lambda-labs-gpu-cloud

Provision GPU cloud instances for distributed ML training by launching on-demand Lambda Labs machines, attaching persistent filesystems, and using Slurm-based multi-node clustering to scale workloads across nodes.

Yes, SSH access is fully supported for Lambda Labs GPU instances, allowing you to directly log into the provisioned machines, configure environments, and manage long-running ML training jobs or batch inference.

Persistent storage is supported through attachable filesystems, ensuring your datasets and model checkpoints persist across sessions when running long-running ML training experiments on Lambda Labs GPU cloud resources.

The best way to run multi-node clustering for ML training is using Slurm-based 1-Click Clusters on Lambda Labs, which provides region-aware GPU selection and pre-installed Lambda Stack software for scalable workloads.

Pre-installed Lambda Stack software is available on provisioned GPU cloud instances to reduce setup friction, providing the necessary frameworks and dependencies to immediately run ML training and inference workloads.