ai-dynamoai-dynamoOfficialยท1 Agent Skills Included

nixl

Accelerate GPU and storage data transfers for AI inference

Accelerates point-to-point data transfers between CPU, GPU memory, and storage for AI inference frameworks like NVIDIA Dynamo. Eliminates slow manual memory handling with a modular plugin system covering UCX, GDS, POSIX, and object storage backends. Includes Python bindings and benchmarking tools so engineers can build, test, and tune high-speed inference pipelines faster.
npx skills add ai-dynamo/nixl --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install NIXL?โ–ผ

Run `npx skills add ai-dynamo/nixl --all -g -y` in your terminal to install all skills in this suite globally.

What is NIXL used for?โ–ผ

NIXL speeds up point-to-point data movement between CPU, GPU, and storage in AI inference systems like NVIDIA Dynamo, using pluggable backends such as UCX and GDS.

Does NIXL work with Python?โ–ผ

Yes. NIXL provides Python bindings installable via pip, and the included skill helps your AI agent write correct agent, descriptor, and transfer code.

Which platforms does NIXL support?โ–ผ

NIXL runs on Linux only, tested on Ubuntu 22.04/24.04 and Fedora, with support for both NVIDIA CUDA and AMD ROCm GPUs.

Can I benchmark NIXL transfer performance?โ–ผ

Yes. The repository includes NIXLBench and KVBench tools to measure transfer throughput and simulate real inference workloads.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’