# Technical Documentation: dusty-nv/jetson-containers > ℹ️ **Provenance:** Multi-Tier Ingestion · [CodeWiki Reference](https://codewiki.google/github.com/dusty-nv/jetson-containers) ### Overview (`README.md`) [](https://www.jetson-ai-lab.com) [](https://pypi.jetson-ai-lab.io) # CUDA Containers for Edge AI & Robotics Modular container build system that provides the latest [**AI/ML packages**](https://pypi.jetson-ai-lab.io/) for [NVIDIA Jetson](https://jetson-ai-lab.com) :rocket::robot: | | | |--------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | **ML** | [`pytorch`](packages/ml/pytorch) [`tensorflow`](packages/ml/tensorflow) [`jax`](packages/ml/jax) [`onnxruntime`](packages/ml/onnxruntime) [`deepstream`](packages/cv/deepstream) [`holoscan`](packages/cv/holoscan) [`CTranslate2`](packages/ml/ctranslate2) [`JupyterLab`](packages/code/jupyterlab) | | **LLM** | [`SGLang`](packages/llm/sglang) [`vLLM`](packages/llm/vllm) [`MLC`](packages/llm/mlc) [`AWQ`](packages/llm/awq) [`transformers`](packages/llm/transformers) [`text-generation-webui`](packages/llm/text-generation-webui) [`text-generation-inference`](packages/llm/text-generation-inference) [`ollama`](packages/llm/ollama) [`optimum`](packages/llm/optimum) [`openai`](packages/llm/openai)[`gptqmodel`](packages/llm/gptqmodel) [`llama.cpp`](packages/llm/llama_cpp) [`llama-factory`](packages/llm/llama-factory) [`exllama`](packages/llm/exllama) [`ktransformers`](packages/llm/ktransformers) [`FlashAttention`](packages/attention/flash-attention) [`DeepSpeed`](packages/llm/deepspeed) [`bitsandbytes`](packages/llm/bitsandbytes) [`xformers`](packages/attention/xformers) [`xgrammar`](packages/llm/xgrammar) | | **VLM** | [`llava`](packages/vlm/llava) [`llama-vision`](packages/vlm/llama-vision) [`VILA`](packages/vlm/vila) [`LITA`](packages/vlm/lita) [`NanoLLM`](packages/llm/nano_llm) [`Prismatic`](packages/vlm/prismatic) [`xtuner`](packages/vlm/xtuner) [`gemma_vlm`](packages/vlm/gemma_vlm) | | **VIT** | [`NanoOWL`](packages/vit/nanoowl) [`NanoSAM`](packages/vit/nanosam) [`Segment Anything (SAM)`](packages/vit/sam) [`Track Anything (TAM)`](packages/vit/tam) [`clip_trt`](packages/vit/clip_trt) | | **RAG** | [`llama-index`](packages/rag/llama-index) [`langchain`](packages/rag/langchain) [`jetson-copilot`](packages/rag/jetson-copilot) [`NanoDB`](packages/vectordb/nanodb) [`FAISS`](packages/vectordb/faiss) [`RAFT`](packages/ml/rapids/raft) | | **L4T** | [`l4t-pytorch`](packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow`](packages/ml/l4t/l4t-tensorflow) [`l4t-ml`](packages/ml/l4t/l4t-ml) [`l4t-diffusion`](packages/ml/l4t/l4t-diffusion) [`l4t-text-generation`](packages/ml/l4t/l4t-text-generation) | | **CUDA** | [`cupy`](packages/ml/numeric/cupy) [`cuda-python`](packages/cuda/cuda-python) [`pycuda`](packages/cuda/pycuda) [`cv-cuda`](packages/cv/cv-cuda) [`opencv:cuda`](packages/cv/opencv) [`numba`](packages/ml/numeric/numba) | | **Robotics** | [`ROS`](packages/physicalAI/ros) [`LeRobot`](packages/physicalAI/lerobot) [`OpenVLA`](packages/physicalAI/vla/openvla) [`3D Diffusion Policy`](packages/cv/diffusion/3d_diffusion_policy) [`Crossformer`](packages/physicalAI/vla/crossformer) [`MimicGen`](packages/physicalAI/sim/mimicgen) [`OpenDroneMap`](packages/physicalAI/opendronemap) [`ZED`](packages/hw/zed) [`openpi`](packages/physicalAI/openpi) | | **Simulation** | [`Isaac Sim`](packages/physicalAI/sim/isaac-sim) [`Genesis`](packages/physicalAI/sim/genesis) [`Habitat Sim`](packages/physicalAI/sim/habitat-sim) [`MimicGen`](packages/physicalAI/sim/mimicgen) [`MuJoCo`](packages/physicalAI/sim/mujoco) [`PhysX`](packages/physicalAI/sim/physx) [`Protomotions`](packages/physicalAI/protomotions) [`RoboGen`](packages/physicalAI/sim/robogen) [`RoboMimic`](packages/physicalAI/sim/robomimic) [`RoboSuite`](packages/physicalAI/sim/robosuite) [`Sapien`](packages/physicalAI/sim/sapien) | | **Graphics** | [`3D Diffusion Policy`](packages/cv/diffusion/3d_diffusion_policy) [`AI Toolkit`](packages/cv/diffusion/ai-toolkit) [`ComfyUI`](packages/cv/diffusion/comfyui) [`Cosmos`](packages/cv/diffusion/cosmos) [`Diffusers`](packages/cv/diffusion/diffusers) [`Diffusion Policy`](packages/cv/diffusion/diffusion_policy) [`FramePack`](packages/cv/diffusion/framepack) [`Stable Diffusion WebUI`](packages/cv/diffusion/stable-diffusion-webui) [`SD.Next`](packages/cv/diffusion/sdnext) [`nerfstudio`](packages/cv/3d/nerf/nerfstudio) [`meshlab`](packages/cv/3d/3dvision/meshlab) [`gsplat`](packages/cv/3d/gaussian_splatting/gsplat) | | **Mamba** | [`mamba`](packages/ml/mamba) [`mambavision`](packages/ml/mamba/mambavision) [`cobra`](packages/ml/mamba/cobra) [`dimba`](packages/ml/mamba/dimba) [`videomambasuite`](packages/ml/mamba/videomambasuite) | | **KANs** | [`pykan`](packages/ml/kans/pykan) [`kat`](packages/ml/kans/kat) | | **xLTSM** | [`xltsm`](packages/ml/xlstm/xlstm) [`mlstm_kernels`](packages/ml/xlstm/mlstm_kernels) | | **Speech** | [`whisper`](packages/speech/whisper) [`whisper_trt`](packages/speech/whisper_trt) [`faster-whisper`](packages/speech/faster-whisper) [`espeak`](packages/speech/espeak) [`whisperx`](packages/speech/whisperx) [`Chatterbox TTS`](packages/speech/chatterbox-tts) [`piper`](packages/speech/piper1-tts) [`riva`](packages/speech/riva-client) [`speech-dispatcher`](packages/speech/speech-dispatcher) [`speaches`](packages/speech/speaches) [`SparkTTS`](packages/speech/spark-tts) [`riva-client`](packages/speech/riva-client) [`audiocraft`](packages/speech/audiocraft) [`voicecraft`](packages/speech/voicecraft) [`voice-pro`](packages/speech/voice-pro) [`xtts`](packages/speech/xtts) | | **Home/IoT** | [`homeassistant-core`](packages/smart-home/homeassistant-core) [`wyoming-whisper`](packages/smart-home/wyoming/wyoming-whisper) [`wyoming-openwakeword`](packages/smart-home/wyoming/wyoming-openwakeword) [`wyoming-piper`](packages/smart-home/wyoming/wyoming-piper) | | **3DPrintObjects** | [`PartPacker`](packages/cv/3d/3dobjects/partpacker) [`Sparc3D`](packages/cv/3d/3dobjects/sparc3d) | | **Agents** | [`OpenClaw`](packages/agents/openclaw) [`OpenFang`](packages/agents/openfang) [`ZeroClaw`](packages/agents/zeroclaw) | See the [**`packages`**](packages) directory for the full list, including pre-built container images for JetPack/L4T. Using the included tools, you can easily combine packages together for building your own containers. Want to run ROS2 with PyTorch and Transformers? No problem - just do the [system setup](/docs/setup.md), and build it on your Jetson: ```bash $ jetson-containers build --name=my_container pytorch transformers ros:humble-desktop ``` There are shortcuts for running containers too - this will pull or build a [`l4t-pytorch`](packages/ml/l4t/l4t-pytorch) image that's compatible: ```bash $ jetson-containers run $(autotag l4t-pytorch) ``` > [`jetson-containers run`](/docs/run.md) launches [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some added defaults (like `--runtime nvidia`, mounted `/data` cache and devices)
> [`autotag`](/docs/run.md#autotag) finds a container image that's compatible with your version of JetPack/L4T - either locally, pulled from a registry, or by building it. If you look at any package's readme (like [`l4t-pytorch`](packages/ml/l4t/l4t-pytorch)), it will have detailed instructions for running it. #### Changing CUDA Versions You can rebuild the container stack for different versions of CUDA by setting the `CUDA_VERSION` variable: ```bash CUDA_VERSION=12.6 jetson-containers build transformers ``` It will then go off and either pull or build all the dependencies needed, including PyTorch and other packages that would be time-consuming to compile. There is a [Pip server](/docs/build.md#pip-server) that caches the wheels to accelerate builds. You can also request specific versions of cuDNN, TensorRT, Python, and PyTorch with similar environment variables like [here](/docs/build.md#changing-versions). ## Documentation * [Package List](/packages) * [Package Definitions](/docs/packages.md) * [System Setup](/docs/setup.md) * [Building Containers](/docs/build.md) * [Running Containers](/docs/run.md) Check out the tutorials at the [**Jetson Generative AI Lab**](https://www.jetson-ai-lab.com)! ## Getting Started Refer to the [System