Overview
Explore a flexible Machine Learning experiment tracking system for Python coders using DVC and Streamlit in this informative talk from PyCon US. Discover how to navigate the complex landscape of data science tools by leveraging existing technologies like Git for versioning and Python IDEs. Learn to build a customizable system that addresses the unique challenges of ML engineering, including data versioning, exploratory work, and model quality assessment. Follow along as the speaker demonstrates training a neural network with TensorFlow to classify cat and dog images, focusing on the tooling rather than the ML algorithm. Gain insights into using DVC (Data Version Control) for versioning data alongside code, building training pipelines, and tracking experiments. See how Streamlit can be utilized to create data exploration apps for interacting with trained models. Witness code samples and live demos showcasing various ways to combine DVC and Streamlit, including building a Streamlit app that selects and tests any DVC-tracked model based on its Git commit.
Syllabus
Talk - Antoine Toubhans: Flexible ML Experiment Tracking System for Python Coders with DVC and St...
Taught by
PyCon US