Scaling Machine Learning Workflows to Big Data with Fugue
CNCF [Cloud Native Computing Foundation] via YouTube
Overview
Explore scaling machine learning workflows to big data using Fugue in this 29-minute conference talk from KubeCon + CloudNativeCon Europe 2022. Learn how to transition from Pandas to distributed computing frameworks like Spark or Dask without reimplementing code. Discover Fugue's open-source abstraction layer that allows data scientists to write framework-agnostic and scale-agnostic code. Follow along as the speakers demonstrate porting native Python code to Spark or Dask with minimal changes, and witness the scaling of data compute from a single machine to a Spark cluster on Kubernetes. Gain insights into lazy evaluation, partitioning, testing, and decoupling logic from execution in big data workflows.
Syllabus
Introduction
Demo Overview
Han Wang Introduction
First Example
Spark
Transformation
Fugue Code
Model
Field Workflow
Results
Physical
Prediction
Pandas vs Spark
Lazy evaluation of Spark
Partitioning
Testing
Fugue
Decouple logic and execution
Demo
Notebook extension
Conclusion
Recap
Taught by
CNCF [Cloud Native Computing Foundation]