Setup](/docs/setup.md) page for tips about setting up your Docker daemon and memory/storage tuning. ```bash # install the container tools git clone https://github.com/dusty-nv/jetson-containers bash jetson-containers/install.sh # automatically pull & run any container jetson-containers run $(autotag l4t-pytorch) ``` To also spin up a local [devpi](https://devpi.net) PyPI + APT caching server during install, pass `--pypi`: ```bash sudo bash jetson-containers/install.sh --pypi ``` This launches a containerized devpi instance with per-architecture indexes (jp6, jp7, sbsa, amd64) and an nginx-based APT cache, so subsequent builds resolve wheels and packages locally instead of hitting the network. Or you can manually run a [container image](https://hub.docker.com/r/dustynv) of your choice without using the helper scripts above: ```bash sudo docker run --runtime nvidia -it --rm --network=host dustynv/l4t-pytorch:r36.2.0 ``` Looking for the old jetson-containers? See the [`legacy`](https://github.com/dusty-nv/jetson-containers/tree/legacy) branch. ### Only Tested and supported Jetpack 6.2 (Cuda 12.6) and JetPack 7 (CUDA 13.x). > [!NOTE] > Ubuntu 24.04 containers for JetPack 6 and JetPack 7 are now available (with CUDA support) > >      `LSB_RELEASE=24.04 jetson-containers build pytorch:2.8` >      `jetson-containers run dustynv/pytorch:2.8-r36.4-cu128-24.04` > > ARM SBSA (Server Base System Architecture) is supported for GH200 / GB200. > To install CUDA 13.0 SBSA wheels for Python 3.12 / 24.04: > >      `uv pip install torch torchvision torchaudio \` >             `--index-url https://pypi.jetson-ai-lab.io/sbsa/cu129` > > See the **[`Ubuntu 24.04`](/docs/build.md#2404-containers)** section of the docs for details and a list of available containers 🤗 > Thanks to all our contributors from **[`Discord`](https://discord.gg/BmqNSK4886)** and AI community for their support 🤗 > [!IMPORTANT] > **Jetson Orin now receives ARM SBSA CUDA in CUDA 13.2** — Starting with [CUDA 13.2](https://developer.nvidia.com/blog/cuda-13-2-introduces-enhanced-cuda-tile-support-and-new-python-features/), the same Arm SBSA CUDA Toolkit is supported on Jetson Orin devices, unifying the toolkit across server-class (GH200/GB200) and embedded (Orin) platforms. This requires **JetPack 7.2** with **CUDA driver R595** minimum. ## Code Style The project uses automated code formatting tools to maintain consistent code style. See [Code Style Guide](docs/code-style.md) for details on: - Setting up formatting tools - Adding your package to formatting checks - Troubleshooting common issues ## Troubleshooting ### InvalidDefaultArgInFrom: Default value for ARG ${BASE_IMAGE} results in empty or invalid base image name This may be caused by corrupted caches. Docker caches both image results and build downloads. Try these simple steps first: - Remove non-essential docker images using: **docker image rm --force \** - Prune resulting docker image cache: **docker image prune -f** - Prune docker build image cache: **docker buildx prune -f** ## Gallery > [Multimodal Voice Chat with LLaVA-1.5 13B on NVIDIA Jetson AGX Orin](https://www.youtube.com/watch?v=9ObzbbBTbcc) (container: [`NanoLLM`](https://dusty-nv.github.io/NanoLLM/))
> [Interactive Voice Chat with Llama-2-70B on NVIDIA Jetson AGX Orin](https://www.youtube.com/watch?v=wzLHAgDxMjQ) (container: [`NanoLLM`](https://dusty-nv.github.io/NanoLLM/))
> [Realtime Multimodal VectorDB on NVIDIA Jetson](https://www.youtube.com/watch?v=ayqKpQNd1Jw) (container: [`nanodb`](/packages/vectordb/nanodb))
> [NanoOWL - Open Vocabulary Object Detection ViT](https://www.jetson-ai-lab.com/tutorial_nanoowl.html) (container: [`nanoowl`](/packages/vit/nanoowl)) > [Live Llava on Jetson AGX Orin](https://youtu.be/X-OXxPiUTuU) (container: [`NanoLLM`](https://dusty-nv.github.io/NanoLLM/)) > [Live Llava 2.0 - VILA + Multimodal NanoDB on Jetson Orin](https://youtu.be/X-OXxPiUTuU) (container: [`NanoLLM`](https://dusty-nv.github.io/NanoLLM/)) > [Small Language Models (SLM) on Jetson Orin Nano](https://www.jetson-ai-lab.com/tutorial_slm.html) (container: [`NanoLLM`](https://dusty-nv.github.io/NanoLLM/)) > [Realtime Video Vision/Language Model with VILA1.5-3b](https://www.jetson-ai-lab.com/tutorial_nano-vlm.html#video-sequences) (container: [`NanoLLM`](https://dusty-nv.github.io/NanoLLM/)) ## Citation Please see [CITATION.cff](CITATION.cff) for citation information. --- ### Overview (`packages/README.md`) # Packages > [`ATTENTION`](#attention) [`BUILD`](#build) [`CODE`](#code) [`CUDA`](#cuda) [`CV`](#cv) [`HW`](#hw) [`LLM`](#llm) [`ML`](#ml) [`MULTIMEDIA`](#multimedia) [`NET`](#net) [`PHYSICAL AI`](#physicalai) [`RAG`](#rag) [`SMART-HOME`](#smart-home) [`SPEECH`](#speech) [`VECTORDB`](#vectordb) [`VIT`](#vit) [`VLM`](#vlm) --- ## Attention Attention mechanism implementations and optimizations. | Package | Path | |---------|------| | `ParaAttention` | [`attention/ParaAttention`](/packages/attention/ParaAttention) | | `block-sparse-attention` | [`attention/block-sparse-attention`](/packages/attention/block-sparse-attention) | | `flash-attention` | [`attention/flash-attention`](/packages/attention/flash-attention) | | `flashinfer` | [`attention/flashinfer`](/packages/attention/flashinfer) | | `flexprefill` | [`attention/flexprefill`](/packages/attention/flexprefill) | | `huggingface_kernels` | [`attention/huggingface_kernels`](/packages/attention/huggingface_kernels) | | `jvp-flash-attention` | [`attention/jvp-flash-attention`](/packages/attention/jvp-flash-attention) | | `log-linear-attention` | [`attention/log-linear-attention`](/packages/attention/log-linear-attention) | | `radial-attention` | [`attention/radial-attention`](/packages/attention/radial-attention) | | `sage-attention` | [`attention/sage-attention`](/packages/attention/sage-attention) | | `sparge-attention` | [`attention/sparge-attention`](/packages/attention/sparge-attention) | | `tilelang` | [`attention/tilelang`](/packages/attention/tilelang) | | `xattention` | [`attention/xattention`](/packages/attention/xattention) | | `xformers` | [`attention/xformers`](/packages/attention/xformers) | --- ## Build Build tools, compilers, and development dependencies. | Package | Path | |---------|------| | `bazel` | [`build/bazel`](/packages/build/bazel) | | `build-essential` | [`build/build-essential`](/packages/build/build-essential) | | `cmake:apt` | [`build/cmake/cmake_apt`](/packages/build/cmake/cmake_apt) | | `cmake:pip` | [`build/cmake/cmake_pip`](/packages/build/cmake/cmake_pip) | | `docker` | [`build/docker`](/packages/build/docker) | | `go` | [`build/go`](/packages/build/go) | | `h5py` | [`build/h5py`](/packages/build/h5py) | | `helm` | [`build/helm`](/packages/build/helm) | | `llvm` | [`build/llvm`](/packages/build/llvm) | | `ninja` | [`build/ninja`](/packages/build/ninja) | | `nodejs` | [`build/nodejs`](/packages/build/nodejs) | | `protobuf:apt` | [`build/protobuf/protobuf_apt`](/packages/build/protobuf/protobuf_apt) | | `protobuf:cpp` | [`build/protobuf/protobuf_cpp`](/packages/build/protobuf/protobuf_cpp) | | `pybind11` | [`build/pybind11`](/packages/build/pybind11) | | `python` | [`build/python`](/packages/build/python) | | `rust` | [`build/rust`](/packages/build/rust) | --- ## Code Development environments and IDEs. | Package | Path | |---------|------| | `jupyterlab` | [`code/jupyterlab`](/packages/code/jupyterlab) | | `openhands` | [`code/openhands`](/packages/code/openhands) | | `vscode` | [`code/vscode`](/packages/code/vscode) | --- ## CUDA CUDA toolkit, libraries, and Python bindings. | Package | Path | |---------|------| | `cuda` | [`cuda/cuda`](/packages/cuda/cuda) | | `cuda-python` | [`cuda/cuda-python`](/packages/cuda/cuda-python) | | `cudastack` | [`cuda/cudastack`](/packages/cuda/cudastack) | |    `cudnn_frontend` | [`cuda/cudastack/cudnn_frontend`](/packages/cuda/cudastack/cudnn_frontend) | | `cutlass` | [`cuda/cutlass`](/packages/cuda/cutlass) | | `pycuda` | [`cuda/pycuda`](/packages/cuda/pycuda) | --- ## CV Computer vision, 3D reconstruction, diffusion models, and related packages. ### cv/3d/3dobjects | Package | Path | |---------|------| | `partpacker` | [`cv/3d/3dobjects/partpacker`](/packages/cv/3d/3dobjects/partpacker) | | `sparc3d` | [`cv/3d/3dobjects/sparc3d`](/packages/cv/3d/3dobjects/sparc3d) | ### cv/3d/3dvision | Package | Path | |---------|------| | `glomap` | [`cv/3d/3dvision/glomap`](/packages/cv/3d/3dvision/glomap) | | `hloc` | [`cv/3d/3dvision/hloc`](/packages/cv/3d/3dvision/hloc) | | `kaolin` | [`cv/3d/3dvision/kaolin`](/packages/cv/3d/3dvision/kaolin) | | `meshlab` | [`cv/3d/3dvision/meshlab`](/packages/cv/3d/3dvision/meshlab) | | `nerfacc` | [`cv/3d/3dvision/nerfacc`](/packages/cv/3d/3dvision/nerfacc) | | `open3d` | [`cv/3d/3dvision/open3d`](/packages/cv/3d/3dvision/open3d) | | `pixsfm` | [`cv/3d/3dvision/pixsfm`](/packages/cv/3d/3dvision/pixsfm) | | `polyscope` | [`cv/3d/3dvision/polyscope`](/packages/cv/3d/3dvision/polyscope) | | `pyceres` | [`cv/3d/3dvision/pyceres`](/packages/cv/3d/3dvision/pyceres) | | `pycolmap` | [`cv/3d/3dvision/pycolmap`](/packages/cv/3d/3dvision/pycolmap) | | `pymeshlab` | [`cv/3d/3dvision/pymeshlab`](/packages/cv/3d/3dvision/pymeshlab) | | `tinycudann` | [`cv/3d/3dvision/tinycudann`](/packages/cv/3d/3dvision/tinycudann) | | `usdcore` | [`cv/3d/3dvision/usdcore`](/packages/cv/3d/3dvision/usdcore) | ### cv/3d/gaussian_splatting | Package | Path | |---------|------| | `3dgrut` | [`cv/3d/gaussian_splatting/3dgrut`](/packages/cv/3d/gaussian_splatting/3dgrut) | | `4k4d` | [`cv/3d/gaussian_splatting/4k4d`](/packages/cv/3d/gaussian_splatting/4k4d) | | `easyvolcap` | [`cv/3d/gaussian_splatting/easyvolcap`](/packages/cv/3d/gaussian_splatting/easyvolcap) | | `fast_gauss` | [`cv/3d/gaussian_splatting/fast_gauss`](/packages/cv/3d/gaussian_splatting/fast_gauss) | | `gsplat` | [`cv/3d/gaussian_splatting/gsplat`](/packages/cv/3d/gaussian_splatting/gsplat) | | `nerfview` | [`cv/3d/gaussian_splatting/nerfview`](/packages/cv/3d/gaussian_splatting/nerfview) | ### cv/3d/nerf | Package | Path | |---------|------| | `fruitnerf` | [`cv/3d/nerf/fruitnerf`](/packages/cv/3d/nerf/fruitnerf) | | `nerfstudio` | [`cv/3d/nerf/nerfstudio`](/packages/cv/3d/nerf/nerfstudio) | ### cv/diffusion | Package | Path | |---------|------| | `3d_diffusion_policy` | [`cv/diffusion/3d_diffusion_policy`](/packages/cv/diffusion/3d_diffusion_policy) | | `ai-toolkit` | [`cv/diffusion/ai-toolkit`](/packages/cv/diffusion/ai-toolkit) | | `cache_edit` | [`cv/diffusion/cache_edit`](/packages/cv/diffusion/cache_edit) | | `comfyui` | [`cv/diffusion/comfyui`](/packages/cv/diffusion/comfyui) | | `diffusers` | [`cv/diffusion/diffusers`](/packages/cv/diffusion/diffusers) | | `diffusion_cpp` | [`cv/diffusion/diffusion_cpp`](/packages/cv/diffusion/diffusion_cpp) | | `diffusion_policy` | [`cv/diffusion/diffusion_policy`](/packages/cv/diffusion/diffusion_policy) | | `fastgen` | [`cv/diffusion/fastgen`](/packages/cv/diffusion/fastgen) | | `framepack` | [`cv/diffusion/framepack`](/packages/cv/diffusion/framepack) | | `sdnext` | [`cv/diffusion/sdnext`](/packages/cv/diffusion/sdnext) | | `self-forcing` | [`cv/diffusion/self-forcing`](/packages/cv/diffusion/self-forcing) | | `stable-diffusion-webui` | [`cv/diffusion/stable-diffusion-webui`](/packages/cv/diffusion/stable-diffusion-webui) | ### cv (other) | Package | Path | |---------|------| | `cv-cuda` | [`cv/cv-cuda`](/packages/cv/cv-cuda) | | `deepstream` | [`cv/deepstream`](/packages/cv/deepstream) | | `holoscan` | [`cv/holoscan`](/packages/cv/holoscan) | | `jetson-inference` | [`cv/jetson-inference`](/packages/cv/jetson-inference) | | `opencv` | [`cv/opencv`](/packages/cv/opencv) | | `vpi` | [`cv/vpi`](/packages/cv/vpi) | --- ## HW Hardware interfaces and device drivers. | Package | Path | |---------|------| | `canable` | [`hw/canable`](/packages/hw/canable) | | `canbus` | [`hw/canbus`](/packages/hw/canbus) | | `cangaroo` | [`hw/cangaroo`](/packages/hw/cangaroo) | | `jetcam` | [`hw/jetcam`](/packages/hw/jetcam) | | `jupyter_clickable_image_widget` | [`hw/jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) | | `librealsense` | [`hw/librealsense`](/packages/hw/librealsense) | | `oled` | [`hw/oled`](/packages/hw/oled) | | `pl2303` | [`hw/pl2303`](/packages/hw/pl2303) | | `zed` | [`hw/zed`](/packages/hw/zed) | --- ## LLM Large language model inference, training, quantization, and serving frameworks. | Package | Path | |---------|------| | `awq` | [`llm/awq`](/packages/llm/awq) | | `bitsandbytes` | [`llm/bitsandbytes`](/packages/llm/bitsandbytes) | | `deepspeed` | [`llm/deepspeed`](/packages/llm/deepspeed) | | `dynamo` | [`llm/dynamo/dynamo`](/packages/llm/dynamo/dynamo) | |    `kai-scheduler` | [`llm/dynamo/kai-scheduler`](/packages/llm/dynamo/kai-scheduler) | |    `mooncake` | [`llm/dynamo/mooncake`](/packages/llm/dynamo/mooncake) | |    `nixl` | [`llm/dynamo/nixl`](/packages/llm/dynamo/nixl) | | `exllama` | [`llm/exllama`](/packages/llm/exllama) | | `gptqmodel` | [`llm/gptqmodel`](/packages/llm/gptqmodel) | | `huggingface_hub` | [`llm/huggingface_hub`](/packages/llm/huggingface_hub) | | `hymba` | [`llm/hymba`](/packages/llm/hymba) | | `ktransformers` | [`llm/ktransformers`](/packages/llm/ktransformers) | | `litellm` | [`llm/litellm`](/packages/llm/litellm) | | `llama-factory` | [`llm/llama-factory`](/packages/llm/llama-factory) | | `llama_cpp` | [`llm/llama_cpp`](/packages/llm/llama_cpp) | | `llamaspeak` | [`llm/llamaspeak`](/packages/llm/llamaspeak) | | `lobe_chat` | [`llm/lobe_chat`](/packages/llm/lobe_chat) | | `local_llm` | [`llm/local_llm`](/packages/llm/local_llm) | | `minference` | [`llm/minference`](/packages/llm/minference) | | `mistral_common` | [`llm/mistral/mistral_common`](/packages/llm/mistral/mistral_common) | | `mistral_rs` | [`llm/mistral/mistral_rs`](/packages/llm/mistral/mistral_rs) | | `mlc` | [`llm/mlc`](/packages/llm/mlc) | | `nano_llm` | [`llm/nano_llm`](/packages/llm/nano_llm) | | `nemo` | [`llm/nemo`](/packages/llm/nemo) | | `nvidia-modelopt` | [`llm/tensorrt_optimizer/nvidia-modelopt`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) | | `ollama` | [`llm/ollama`](/packages/llm/ollama) | | `open-webui` | [`llm/open-webui`](/packages/llm/open-webui) | | `openai` | [`llm/openai`](/packages/llm/openai) | | `optimum` | [`llm/optimum`](/packages/llm/optimum) | | `sglang` | [`llm/sglang`](/packages/llm/sglang) | | `sudonim` | [`llm/sudonim`](/packages/llm/sudonim) | | `tensorrt_llm` | [`llm/tensorrt_optimizer/tensorrt_llm`](/packages/llm/tensorrt_optimizer/tensorrt_llm) | | `text-generation-inference` | [`llm/text-generation-inference`](/packages/llm/text-generation-inference) | | `text-generation-webui` | [`llm/text-generation-webui`](/packages/llm/text-generation-webui) | | `transformers` | [`llm/transformers`](/packages/llm/transformers) | | `unsloth` | [`llm/unsloth`](/packages/llm/unsloth) | | `vllm` | [`llm/vllm`](/packages/llm/vllm) | | `xgrammar` | [`llm/xgrammar`](/packages/llm/xgrammar) | --- ## ML Machine learning frameworks, PyTorch ecosystem, numeric libraries, and RAPIDS. ### ml/pytorch | Package | Path | |---------|------| | `pytorch` | [`ml/pytorch`](/packages/ml/pytorch) | | `apex` | [`ml/pytorch/apex`](/packages/ml/pytorch/apex) | | `torch2trt` | [`ml/pytorch/torch2trt`](/packages/ml/pytorch/torch2trt) | | `torch3d` | [`ml/pytorch/torch3d`](/packages/ml/pytorch/torch3d) | | `torch_tensorrt` | [`ml/pytorch/torch_tensorrt`](/packages/ml/pytorch/torch_tensorrt) | | `torchao` | [`ml/pytorch/torchao`](/packages/ml/pytorch/torchao) | | `torchaudio` | [`ml/pytorch/torchaudio`](/packages/ml/pytorch/torchaudio) | | `torchcodec` | [`ml/pytorch/torchcodec`](/packages/ml/pytorch/torchcodec) | | `torchsaver` | [`ml/pytorch/torchsaver`](/packages/ml/pytorch/torchsaver) | | `torchsde` | [`ml/pytorch/torchsde`](/packages/ml/pytorch/torchsde) | | `torchtext` | [`ml/pytorch/torchtext`](/packages/ml/pytorch/torchtext) | | `torchvision` | [`ml/pytorch/torchvision`](/packages/ml/pytorch/torchvision) | ### ml/tensorflow | Package | Path | |---------|------| | `tensorflow` | [`ml/tensorflow`](/packages/ml/tensorflow) | | `tensorboard` | [`ml/tensorflow/tensorboard`](/packages/ml/tensorflow/tensorboard) | ### ml/jax | Package | Path | |---------|------| | `jax` | [`ml/jax`](/packages/ml/jax) | ### ml/mamba | Package | Path | |---------|------| | `causalconv1d` | [`ml/mamba/causalconv1d`](/packages/ml/mamba/causalconv1d) | | `cobra` | [`ml/mamba/cobra`](/packages/ml/mamba/cobra) | | `dimba` | [`ml/mamba/dimba`](/packages/ml/mamba/dimba) | | `mamba` | [`ml/mamba/mamba`](/packages/ml/mamba/mamba) | | `mambavision` | [`ml/mamba/mambavision`](/packages/ml/mamba/mambavision) | | `videomambasuite` | [`ml/mamba/videomambasuite`](/packages/ml/mamba/videomambasuite) | | `zigma` | [`ml/mamba/zigma`](/packages/ml/mamba/zigma) | ### ml/numeric | Package | Path | |---------|------| | `arrow` | [`ml/numeric/arrow`](/packages/ml/numeric/arrow) | | `cupy` | [`ml/numeric/cupy`](/packages/ml/numeric/cupy) | | `numba` | [`ml/numeric/numba`](/packages/ml/numeric/numba) | | `numpy` | [`ml/numeric/numpy`](/packages/ml/numeric/numpy) | | `warp` | [`ml/numeric/warp`](/packages/ml/numeric/warp) | ### ml/rapids | Package | Path | |---------|------| | `cudf` | [`ml/rapids/cudf`](/packages/ml/rapids/cudf) | | `cuml` | [`ml/rapids/cuml`](/packages/ml/rapids/cuml) | | `raft` | [`ml/rapids/raft`](/packages/ml/rapids/raft) | ### ml/kans | Package | Path | |---------|------| | `kat` | [`ml/kans/kat`](/packages/ml/kans/kat) | | `pykan` | [`ml/kans/pykan`](/packages/ml/kans/pykan) | ### ml/xlstm | Package | Path | |---------|------| | `mlstm_kernels` | [`ml/xlstm/mlstm_kernels`](/packages/ml/xlstm/mlstm_kernels) | | `pltsm` | [`ml/xlstm/pltsm`](/packages/ml/xlstm/pltsm) | | `xlstm` | [`ml/xlstm/xlstm`](/packages/ml/xlstm/xlstm) | ### ml/apache | Package | Path | |---------|------| | `tvm` | [`ml/apache/tvm`](/packages/ml/apache/tvm) | | `tvm-ffi` | [`ml/apache/tvm-ffi`](/packages/ml/apache/tvm-ffi) | ### ml (other) | Package | Path | |---------|------| | `aim` | [`ml/aim`](/packages/ml/aim) | | `ctranslate2` | [`ml/ctranslate2`](/packages/ml/ctranslate2) | | `dli-nano-ai` | [`ml/dli/dli-nano-ai`](/packages/ml/dli/dli-nano-ai) | | `onnx` | [`ml/onnx`](/packages/ml/onnx) | | `onnxruntime` | [`ml/onnxruntime`](/packages/ml/onnxruntime) | | `onnxruntime_genai` | [`ml/onnxruntime_genai`](/packages/ml/onnxruntime_genai) | | `transformer-engine` | [`ml/transformer-engine`](/packages/ml/transformer-engine) | | `triton` | [`ml/triton`](/packages/ml/triton) | | `tritonserver` | [`ml/tritonserver`](/packages/ml/tritonserver) | ### ml/l4t | Package | Path | |---------|------| | `l4t-diffusion` | [`ml/l4t/l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) | | `l4t-dynamo` | [`ml/l4t/l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) | | `l4t-ml` | [`ml/l4t/l4t-ml`](/packages/ml/l4t/l4t-ml) | | `l4t-pytorch` | [`ml/l4t/l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) | | `l4t-tensorflow` | [`ml/l4t/l4t-tensorflow`](/packages/ml/l4t/l4t-tensorflow) | | `l4t-text-generation` | [`ml/l4t/l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) | --- ## Multimedia Media processing, video codecs, and streaming. | Package | Path | |---------|------| | `decord` | [`multimedia/decord`](/packages/multimedia/decord) | | `ffmpeg` | [`multimedia/ffmpeg`](/packages/multimedia/ffmpeg) | | `gstreamer` | [`multimedia/gstreamer`](/packages/multimedia/gstreamer) | | `jetson-utils` | [`multimedia/jetson-utils`](/packages/multimedia/jetson-utils) | | `libcom` | [`multimedia/libcom`](/packages/multimedia/libcom) | | `opengl` | [`multimedia/opengl`](/packages/multimedia/opengl) | | `pyav` | [`multimedia/pyav`](/packages/multimedia/pyav) | | `sound-utils` | [`multimedia/sound-utils`](/packages/multimedia/sound-utils) | | `video-codec-sdk` | [`multimedia/video-codec-sdk`](/packages/multimedia/video-codec-sdk) | | `vulkan` | [`multimedia/vulkan`](/packages/multimedia/vulkan) | --- ## Net Networking, proxies, and self-hosted infrastructure. | Package | Path | |---------|------| | `devpi` | [`net/devpi`](/packages/net/devpi) | | `https-portal` | [`net/https-portal`](/packages/net/https-portal) | | `mkdocs` | [`net/mkdocs`](/packages/net/mkdocs) | | `nginx_proxy_manager` | [`net/nginx_proxy_manager`](/packages/net/nginx_proxy_manager) | | `portainer` | [`net/portainer`](/packages/net/portainer) | | `pypi.dev` | [`net/pypi.dev`](/packages/net/pypi.dev) | | `tailscale` | [`net/tailscale`](/packages/net/tailscale) | | `yacht` | [`net/yacht`](/packages/net/yacht) | --- ## PhysicalAI Robotics, simulation, embodied AI, and vision-language-action models. ### physicalAI/sim | Package | Path | |---------|------| | `genesis` | [`physicalAI/sim/genesis`](/packages/physicalAI/sim/genesis) | |    `quadrants` | [`physicalAI/sim/genesis/quadrants`](/packages/physicalAI/sim/genesis/quadrants) | |    `splashSurf` | [`physicalAI/sim/genesis/splashSurf`](/packages/physicalAI/sim/genesis/splashSurf) | |    `vtk` | [`physicalAI/sim/genesis/vtk`](/packages/physicalAI/sim/genesis/vtk) | | `habitat-sim` | [`physicalAI/sim/habitat-sim`](/packages/physicalAI/sim/habitat-sim) | | `isaac-sim` | [`physicalAI/sim/isaac-sim`](/packages/physicalAI/sim/isaac-sim) | | `mimicgen` | [`physicalAI/sim/mimicgen`](/packages/physicalAI/sim/mimicgen) | | `mujoco` | [`physicalAI/sim/mujoco`](/packages/physicalAI/sim/mujoco) | | `newton` | [`physicalAI/sim/newton`](/packages/physicalAI/sim/newton) | | `physx` | [`physicalAI/sim/physx`](/packages/physicalAI/sim/physx) | | `robogen` | [`physicalAI/sim/robogen`](/packages/physicalAI/sim/robogen) | | `robomimic` | [`physicalAI/sim/robomimic`](/packages/physicalAI/sim/robomimic) | | `robosuite` | [`physicalAI/sim/robosuite`](/packages/physicalAI/sim/robosuite) | | `sapien` | [`physicalAI/sim/sapien`](/packages/physicalAI/sim/sapien) | ### physicalAI/vla | Package | Path | |---------|------| | `crossformer` | [`physicalAI/vla/crossformer`](/packages/physicalAI/vla/crossformer) | | `isaac-gr00t` | [`physicalAI/vla/isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) | | `octo` | [`physicalAI/vla/octo`](/packages/physicalAI/vla/octo) | | `openvla` | [`physicalAI/vla/openvla`](/packages/physicalAI/vla/openvla) | | `robopoint` | [`physicalAI/vla/robopoint`](/packages/physicalAI/vla/robopoint) | ### physicalAI/cosmos | Package | Path | |---------|------| | `cosmos-policy` | [`physicalAI/cosmos/cosmos-policy`](/packages/physicalAI/cosmos/cosmos-policy) | | `cosmos-predict2` | [`physicalAI/cosmos/cosmos-predict2`](/packages/physicalAI/cosmos/cosmos-predict2) | | `cosmos-reason1` | [`physicalAI/cosmos/cosmos-reason1`](/packages/physicalAI/cosmos/cosmos-reason1) | | `cosmos-transfer1` | [`physicalAI/cosmos/cosmos-transfer1`](/packages/physicalAI/cosmos/cosmos-transfer1) | | `cosmos_diffusion_renderer` | [`physicalAI/cosmos/cosmos_diffusion_renderer`](/packages/physicalAI/cosmos/cosmos_diffusion_renderer) | ### physicalAI (other) | Package | Path | |---------|------| | `Isaac-GR00T` | [`physicalAI/Isaac-GR00T`](/packages/physicalAI/Isaac-GR00T) | | `isaac-ros` | [`physicalAI/isaac-ros`](/packages/physicalAI/isaac-ros) | | `lerobot` | [`physicalAI/lerobot`](/packages/physicalAI/lerobot) | | `opendronemap` | [`physicalAI/opendronemap`](/packages/physicalAI/opendronemap) | | `openpi` | [`physicalAI/openpi`](/packages/physicalAI/openpi) | | `protomotions` | [`physicalAI/protomotions`](/packages/physicalAI/protomotions) | | `ros` | [`physicalAI/ros`](/packages/physicalAI/ros) | --- ## RAG Retrieval-augmented generation and knowledge tools. | Package | Path | |---------|------| | `graphiti` | [`rag/graphiti`](/packages/rag/graphiti) | | `jetson-copilot` | [`rag/jetson-copilot`](/packages/rag/jetson-copilot) | | `langchain` | [`rag/langchain`](/packages/rag/langchain) | | `llama-index` | [`rag/llama-index`](/packages/rag/llama-index) | | `memvid` | [`rag/memvid`](/packages/rag/memvid) | | `n8n` | [`rag/n8n`](/packages/rag/n8n) | --- ## Smart-Home Home automation, voice assistants, and Wyoming protocol servers. | Package | Path | |---------|------| | `homeassistant-base` | [`smart-home/homeassistant-base`](/packages/smart-home/homeassistant-base) | | `homeassistant-core` | [`smart-home/homeassistant-core`](/packages/smart-home/homeassistant-core) | | `ciso8601` | [`smart-home/dependencies/ciso8601`](/packages/smart-home/dependencies/ciso8601) | | `psutil-home-assistant` | [`smart-home/dependencies/psutil-home-assistant`](/packages/smart-home/dependencies/psutil-home-assistant) | | `wyoming-assist-microphone` | [`smart-home/wyoming/wyoming-assist-microphone`](/packages/smart-home/wyoming/wyoming-assist-microphone) | | `wyoming-openwakeword` | [`smart-home/wyoming/wyoming-openwakeword`](/packages/smart-home/wyoming/wyoming-openwakeword) | | `wyoming-piper` | [`smart-home/wyoming/wyoming-piper`](/packages/smart-home/wyoming/wyoming-piper) | | `wyoming-whisper` | [`smart-home/wyoming/wyoming-whisper`](/packages/smart-home/wyoming/wyoming-whisper) | --- ## Speech