nano_llm

nano_llm

CONTAINERS IMAGES RUN BUILD

Note

NanoLLM is a lightweight, optimized library for LLM inference and multimodal agents. For more info, see these resources:

• Repo - github.com/dusty-nv/NanoLLM
• Docs - dusty-nv.github.io/NanoLLM
• Jetson AI Lab - Live Llava, NanoVLM, SLM

CONTAINERS

nano_llm:main
Aliases	nano_llm
Requires	L4T ['>=35']
Dependencies	build-essential cuda:12.2 cudnn:8.9 python numpy cmake onnx pytorch:2.2 torchvision huggingface_hub rust transformers awq mlc riva-client:python tensorrt opencv gstreamer jetson-inference torch2trt torchaudio onnxruntime piper-tts whisper_trt cuda-python faiss faiss_lite clip_trt nanodb
Dockerfile	Dockerfile

CONTAINER IMAGES

Repository/Tag	Date	Arch	Size
dustynv/nano_llm:24.4-r35.4.1	2024-04-15	arm64	8.5GB
dustynv/nano_llm:24.4-r36.2.0	2024-04-15	arm64	9.7GB
dustynv/nano_llm:24.4.1-r35.4.1	2024-04-19	arm64	10.2GB
dustynv/nano_llm:24.4.1-r36.2.0	2024-04-19	arm64	11.1GB
dustynv/nano_llm:24.5-r35.4.1	2024-05-03	arm64	10.2GB
dustynv/nano_llm:24.5-r36.2.0	2024-05-03	arm64	11.2GB
dustynv/nano_llm:24.5.1-r35.4.1	2024-05-19	arm64	10.0GB
dustynv/nano_llm:24.5.1-r36.2.0	2024-05-19	arm64	10.6GB
dustynv/nano_llm:24.6-r36.2.0	2024-06-09	arm64	10.8GB
dustynv/nano_llm:r35.4.1	2024-05-15	arm64	10.0GB
dustynv/nano_llm:r36.2.0	2024-06-29	arm64	10.8GB

_{Container images are compatible with other minor versions of JetPack/L4T:
    • L4T R32.7 containers can run on other versions of L4T R32.7 (JetPack 4.6+)
    • L4T R35.x containers can run on other versions of L4T R35.x (JetPack 5.1+)}

RUN CONTAINER

To start the container, you can use jetson-containers run and autotag, or manually put together a docker run command:

# automatically pull or build a compatible container image
jetson-containers run $(autotag nano_llm)

# or explicitly specify one of the container images above
jetson-containers run dustynv/nano_llm:r36.2.0

# or if using 'docker run' (specify image and mounts/ect)
sudo docker run --runtime nvidia -it --rm --network=host dustynv/nano_llm:r36.2.0

^{jetson-containers run forwards arguments to docker run with some defaults added (like --runtime nvidia, mounts a /data cache, and detects devices)
autotag finds a container image that's compatible with your version of JetPack/L4T - either locally, pulled from a registry, or by building it.}

To mount your own directories into the container, use the -v or --volume flags:

jetson-containers run -v /path/on/host:/path/in/container $(autotag nano_llm)

To launch the container running a command, as opposed to an interactive shell:

jetson-containers run $(autotag nano_llm) my_app --abc xyz

You can pass any options to it that you would to docker run, and it'll print out the full command that it constructs before executing it.

BUILD CONTAINER

If you use autotag as shown above, it'll ask to build the container for you if needed. To manually build it, first do the system setup, then run:

jetson-containers build nano_llm

The dependencies from above will be built into the container, and it'll be tested during. Run it with --help for build options.