Andromeda Documentation
Guides and reference for running workloads on your Andromeda GPU clusters, from your first login to long-running Slurm and Kubernetes jobs.
Connect over SSH or kubectl, configure Slurm, automate access with the Business and Platform APIs, and monitor every node you reserve.
Start here
# 1. Sign in
# 2. Find your cluster
# 3. Connect and run on all 8 GPUs of a node
From sign-in to
nvidia-smi in about ten minutes. Full walkthrough →GuideRun the nanoGPT speedrunTrain modded-nanogpt on one 8-GPU H100 node as a Slurm batch job, in a CUDA 13 container started through Pyxis.Read the guide →
Common tasks
The pages teams reach for most, across every product.