The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download the official Triton Inference Server container image from the NVIDIA Container Registry (nvcr.io). This is a well-known and trusted source for machine learning infrastructure.
[COMMAND_EXECUTION]: Includes a standard docker run command for launching the server with GPU support and necessary port mappings (8000-8002). The volume mount (/models) is a required configuration for the server to access model files.
[COMMAND_EXECUTION]: Provides instructions to install the tritonclient Python package via pip, which is the standard library for interacting with the inference server.
[COMMAND_EXECUTION]: Suggests using torch.onnx.export for model conversion, which is a standard development practice in the machine learning lifecycle.

triton_inference_server