Speech recognition, text-to-speech, and audio processing. | Package | Path | |---------|------| | `audiocraft` | [`speech/audiocraft`](/packages/speech/audiocraft) | | `chatterbox-tts` | [`speech/chatterbox-tts`](/packages/speech/chatterbox-tts) | | `espeak` | [`speech/espeak`](/packages/speech/espeak) | | `faster-whisper` | [`speech/faster-whisper`](/packages/speech/faster-whisper) | | `kokoro-tts-fastapi` | [`speech/kokoro-tts/kokoro-tts-fastapi`](/packages/speech/kokoro-tts/kokoro-tts-fastapi) | | `kokoro-tts-hf` | [`speech/kokoro-tts/kokoro-tts-hf`](/packages/speech/kokoro-tts/kokoro-tts-hf) | | `kokoro-tts-onnx` | [`speech/kokoro-tts/kokoro-tts-onnx`](/packages/speech/kokoro-tts/kokoro-tts-onnx) | | `piper1-tts` | [`speech/piper1-tts`](/packages/speech/piper1-tts) | | `riva-client` | [`speech/riva-client`](/packages/speech/riva-client) | | `spark-tts` | [`speech/spark-tts`](/packages/speech/spark-tts) | | `speaches` | [`speech/speaches`](/packages/speech/speaches) | | `speech-dispatcher` | [`speech/speech-dispatcher`](/packages/speech/speech-dispatcher) | | `voice-pro` | [`speech/voice-pro`](/packages/speech/voice-pro) | | `voicecraft` | [`speech/voicecraft`](/packages/speech/voicecraft) | | `whisper` | [`speech/whisper`](/packages/speech/whisper) | | `whisper_trt` | [`speech/whisper_trt`](/packages/speech/whisper_trt) | | `whisperx` | [`speech/whisperx`](/packages/speech/whisperx) | | `xtts` | [`speech/xtts`](/packages/speech/xtts) | --- ## VectorDB Vector databases and similarity search. | Package | Path | |---------|------| | `faiss` | [`vectordb/faiss`](/packages/vectordb/faiss) | | `faiss_lite` | [`vectordb/faiss_lite`](/packages/vectordb/faiss_lite) | | `nanodb` | [`vectordb/nanodb`](/packages/vectordb/nanodb) | --- ## VIT Vision transformer models and segmentation. | Package | Path | |---------|------| | `clip_trt` | [`vit/clip_trt`](/packages/vit/clip_trt) | | `efficientvit` | [`vit/efficientvit`](/packages/vit/efficientvit) | | `nanoowl` | [`vit/nanoowl`](/packages/vit/nanoowl) | | `nanosam` | [`vit/nanosam`](/packages/vit/nanosam) | | `sam` | [`vit/sam`](/packages/vit/sam) | | `sapiens` | [`vit/sapiens`](/packages/vit/sapiens) | | `tam` | [`vit/tam`](/packages/vit/tam) | --- ## VLM Vision-language models for multimodal understanding. | Package | Path | |---------|------| | `gemma_vlm` | [`vlm/gemma_vlm`](/packages/vlm/gemma_vlm) | | `lita` | [`vlm/lita`](/packages/vlm/lita) | | `llama-vision` | [`vlm/llama-vision`](/packages/vlm/llama-vision) | | `llava` | [`vlm/llava`](/packages/vlm/llava) | | `minigpt4` | [`vlm/minigpt4`](/packages/vlm/minigpt4) | | `prismatic` | [`vlm/prismatic`](/packages/vlm/prismatic) | | `videollama` | [`vlm/videollama`](/packages/vlm/videollama) | | `vila` | [`vlm/vila`](/packages/vlm/vila) | | `vila-microservice` | [`vlm/vila-microservice`](/packages/vlm/vila-microservice) | | `xtuner` | [`vlm/xtuner`](/packages/vlm/xtuner) | --- ### Overview (`deprecated/auto_awq/README.md`) # auto_awq > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`auto_awq:0.2.4`** | | | :-- | :-- | |    Aliases | `auto_awq` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`cuda`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/onnx) [`pytorch:2.2`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag auto_awq) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host auto_awq:35.2.1 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag auto_awq) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag auto_awq) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build auto_awq ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/gptq-for-llama/README.md`) # gptq-for-llama > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`gptq-for-llama`** | | | :-- | :-- | |    Builds | [](https://github.com/dusty-nv/jetson-containers/actions/workflows/gptq-for-llama_jp60.yml) [](https://github.com/dusty-nv/jetson-containers/actions/workflows/gptq-for-llama_jp51.yml) | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`cuda`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/onnx) [`pytorch:2.2`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/gptq-for-llama:r35.2.1`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) `(2023-12-06, 5.9GB)`
[`dustynv/gptq-for-llama:r35.3.1`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) `(2023-12-14, 5.9GB)`
[`dustynv/gptq-for-llama:r35.4.1`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) `(2023-12-15, 5.9GB)`
[`dustynv/gptq-for-llama:r36.2.0`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) `(2024-01-02, 7.6GB)` | |    Notes | https://github.com/oobabooga/GPTQ-for-LLaMa |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/gptq-for-llama:r35.2.1`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) | `2023-12-06` | `arm64` | `5.9GB` | |   [`dustynv/gptq-for-llama:r35.3.1`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) | `2023-12-14` | `arm64` | `5.9GB` | |   [`dustynv/gptq-for-llama:r35.4.1`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) | `2023-12-15` | `arm64` | `5.9GB` | |   [`dustynv/gptq-for-llama:r36.2.0`](https://hub.docker.com/r/dustynv/gptq-for-llama/tags) | `2024-01-02` | `arm64` | `7.6GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag gptq-for-llama) # or explicitly specify one of the container images above jetson-containers run dustynv/gptq-for-llama:r36.2.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/gptq-for-llama:r36.2.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag gptq-for-llama) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag gptq-for-llama) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build gptq-for-llama ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/piper-tts/README.md`) # piper-tts > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`piper-tts`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchaudio`](/packages/pytorch/torchaudio) [`tensorrt`](/packages/cuda/tensorrt) [`onnxruntime`](/packages/ml/onnxruntime) [`espeak`](/packages/speech/espeak) | |    Dependants | [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`speaches`](/packages/speech/speaches) [`wyoming-piper:1.6.2`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-piper:master`](/packages/smart-home/wyoming/wyoming-piper) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/piper-tts:r35.4.1`](https://hub.docker.com/r/dustynv/piper-tts/tags) `(2024-04-07, 5.5GB)`
[`dustynv/piper-tts:r36.2.0`](https://hub.docker.com/r/dustynv/piper-tts/tags) `(2024-04-07, 6.7GB)`
[`dustynv/piper-tts:r36.4.0`](https://hub.docker.com/r/dustynv/piper-tts/tags) `(2025-02-19, 6.3GB)`
[`dustynv/piper-tts:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/piper-tts/tags) `(2025-03-03, 5.6GB)` |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/piper-tts:r35.4.1`](https://hub.docker.com/r/dustynv/piper-tts/tags) | `2024-04-07` | `arm64` | `5.5GB` | |   [`dustynv/piper-tts:r36.2.0`](https://hub.docker.com/r/dustynv/piper-tts/tags) | `2024-04-07` | `arm64` | `6.7GB` | |   [`dustynv/piper-tts:r36.4.0`](https://hub.docker.com/r/dustynv/piper-tts/tags) | `2025-02-19` | `arm64` | `6.3GB` | |   [`dustynv/piper-tts:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/piper-tts/tags) | `2025-03-03` | `arm64` | `5.6GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag piper-tts) # or explicitly specify one of the container images above jetson-containers run dustynv/piper-tts:r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/piper-tts:r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag piper-tts) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag piper-tts) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build piper-tts ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/shape-llm/README.md`) # shape-llm > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`shape-llm`** | | | :-- | :-- | |    Aliases | `shape_llm` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) [`cuda:12.2`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.2`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`bitsandbytes`](/packages/llm/bitsandbytes) [`flash-attention`](/packages/attention/flash-attention) [`opencv:4.11.0`](/packages/opencv) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag shape-llm) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host shape-llm:36.3.