The Agent Skills Directory

[SAFE]: The skill documents the use of official NVIDIA tools and libraries for LLM optimization.\n- [EXTERNAL_DOWNLOADS]: Fetches official Docker images from NVIDIA's repository (nvidia/tensorrt_llm).\n- [EXTERNAL_DOWNLOADS]: Installs the tensorrt_llm package from the official Python package registry.\n- [EXTERNAL_DOWNLOADS]: References and downloads LLM models from Meta's official HuggingFace repository (meta-llama/Meta-Llama-3-8B).\n- [COMMAND_EXECUTION]: Provides standard command-line examples for starting inference servers and running benchmarks using the trtllm-serve utility and standard Python scripts.\n- [CREDENTIALS_UNSAFE]: References API key authentication for production serving but correctly uses environment variable placeholders ($API_KEY) instead of hardcoded secrets.

tensorrt-llm