cuda

cuda

CONTAINERS IMAGES RUN BUILD

CONTAINERS

cuda:12.2
Builds
Requires	L4T ==35.*
Dependencies	build-essential
Dependants	cuda:12.2-samples
Dockerfile	Dockerfile
Images	dustynv/cuda:12.2-r36.2.0 (2023-12-05, 3.4GB) dustynv/cuda:12.2-samples-r36.2.0 (2023-12-07, 4.8GB)

cuda:12.2-samples
Builds
Requires	L4T ==35.*
Dependencies	build-essential cuda:12.2 python cmake
Dockerfile	Dockerfile.samples
Images	dustynv/cuda:12.2-samples-r36.2.0 (2023-12-07, 4.8GB)
Notes	CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples

cuda:11.8
Requires	L4T ==35.*
Dependencies	build-essential
Dependants	cuda:11.8-samples
Dockerfile	Dockerfile

cuda:11.8-samples
Requires	L4T ==35.*
Dependencies	build-essential cuda:11.8 python cmake
Dockerfile	Dockerfile.samples
Notes	CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples

cuda:11.4
Aliases	cuda
Requires	L4T <36
Dependencies	build-essential
Dependants	arrow:12.0.1 arrow:14.0.1 arrow:5.0.0 audiocraft auto_awq auto_gptq awq awq:dev bitsandbytes cuda-python cuda:11.4-samples cudf:21.10.02 cudf:23.10.03 cudnn cudnn:8.9 cuml cupy deepstream efficientvit exllama:v1 exllama:v2 faiss faiss:lite gptq-for-llama gstreamer jetson-inference jetson-utils l4t-diffusion l4t-ml l4t-pytorch l4t-tensorflow:tf1 l4t-tensorflow:tf2 l4t-text-generation langchain langchain:samples llama_cpp:ggml llama_cpp:gguf llava local_llm minigpt4 mlc:3feed05 mlc:3feed05-builder mlc:51fb0f4 mlc:51fb0f4-builder mlc:5584cac mlc:5584cac-builder mlc:9bf5723 mlc:9bf5723-builder mlc:dev mlc:dev-builder nanodb nanoowl nanosam nemo numba onnxruntime opencv:4.5.0 opencv:4.5.0-builder opencv:4.8.1 opencv:4.8.1-builder optimum pycuda pytorch:1.10 pytorch:1.11 pytorch:1.12 pytorch:1.13 pytorch:1.9 pytorch:2.0 pytorch:2.0-distributed pytorch:2.1 pytorch:2.1-builder pytorch:2.1-distributed raft realsense ros:foxy-desktop ros:foxy-ros-base ros:foxy-ros-core ros:galactic-desktop ros:galactic-ros-base ros:galactic-ros-core ros:humble-desktop ros:humble-ros-base ros:humble-ros-core ros:iron-desktop ros:iron-ros-base ros:iron-ros-core ros:melodic-desktop ros:melodic-ros-base ros:melodic-ros-core ros:noetic-desktop ros:noetic-ros-base ros:noetic-ros-core sam stable-diffusion stable-diffusion-webui tam tensorflow tensorflow2 tensorrt tensorrt:8.6 text-generation-inference text-generation-webui:1.7 text-generation-webui:6a7cd01 text-generation-webui:main torch2trt torch_tensorrt torchaudio torchvision transformers transformers:git transformers:nvgpt tvm whisper whisperx xformers zed

cuda:11.4-samples
Aliases	cuda:samples
Requires	L4T <36
Dependencies	build-essential cuda:11.4 python cmake
Dockerfile	Dockerfile.samples
Notes	CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples

CONTAINER IMAGES

Repository/Tag	Date	Arch	Size
dustynv/cuda:12.2-r36.2.0	2023-12-05	arm64	3.4GB
dustynv/cuda:12.2-samples-r36.2.0	2023-12-07	arm64	4.8GB

_{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 the run.sh/autotag helpers or manually put together a docker run command:

# automatically pull or build a compatible container image
./run.sh $(./autotag cuda)

# or explicitly specify one of the container images above
./run.sh dustynv/cuda:12.2-samples-r36.2.0

# or if using 'docker run' (specify image and mounts/ect)
sudo docker run --runtime nvidia -it --rm --network=host dustynv/cuda:12.2-samples-r36.2.0

^{run.sh 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:

./run.sh -v /path/on/host:/path/in/container $(./autotag cuda)

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

./run.sh $(./autotag cuda) my_app --abc xyz

You can pass any options to run.sh 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:

./build.sh cuda

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