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag shape-llm) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag shape-llm) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build shape-llm ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/stable-diffusion/README.md`) # stable-diffusion > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build) Generate images from text (txt2img) or from other images (img2img) * stable-diffusion: https://github.com/CompVis/stable-diffusion (`/opt/stable-diffusion`) * with memory optimizations: https://github.com/basujindal/stable-diffusion (`/opt/stable-diffusion/optimizedSD`) * tested on `stable-diffusion-1.4` model: https://huggingface.co/CompVis/stable-diffusion-v-1-4-original See the [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) container for a faster implementation with a web interface. ### txt2img Download the [stable-diffusion-1.4](https://huggingface.co/CompVis/stable-diffusion-v-1-4-original) model (`sd-v1-4.ckpt`) ```bash wget https://huggingface.co/CompVis/stable-diffusion-v-1-4-original/resolve/main/sd-v1-4.ckpt -O /data/models/stable-diffusion/sd-v1-4.ckpt ``` Then run this in the container to generate images (by default, six 512x512 images with 50 refinement steps) ```bash cd /opt/stable-diffusion && python3 scripts/txt2img.py --plms \ --ckpt /data/models/stable-diffusion/sd-v1-4.ckpt \ --outdir /data/images/stable-diffusion \ --prompt "a photograph of an astronaut riding a horse" ``` See here for options: https://github.com/CompVis/stable-diffusion#reference-sampling-script For just one 512x512 image with 25 steps: ```bash cd /opt/stable-diffusion && python3 scripts/txt2img.py --plms \ --n_samples 1 --n_iter 1 --ddim_steps 25 \ --ckpt /data/models/stable-diffusion/sd-v1-4.ckpt \ --outdir /data/images/stable-diffusion \ --prompt "two robots walking in the woods" ``` * Change the image resolution with `--W` and `--H` (the default is 512x512) * Change the `--seed` to have the images be different (the default seed is 42) For Jetson Orin Nano and reduced memory usage: ```bash cd /opt/stable-diffusion && python3 optimizedSD/optimized_txt2img.py \ --sampler plms --seed 42 \ --n_samples 1 --n_iter 1 --ddim_steps 25 \ --ckpt /data/models/stable-diffusion/sd-v1-4.ckpt \ --outdir /data/images/stable-diffusion \ --prompt "a photograph of an astronaut riding a horse" ``` To run all these steps from a script, see [`stable-diffusion/test.sh`](/packages/diffusion/stable-diffusion/test.sh)
CONTAINERS
| **`stable-diffusion`** | | | :-- | :-- | |    Builds | [](https://github.com/dusty-nv/jetson-containers/actions/workflows/stable-diffusion_jp51.yml) | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`cuda`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/onnx) [`pytorch:2.2`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) | |    Dependants | [`l4t-diffusion`](/packages/l4t/l4t-diffusion) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/stable-diffusion:r35.2.1`](https://hub.docker.com/r/dustynv/stable-diffusion/tags) `(2023-12-14, 6.1GB)`
[`dustynv/stable-diffusion:r35.3.1`](https://hub.docker.com/r/dustynv/stable-diffusion/tags) `(2023-12-12, 6.1GB)`
[`dustynv/stable-diffusion:r35.4.1`](https://hub.docker.com/r/dustynv/stable-diffusion/tags) `(2023-12-15, 6.1GB)` | |    Notes | disabled on JetPack 4 |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/stable-diffusion:r35.2.1`](https://hub.docker.com/r/dustynv/stable-diffusion/tags) | `2023-12-14` | `arm64` | `6.1GB` | |   [`dustynv/stable-diffusion:r35.3.1`](https://hub.docker.com/r/dustynv/stable-diffusion/tags) | `2023-12-12` | `arm64` | `6.1GB` | |   [`dustynv/stable-diffusion:r35.4.1`](https://hub.docker.com/r/dustynv/stable-diffusion/tags) | `2023-12-15` | `arm64` | `6.1GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag stable-diffusion) # or explicitly specify one of the container images above jetson-containers run dustynv/stable-diffusion:r35.4.1 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/stable-diffusion:r35.4.1 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag stable-diffusion) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag stable-diffusion) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build stable-diffusion ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`.github/workflows/scripts/README.md`) # GitHub Actions Workflow Scripts This directory contains reusable scripts for GitHub Actions workflows in the jetson-containers project. ## Scripts ### Core Scripts - **`step-pre-checkout-cleanup.sh`** - Cleans up permission-restricted files before checkout - **`step-system-info.sh`** - Displays Jetson system information - **`step-env-info.sh`** - Records comprehensive environment variables and system info - **`step-clean-environment.sh`** - Cleans up Docker, cache, and permission-restricted directories - **`step-git-status.sh`** - Shows git repository state and recent history ### Build Scripts - **`step-build-package.sh `** - Builds and tests a specific package - **`step-analyze-results.sh`** - Analyzes build results and outputs status - **`step-test-results-summary.sh`** - Generates a comprehensive test results summary ### Workflow Generation - **`workflow-template.yml`** - Template for creating new package workflow files - **`generate-workflow.sh [platform]`** - Generates a new workflow file from template ## Usage ### Creating a New Workflow To create a new workflow for a package and platform: ```bash cd .github/workflows/scripts ./generate-workflow.sh vllm orin ./generate-workflow.sh vllm thor ``` This will create: - `../workflows/pr-on-dev_vllm_orin.yml` - `../workflows/pr-on-dev_vllm_thor.yml` ### Creating Multiple Workflows ```bash cd .github/workflows/scripts # For Orin platform ./generate-workflow.sh vllm orin ./generate-workflow.sh cudnn orin ./generate-workflow.sh nccl orin # For Thor platform ./generate-workflow.sh vllm thor ./generate-workflow.sh cudnn thor ./generate-workflow.sh nccl thor ``` ### Platform Support - **`orin`** - Jetson Orin (default if platform not specified) - **`thor`** - Jetson Thor - **Any other platform** - Just specify the platform name ### Manual Workflow Creation 1. Copy `workflow-template.yml` to `../workflows/pr-on-dev__.yml` 2. Replace all instances of `PACKAGE_NAME` with the actual package name 3. Replace all instances of `PLATFORM` with the actual platform name 4. Update the workflow name as needed ## Script Parameters ### step-build-package.sh - **Required**: `` - The package to build and test - **Example**: `./step-build-package.sh vllm` ### step-analyze-results.sh - **Required**: ` ` - **Optional**: `[failure_phase] [failure_stage] [failure_component]` ### step-test-results-summary.sh - **Required**: ` ` - **Optional**: `[failure_phase] [failure_stage] [failure_component]` ## Benefits - **Reusable**: Same scripts work for all packages - **Maintainable**: Update scripts once, affects all workflows - **Consistent**: All workflows have the same behavior and output format - **Easy to create**: Generate new workflows with a single command --- ### Overview (`deprecated/cuda/cuda-python/README.md`) # cuda-python > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cuda-python:12.2`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) [`cuda:12.2`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.4`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu124`](/packages/cuda/cuda) [`cuda:12.4`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.6`** | | | :-- | :-- | |    Aliases | `cuda-python` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dependants | [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`cudf:23.10.03`](/packages/ml/rapids/cudf) [`cudf:25.08.00`](/packages/ml/rapids/cudf) [`cuml`](/packages/ml/rapids/cuml) [`cutlass:4.1.0`](/packages/cuda/cutlass) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faiss_lite`](/packages/vectordb/faiss_lite) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-index:samples`](/packages/rag/llama-index) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`minference:0.1.7`](/packages/llm/minference) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`raft`](/packages/ml/rapids/raft) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`sudonim:hf`](/packages/llm/sudonim) [`tam`](/packages/vit/tam) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vscode:transformers`](/packages/code/vscode) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtuner`](/packages/vlm/xtuner) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.8`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu128`](/packages/cuda/cuda) [`cuda:12.8`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.9`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu129`](/packages/cuda/cuda) [`cuda:12.9`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:13.0`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu130`](/packages/cuda/cuda) [`cuda:13.0`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/cuda-python:builder-r35.4.