Huggingface Accelerate

Wraps HuggingFace's Accelerate library to handle distributed training without rewriting your PyTorch code. You add four lines (import Accelerator, initialize it, wrap your model/optimizer/dataloader) and suddenly your script runs on multiple GPUs, TPUs, or mixed precision setups. Useful when you've outgrown single-GPU training but don't want to dive into the complexity of torch.distributed directly. The skill passed Gen Agent Trust Hub and Socket audits but got a warning from Snyk, so check dependencies if security matters for your deployment. It's been installed 262 times and comes from orchestra-research's AI research skill collection.

Install to Claude Code

npx -y skills add orchestra-research/ai-research-skills --skill huggingface-accelerate --agent claude-code