tianle91 / faas

Forests As A Service

Forests As A Service

Easy training and prediction service.

docker compose up

Then navigate to the ui.

Guide

Start by uploading a dataset.

Configure model training by specifying what the target and features columns and setting special columns.

Train a model.

Run predictions.

Evaluate model performance.

faas's Issues

use pyarrow for pandas serialization

monetization

Monetizing API with stripe https://www.youtube.com/watch?v=MbqSMgMAzxU

stop spark session after set time

try storing spark session in session state

no ingestion if app ran for too long

multiple select radio boxes if reupload new data

only one set is active though - the old one is greyed out

comparisons with other models

Resumable training

Clean up guide screenshots by selecting ui elements instead of the entire screen

show off what's possible in the ui

Allow selectable positive transformation

Option box to allow for it

Defaults to true if more than 99% of observations has the same sign and those that don't has less than 1% the magnitude of those that do

support for multiple uploads

Is a pandas implementation faster and easier?

it's also easier to keep things in memory instead of keeping a spark session alive.

drop streamlit and split up the app script into frontend + backend