together-gpu-clusters

Provision GPU clusters on Together AI by using the Python or TypeScript SDKs to specify your region, GPU type, and cluster size. You can also use the tcloud CLI, Terraform, SkyPilot, or REST API.

Together AI HPC workloads support NVIDIA H100, H200, B200, L40, and RTX-6000 GPUs. You can orchestrate your clusters using either Kubernetes or Slurm, enabling autoscaling, health checks, and multi-node management.

Yes, you can manage on-demand GPU capacity for ML pipelines using Terraform or SkyPilot. The platform also supports the tcloud CLI, REST API, and Python/TypeScript SDKs for cluster management and accessibility.

Yes, Together AI supports shared volumes with InfiniBand and nvme-backed storage for high-throughput ML pipelines. This enables scalable multi-node distributed training and batch inference.

The best way to run large-scale distributed training is provisioning on-demand or reserved GPU clusters via Together AI. You can orchestrate multi-node management and autoscaling using Kubernetes or Slurm.

Together AI batch inference and HPC simulations rely on available on-demand or reserved capacity for specified NVIDIA GPU types. Users must manage cluster orchestration through provided tools like Kubernetes or Slurm.