Lambda Labs — управление облачными GPU

★ 7.7 · devops

lambda-labs is a Claude Code skill that manages Lambda Labs GPU cloud instances for ML training and inference workloads. It covers the full infrastructure lifecycle: launching and terminating instances via the lambda-cloud-client Python library or curl, managing SSH keys, and attaching persistent filesystems that retain data across restarts. Available GPUs range from V100 to B200 SXM6 with per-minute billing and no egress fees; for large-scale distributed training, 1-Click Clusters provide Slurm-managed setups of 16–512 GPUs connected over InfiniBand. Every instance ships with Lambda Stack pre-installed — Ubuntu 22.04, CUDA 12.x, PyTorch, TensorFlow, JAX, and JupyterLab — so environments are ready without manual setup. The skill is built for teams running long training jobs, multi-node experiments, or dedicated inference workloads across 12+ global regions.