Benchmark tooling for common models and operations
- Mentors
- Brian Chen, Kyle Daruwalla
- Organization
- NumFOCUS
- Technologies
- julia, GitHub Actions
- Topics
- machine learning, devops
FluxML is a deep learning stack built in 100% Julia code, which aims to provide high-speed and light-weight abstraction of GPU, auto-differentiation and deep learning model composites. This project will introduce a specific benchmarking system to adapt to the current requirements of the FluxML community. The system includes GitHub Actions workflows that are triggered by specific events, a FluxML-specific benchmarking tool and a repository for benchmarking code.