NVIDIAOfficial·26 Agent Skills Included

TensorRT-LLM

Optimize, test, and serve LLM inference on NVIDIA GPUs

Accelerates large language model inference on NVIDIA GPUs with optimized kernels, serving configs, and benchmarking tools. Guides AI agents through model onboarding, AutoDeploy graph transforms, accuracy debugging, and performance profiling without manual trial and error. Automates test execution, Slurm cluster jobs, Docker builds, and kernel optimization so engineers ship faster inference with less guesswork.

14,522|2,712

↗

npx skills add NVIDIA/TensorRT-LLM --all -g -y

Available:

Gives the AI agent the repository's architecture map, coding rules, test and serve commands, and commit conventions so it can modify and validate TensorRT-LLM code correctly.

All Skills in This Repository (26)

Pure Emerald Level Indicators

📦 In Repo

NVIDIA

ad-pipeline-failure-pr

Analyze GitLab AutoDeploy pipeline failures and group issues into root-cause buckets.

ad-model-onboard

Translate HuggingFace models into prefill-only AutoDeploy custom models with equivalence tests.

perf-optimize

Iteratively applies and verifies TensorRT-LLM serving optimizations through benchmark-gated optimization rounds.

perf-analyze

Diagnoses TensorRT-LLM serving deployments via benchmarking, SOL projection, and kernel-level profiling.

exec-local-docker

Executes TensorRT-LLM workloads in local Docker containers with background monitoring and log-based result reporting.

exec-local-slurm

Submit and monitor Slurm jobs on a local cluster with persistent node allocation reuse.

ad-sharding-ir-port

Adds sharding-aware IR hints to TensorRT-LLM AutoDeploy custom model files in place.

ad-graph-dump

Enable and interpret TensorRT-LLM AutoDeploy FX graph dumps via AD_DUMP_GRAPHS_DIR.

ad-conf-check

Verifies AutoDeploy YAML config application by cross-referencing server logs and graph dumps.

ad-add-fusion-transformation

Implement and register AutoDeploy fusion transform passes in TensorRT-LLM graph pipelines.

trtllm-case-executor

Executes TensorRT-LLM tests, benchmarks, and evaluations locally or via Slurm job submission.

perf-optimization

Coordinates GPU performance optimization by routing tasks to kernel and profiling specialists.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install TensorRT-LLM skills?▼

Run `npx skills add NVIDIA/TensorRT-LLM --all -g -y` in your terminal to install the full skill suite globally.

What do the TensorRT-LLM agent skills do?▼

They let an AI agent run tests, benchmarks, and evaluations, onboard new HuggingFace models, debug accuracy regressions, and optimize GPU kernels inside a TensorRT-LLM checkout.

Can AI agents profile and fix slow LLM inference?▼

Yes. The perf skills analyze nsys and ncu traces, classify bottlenecks as compute, memory, or host-bound, and apply proven optimizations like CUDA graphs and kernel fusion.

Do these skills work on Slurm GPU clusters?▼

Yes. Dedicated skills handle environment detection, container builds, and job submission on local or remote Slurm clusters, including multi-node runs.

Which AI agents are compatible with these skills?▼

They follow the standard SKILL.md format and work with Claude Code, Codex, and other agents that support skill-based workflows.