The Agent Skills Directory

[SAFE]: No security issues were detected in this skill. The instructions are purely technical and focus on refactoring training code for distributed environments. All external links point to official documentation from the PyTorch organization (docs.pytorch.org) or well-known, high-reputation repositories (github.com/pytorch/examples, github.com/pytorch/torchtitan). The use of standard environment variables like LOCAL_RANK and the recommended execution of torchrun are standard, legitimate practices for distributed model training and do not involve unauthorized privilege escalation or dangerous execution patterns.

pytorch-fsdp2