nv-legate/cupynumeric
NumPy and SciPy on Multi-Node Multi-GPU systems
What it solves
It addresses the difficulty of scaling NumPy workflows to high-performance computing environments. Instead of rewriting code for distributed systems or GPUs, users can run their existing NumPy-based data analysis and machine learning workflows on massive hardware clusters.
How it works
cuPyNumeric implements the NumPy API on top of the Legate framework. This allows it to distribute computations across a single CPU, a single GPU, and thousands of GPUs across multiple nodes without requiring significant code changes from the user.
Who it’s for
It is designed for researchers and developers working with large-scale data analysis, complex simulations, and machine learning who need to scale their computations from a single machine to distributed GPU clusters.
Highlights
- High-performance array computing
- Implements the NumPy API for seamless transition
- Scales from single CPU to thousands of GPUs
- Distributed system support across multiple nodes
Related
- Project
- Project
- Project
- Project
- Project