Skip to content

OpenG2G

GPU-to-Grid Simulation under LLM Workloads

A modular Python framework for studying how GPU-level controls affect distribution-level voltages.

Get Started Library Guide Read the Paper

Datacenter Backends

Trace-replay from real GPU benchmarks, or live GPU power via Zeus

Grid Backends

AC power flow on IEEE test feeders via OpenDSS, with tap control and sensitivity estimation

Controllers

Ships with OFO batch optimization and tap scheduling; subclass to write your own

Multi-Rate Coordinator

Compose components at different timesteps into a single simulation