PyTorch skill for PyTorch model development, datasets, training loops, distributed training, export, debugging, and production readiness. Use when Codex must inspect an existing repository, plan or implement changes, preserve working behavior, avoid unnecessary dependencies, and validate the result.
