libra

Why we included this project

Libra shrinks the usual machine learning workflow down to a few commands a non-specialist can actually follow. You hand it a tabular dataset, ask it to model a target column in plain English, and it does the preprocessing, builds the network, trains, and plots results, then hands back the model, metrics, and an interpretation in one object. For analysts, product people, and small teams that want a quick baseline without setting up a PyTorch or Keras project from scratch, that is a genuinely useful shortcut. Engineers who get promising first results can still reach past the wrapper to tune the underlying scikit-learn and Keras parameters, since everything runs through the same client object. It won't replace a serious production deep-learning stack, but for everyday tabular experiments it saves real time.