Audience
Developers, researchers, and teams running GPU-accelerated machine-learning and other compute-intensive workloads.
About supercomputer
supercomputer is a hosted cloud-compute service that provides GPU and CPU sessions for machine-learning and other compute-intensive workloads. Developers, researchers, and teams can launch and manage sessions from a command-line interface or Python SDK, run code on configurable GPU instances, and preserve workspace files between sessions. The service supports model training and fine-tuning, inference, experiments, evaluations, batch processing, hyperparameter sweeps, and data-processing jobs. Users can upload scripts and datasets, execute commands in persistent sessions, and retrieve results or save them in shared volumes. Compute charges accrue only while sessions are running, and stopped sessions retain their workspace for later use. The service uses prepaid credits and bills compute by the second, with rates varying by GPU configuration.