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2024-03-26` | `arm64` | `5.1GB` | |   [`dustynv/cuda-python:builder-r36.2.0`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2024-03-11` | `arm64` | `3.6GB` | |   [`dustynv/cuda-python:r35.2.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-12-06` | `arm64` | `5.0GB` | |   [`dustynv/cuda-python:r35.3.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-08-29` | `arm64` | `5.0GB` | |   [`dustynv/cuda-python:r35.4.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-12-06` | `arm64` | `5.0GB` | |   [`dustynv/cuda-python:r36.2.0`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-12-06` | `arm64` | `3.5GB` | |   [`dustynv/cuda-python:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2025-03-03` | `arm64` | `2.3GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cuda-python) # or explicitly specify one of the container images above jetson-containers run dustynv/cuda-python:r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/cuda-python:r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cuda-python) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cuda-python) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cuda-python ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/cuda/README.md`) # cuda > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cuda:12.2`** | | | :-- | :-- | |    Requires | `L4T ['==36.*', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.2`](/packages/cuda/cuda-python) [`cuda:12.2-samples`](/packages/cuda/cuda) [`cudnn:8.9`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/cuda:12.2-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) `(2023-12-05, 3.4GB)`
[`dustynv/cuda:12.2-samples-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) `(2023-12-07, 4.8GB)` | | **`pip_cache:cu122`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | | |    Dependants | [`cuda-python:12.2`](/packages/cuda/cuda-python) [`cuda:12.2`](/packages/cuda/cuda) [`cuda:12.2-samples`](/packages/cuda/cuda) [`cudnn:8.9`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.2-samples`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) [`cuda:12.2`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Images | [`dustynv/cuda:12.2-samples-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) `(2023-12-07, 4.8GB)` | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.4`** | | | :-- | :-- | |    Requires | `L4T ['==36.*', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu124`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.4`](/packages/cuda/cuda-python) [`cuda:12.4-samples`](/packages/cuda/cuda) [`cudnn:9.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu124`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | | |    Dependants | [`cuda-python:12.4`](/packages/cuda/cuda-python) [`cuda:12.4`](/packages/cuda/cuda) [`cuda:12.4-samples`](/packages/cuda/cuda) [`cudnn:9.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.4-samples`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu124`](/packages/cuda/cuda) [`cuda:12.4`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.6`** | | | :-- | :-- | |    Aliases | `cuda` | |    Requires | `L4T ['==36.*', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) | |    Dependants | [`3d_diffusion_policy`](/packages/diffusion/3d_diffusion_policy) [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`4k4d:0.0.0`](/packages/3d/gaussian_splatting/4k4d) [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`apex:0.1`](/packages/pytorch/apex) [`arrow:12.0.1`](/packages/numeric/arrow) [`arrow:14.0.1`](/packages/numeric/arrow) [`arrow:19.0.1`](/packages/numeric/arrow) [`arrow:20.0.0`](/packages/numeric/arrow) [`arrow:5.0.0`](/packages/numeric/arrow) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`bitsandbytes:0.39.1`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.4`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.5`](/packages/llm/bitsandbytes) [`bitsandbytes:0.46.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.47.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.48.0`](/packages/llm/bitsandbytes) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`cache_dit:0.2.0`](/packages/diffusion/cache_edit) [`causalconv1d:1.6.0`](/packages/ml/mamba/causalconv1d) [`chatterbox-tts`](/packages/speech/chatterbox-tts) [`clip_trt`](/packages/vit/clip_trt) [`cobra:0.0.1`](/packages/ml/mamba/cobra) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`crossformer`](/packages/physicalAI/vla/crossformer) [`ctranslate2:4.6.0`](/packages/ml/ctranslate2) [`ctranslate2:master`](/packages/ml/ctranslate2) [`cuda-python:12.6`](/packages/cuda/cuda-python) [`cuda:12.6-samples`](/packages/cuda/cuda) [`cudf:23.10.03`](/packages/ml/rapids/cudf) [`cudf:25.08.00`](/packages/ml/rapids/cudf) [`cudnn:9.3`](/packages/cuda/cudnn) [`cudnn:9.4`](/packages/cuda/cudnn) [`cuml`](/packages/ml/rapids/cuml) [`cupy`](/packages/numeric/cupy) [`cutlass:4.1.0`](/packages/cuda/cutlass) [`cv-cuda:0.15`](/packages/cv/cv-cuda) [`cv-cuda:0.15-cpp`](/packages/cv/cv-cuda) [`decord2:1.0.0`](/packages/multimedia/decord) [`deepspeed-kernels:0.1.0`](/packages/llm/deepspeed/deepspeed-kernels) [`deepspeed:0.15.2`](/packages/llm/deepspeed) [`deepspeed:0.18.0`](/packages/llm/deepspeed) [`deepspeed:0.9.5`](/packages/llm/deepspeed) [`deepstream`](/packages/cv/deepstream) [`diffusers:0.35.0`](/packages/diffusion/diffusers) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dimba:1.0`](/packages/ml/mamba/dimba) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`easyvolcap:0.0.0`](/packages/3d/gaussian_splatting/easyvolcap) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faiss:1.12.0`](/packages/vectordb/faiss) [`faiss_lite`](/packages/vectordb/faiss_lite) [`fast_gauss:1.0.0`](/packages/3d/gaussian_splatting/fast_gauss) [`faster-whisper`](/packages/speech/faster-whisper) [`ffmpeg:7.1`](/packages/multimedia/ffmpeg) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flashinfer:0.2.1.post2`](/packages/attention/flash-infer) [`flashinfer:0.2.2`](/packages/attention/flash-infer) [`flashinfer:0.2.2.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.6.