How-to guides#

This is a page of how-to guides. Guides are basically answers to the “How to … with Cascade” type of questions. They should be practical and may require some background knowledge of Cascade.

If you are new to the library, you may want to start from the Tutorial page.

Pipelines#

Build a pipeline

How to build data processing pipelines with cascade.data module

Track errors in a dataset

How to find the source of and exception deep inside a pipeline

Version data pipeline

How to track pipeline changes, add metadata and avoid version explosions

Experiment tracking#

Track model training experiment

How to conduct experiments with Cascade

Track logs

How to save a log of a machine learning script

Use a Trainer

How to make model training even easier with Trainers

Track a file

How to log files along with saved model

Track an existing functional pipeline

If you already have your Python functions and want to track them

CLI#

How to use Cascade CLI

Some things are more convenient when done in command line

Query experiment results

Just like with database, but without database

Integrations#

scikit-learn

Effectively track your sklearn experiments

PyTorch

To be written soon!

Experiment management#

Link Dataset to a Model

You can link anything to anything basically

Comment on results of an experiment

Filling bare numbers with some meaning and starting discussions

Add Tags

It is about building your own system

Cascade Recipes#

Browse a collection of real-life small projects built with Cascade. Each use-case shows how Cascade can be used with full working code examples.

01 - Track ML experiments with Cascade