The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains instructional code that uses the subprocess module to execute system commands.
Examples include running nvidia-smi to verify GPU availability and pip list to inspect the container environment.
It also uses subprocesses for orchestrating distributed machine learning training via torch.distributed.launch.
[EXTERNAL_DOWNLOADS]: The skill defines container images that fetch and install dependencies from external sources.
It installs standard machine learning libraries such as torch, transformers, and accelerate from official registries.
It includes an example of installing the transformers library directly from HuggingFace's official GitHub repository.
[REMOTE_CODE_EXECUTION]: The skill documents the deployment of local Python code to Modal's remote serverless infrastructure.
It describes features for running code in isolated environments using modal.Sandbox, which is a core capability of the infrastructure platform for testing and execution.

modal-serverless-gpu