_config(
    page_title="Docker Workshop", layout="wide", page_icon="./images/flask.png"
)

st.image(["./images/ekidatacraft.png"],width=300)

def sidebar_controllers():
    dataset, n_samples, train_noise, test_noise, n_classes = dataset_selector()
    model_type, model = model_selector()
    x_train, y_train, x_test, y_test = generate_data(
        dataset, n_samples, train_noise, test_noise, n_classes
    )
    st.sidebar.header("Feature engineering")
    degree = polynomial_degree_selector()
    footer()

    return (
        dataset,
        n_classes,
        model_type,
        model,
        x_train,
        y_train,
        x_test,
        y_test,
        degree,
        train_noise,
        test_noise,
        n_samples,
    )


def body(
    x_train, x_test, y_train, y_test, degree, model, model_type, train_noise, test_noise
):
    introduction()
    col1, col2 = st.beta_columns((1, 1))

    with col1:
        plot_placeholder = st.empty()

    with col2:
        duration_placeholder = st.empty()
        model_url_placeholder = st.empty()
        code_header_placeholder = st.empty()
        snippet_placeholder = st.empty()
        tips_header_placeholder = st.empty()
        tips_placeholder = st.empty()

    x_train, x_test = add_polynomial_features(x_train, x_test, degree)
    model_url = get_model_url(model_type)

    (
        model,
        train_accuracy,
        train_f1,
        test_accuracy,
        test_f1,
        duration,
    ) = train_model(model, x_train, y_train, x_test, y_test)

    metrics = {
        "train_accuracy": train_accuracy,
        "train_f1": train_f1,
        "test_accuracy": test_accuracy,
        "test_f1": test_f1,
    }

    snippet = generate_snippet(
        model, model_type, n_samples, train_noise, test_noise, dataset, degree
    )

    model_tips = get_model_tips(model_type)

    fig = plot_decision_boundary_and_metrics(
        model, x_train, y_train, x_test, y_test, metrics
    )

    plot_placeholder.plotly_chart(fig, True)
    duration_placeholder.warning(f"Training took {duration:.3f} seconds")
    model_url_placeholder.markdown(model_url)
    code_header_placeholder.header("**Retrain the same model in Python**")
    snippet_placeholder.code(snippet)
    tips_header_placeholder.header(f"**Tips on the {model_type} 💡 **")
    tips_placeholder.info(model_tips)


if __name__ == "__main__":
    (
        dataset,
        n_classes,
        model_type,
        model,
        x_train,
        y_train,
        x_test,
        y_test,
        degree,
        train_noise,
        test_noise,
        n_samples,
    ) = sidebar_controllers()
    body(
        x_train,
        x_test,
        y_train,
        y_test,
        degree,
        model,
        model_type,
        train_noise,
        test_noise,
    )