... streamlit.

Sometimes you'd like to show a demo of working code without a jupyter notebook around it. The code is useful, but it might scare some stakeholders away. For those moments streamlit can be a great option.


We're using scikit-learn in this demo. Should you be unfamilar with it, feel free to just copy and paste the code below. The goal of the video is to explain streamlit, not that we worry about understanding the algorithm in detail.

This is the app.py that we end up with at the end of the video;

import numpy as np
import streamlit as st
import matplotlib.pylab as plt
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import AdaBoostRegressor

n = 1000
x = np.linspace(0, 6, n)
X = np.linspace(0, 6, n)[:, np.newaxis]
y = np.sin(X).ravel() + np.sin(6 * X).ravel() + np.random.random(n) * 0.3

n_est = st.sidebar.slider("n_est", min_value=1, max_value=5_000, step=1)

def make_predictions(n_est):
    mod1 = DecisionTreeRegressor(max_depth=4)
    y1 = mod1.fit(X,y).predict(X)
    y2 = AdaBoostRegressor(mod1, n_estimators=n_est).fit(X, y).predict(X)
    return y1, y2

y1, y2 = make_predictions(n_est=n_est)

if st.sidebar.checkbox("Toggle ScatterChart"):
    plt.scatter(x, y, alpha=0.1)
plt.plot(x, y1, label="just a tree")
plt.plot(x, y2, label=f"adaboost-{n_est}")


Note that we run this app from the command line by running;

streamlit run app.py

Feedback? See an issue? Something unclear? Feel free to mention it here.

If you want to be kept up to date, consider getting the newsletter.