What problem does it solve?
Procuring and configuring dedicated GPU hardware for machine learning training is slow and expensive. This Skill provides direct programmatic access to Lambda Labs cloud instances, enabling rapid provisioning of H100, A100, and B200 GPUs with pre-installed ML stacks.
Core Features & Use Cases
- Instance Lifecycle Management: Launch, monitor, and terminate single-GPU or multi-node GPU instances via Python API or curl commands.
- Persistent Storage: Configure NFS filesystems to retain datasets, checkpoints, and models across instance terminations.
- Distributed Training Setup: Deploy PyTorch DDP, FSDP, or DeepSpeed across 1-Click Slurm clusters with InfiniBand networking.
- Use Case: A research team needs to fine-tune a 70B parameter LLM. Use this Skill to programmatically launch an 8x H100 cluster, attach a persistent filesystem for checkpoints, run distributed training, and automatically terminate the instance upon completion to minimize costs.
Quick Start
Launch a 1x H100 instance on Lambda Labs in the us-west-1 region using my SSH key named training-key and attach the dataset-storage filesystem.