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.7`](/packages/attention/flash-infer) [`flashinfer:0.2.8`](/packages/attention/flash-infer) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`gdrcopy`](/packages/cuda/gdrcopy) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`genesis-world:0.2.2`](/packages/sim/genesis) [`glomap:2.0.0`](/packages/3d/3dvision/glomap) [`gptqmodel:3.0.1`](/packages/llm/gptqmodel) [`gsplat:1.5.3`](/packages/3d/gaussian_splatting/gsplat) [`gstreamer`](/packages/multimedia/gstreamer) [`habitat-sim:0.3.4`](/packages/sim/habitat-sim) [`hloc:1.4`](/packages/3d/3dvision/hloc) [`hloc:1.5`](/packages/3d/3dvision/hloc) [`holoscan`](/packages/cv/holoscan) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaaclab:2.2.0`](/packages/sim/isaac-sim/isaac-lab) [`isaacsim:5.0.0`](/packages/sim/isaac-sim) [`jax:0.4.38`](/packages/ml/jax) [`jax:0.6.3`](/packages/ml/jax) [`jetcam`](/packages/hw/jetcam) [`jetson-copilot`](/packages/rag/jetson-copilot) [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`jetson-inference:main`](/packages/cv/jetson-inference) [`jetson-utils:v1`](/packages/multimedia/jetson-utils) [`jetson-utils:v2`](/packages/multimedia/jetson-utils) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`kai_scheduler:0.5.5`](/packages/llm/dynamo/kai-scheduler) [`kaolin:0.18.0`](/packages/3d/3dvision/kaolin) [`kat:1`](/packages/ml/kans/kat) [`kokoro-tts:fastapi`](/packages/speech/kokoro-tts/kokoro-tts-fastapi) [`kokoro-tts:hf`](/packages/speech/kokoro-tts/kokoro-tts-hf) [`kokoro-tts:onnx`](/packages/speech/kokoro-tts/kokoro-tts-onnx) [`ktransformers:0.3.3`](/packages/llm/ktransformers) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain`](/packages/rag/langchain) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`libcom:0.1.0`](/packages/multimedia/libcom) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-index`](/packages/rag/llama-index) [`llama-index:samples`](/packages/rag/llama-index) [`llama-vision`](/packages/vlm/llama-vision) [`llama_cpp:0.2.57`](/packages/llm/llama_cpp) [`llama_cpp:0.2.70`](/packages/llm/llama_cpp) [`llama_cpp:0.2.83`](/packages/llm/llama_cpp) [`llama_cpp:0.2.90`](/packages/llm/llama_cpp) [`llama_cpp:0.3.1`](/packages/llm/llama_cpp) [`llama_cpp:0.3.2`](/packages/llm/llama_cpp) [`llama_cpp:0.3.5`](/packages/llm/llama_cpp) [`llama_cpp:0.3.6`](/packages/llm/llama_cpp) [`llama_cpp:0.3.7`](/packages/llm/llama_cpp) [`llama_cpp:0.3.8`](/packages/llm/llama_cpp) [`llama_cpp:0.3.9`](/packages/llm/llama_cpp) [`llama_cpp:0.4.0`](/packages/llm/llama_cpp) [`llama_cpp:b5255`](/packages/llm/llama_cpp) [`llama_cpp:b5833`](/packages/llm/llama_cpp) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`log-linear-attention:0.0.1`](/packages/attention/log-linear-attention) [`mamba:2.2.5`](/packages/ml/mamba/mamba) [`mambavision:1.0`](/packages/ml/mamba/mambavision) [`memvid:0.1.4`](/packages/rag/memvid) [`mimicgen`](/packages/sim/mimicgen) [`minference:0.1.7`](/packages/llm/minference) [`minigpt4`](/packages/vlm/minigpt4) [`mistral_rs`](/packages/llm/mistral_rs) [`mlc:0.1.0`](/packages/llm/mlc) [`mlc:0.1.1`](/packages/llm/mlc) [`mlc:0.1.2`](/packages/llm/mlc) [`mlc:0.1.3`](/packages/llm/mlc) [`mlc:0.1.4`](/packages/llm/mlc) [`mlc:0.19.0`](/packages/llm/mlc) [`mlc:0.20.0`](/packages/llm/mlc) [`mlc:0.21.0`](/packages/llm/mlc) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`mooncake:0.3.5`](/packages/llm/dynamo/mooncake) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nanoowl`](/packages/vit/nanoowl) [`nanosam`](/packages/vit/nanosam) [`nccl`](/packages/cuda/nccl) [`nemo`](/packages/llm/nemo) [`nerfacc:0.5.3`](/packages/3d/3dvision/nerfacc) [`nerfacc:0.5.4`](/packages/3d/3dvision/nerfacc) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`nerfview:0.1.4`](/packages/3d/gaussian_splatting/nerfview) [`nixl:0.3.2`](/packages/llm/dynamo/nixl) [`numba`](/packages/numeric/numba) [`nvdiffrast:0.3.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer/nvdiffrast) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`nvpl`](/packages/cuda/nvpl) [`octo`](/packages/physicalAI/vla/octo) [`ollama:0.10.0`](/packages/llm/ollama) [`ollama:0.4.0`](/packages/llm/ollama) [`ollama:0.5.1`](/packages/llm/ollama) [`ollama:0.5.5`](/packages/llm/ollama) [`ollama:0.5.7`](/packages/llm/ollama) [`ollama:0.6.7`](/packages/llm/ollama) [`ollama:0.7.0`](/packages/llm/ollama) [`ollama:0.8.0`](/packages/llm/ollama) [`ollama:0.9.6`](/packages/llm/ollama) [`onnxruntime:1.19.2`](/packages/ml/onnxruntime) [`onnxruntime:1.20`](/packages/ml/onnxruntime) [`onnxruntime:1.20.1`](/packages/ml/onnxruntime) [`onnxruntime:1.21`](/packages/ml/onnxruntime) [`onnxruntime:1.22`](/packages/ml/onnxruntime) [`onnxruntime_genai:0.8.5`](/packages/ml/onnxruntime_genai) [`open3d:1.19.0`](/packages/3d/3dvision/open3d) [`opencv:4.10.0`](/packages/cv/opencv) [`opencv:4.10.0-meta`](/packages/cv/opencv) [`opencv:4.11.0`](/packages/cv/opencv) [`opencv:4.11.0-meta`](/packages/cv/opencv) [`opencv:4.12.0`](/packages/cv/opencv) [`opencv:4.12.0-meta`](/packages/cv/opencv) [`opencv:4.8.1`](/packages/cv/opencv) [`opencv:4.8.1-deb`](/packages/cv/opencv) [`opencv:4.8.1-meta`](/packages/cv/opencv) [`opendronemap`](/packages/robots/opendronemap) [`opendronemap:node`](/packages/robots/opendronemap) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`optimum`](/packages/llm/optimum) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`physx`](/packages/sim/physx) [`piper-tts`](/packages/speech/piper-tts) [`piper1-tts:1.3.0`](/packages/speech/piper1-tts) [`pixsfm:1.0`](/packages/3d/3dvision/pixsfm) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`protomotions:2.5.0`](/packages/robots/protomotions) [`pyav`](/packages/multimedia/pyav) [`pycolmap:3.12`](/packages/3d/3dvision/pycolmap) [`pycolmap:3.13`](/packages/3d/3dvision/pycolmap) [`pycuda`](/packages/cuda/pycuda) [`pykan:0.2.9`](/packages/ml/kans/pykan) [`pytorch3d:0.7.8`](/packages/pytorch/torch3d) [`pytorch3d:0.7.9`](/packages/pytorch/torch3d) [`pytorch:2.1`](/packages/pytorch) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`raft`](/packages/ml/rapids/raft) [`realsense`](/packages/hw/realsense) [`robogen`](/packages/sim/robogen) [`robomimic`](/packages/sim/robomimic) [`robopoint`](/packages/physicalAI/vla/robopoint) [`robosuite`](/packages/sim/robosuite) [`ros:foxy-desktop`](/packages/robots/ros) [`ros:foxy-ros-base`](/packages/robots/ros) [`ros:foxy-ros-core`](/packages/robots/ros) [`ros:galactic-desktop`](/packages/robots/ros) [`ros:galactic-ros-base`](/packages/robots/ros) [`ros:galactic-ros-core`](/packages/robots/ros) [`ros:humble-desktop`](/packages/robots/ros) [`ros:humble-ros-base`](/packages/robots/ros) [`ros:humble-ros-core`](/packages/robots/ros) [`ros:iron-desktop`](/packages/robots/ros) [`ros:iron-ros-base`](/packages/robots/ros) [`ros:iron-ros-core`](/packages/robots/ros) [`ros:jazzy-desktop`](/packages/robots/ros) [`ros:jazzy-ros-base`](/packages/robots/ros) [`ros:jazzy-ros-core`](/packages/robots/ros) [`ros:noetic-desktop`](/packages/robots/ros) [`ros:noetic-ros-base`](/packages/robots/ros) [`ros:noetic-ros-core`](/packages/robots/ros) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sapien`](/packages/sim/sapien) [`sapiens`](/packages/vit/sapiens) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sound-utils`](/packages/multimedia/sound-utils) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`spark-tts`](/packages/speech/spark-tts) [`speaches`](/packages/speech/speaches) [`splashsurf`](/packages/sim/genesis/splashSurf) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim:hf`](/packages/llm/sudonim) [`quadrants:1.8.0`](/packages/sim/genesis/quadrants) [`tam`](/packages/vit/tam) [`tensorflow2:2.16.1`](/packages/ml/tensorflow) [`tensorflow2:2.18.0`](/packages/ml/tensorflow) [`tensorflow2:2.19.0`](/packages/ml/tensorflow) [`tensorflow2:2.20.0`](/packages/ml/tensorflow) [`tensorflow2:2.21.0`](/packages/ml/tensorflow) [`tensorflow_graphics:2.18.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.19.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.20.0`](/packages/ml/tensorflow/graphics) [`tensorflow_text:2.18.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.19.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.20.0`](/packages/ml/tensorflow/text) [`tensorrt:10.3`](/packages/cuda/tensorrt) [`tensorrt:10.4`](/packages/cuda/tensorrt) [`tensorrt:10.5`](/packages/cuda/tensorrt) [`tensorrt:10.7`](/packages/cuda/tensorrt) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-inference`](/packages/llm/text-generation-inference) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`tinycudann:1.7`](/packages/3d/3dvision/tinycudann) [`torch-memory-saver:0.0.7`](/packages/pytorch/torchsaver) [`torch2trt`](/packages/pytorch/torch2trt) [`torch_tensorrt`](/packages/pytorch/torch_tensorrt) [`torchao:0.12.0`](/packages/pytorch/torchao) [`torchao:0.13.0`](/packages/pytorch/torchao) [`torchaudio:2.1.0`](/packages/pytorch/torchaudio) [`torchaudio:2.2.2`](/packages/pytorch/torchaudio) [`torchaudio:2.3.0`](/packages/pytorch/torchaudio) [`torchaudio:2.4.0`](/packages/pytorch/torchaudio) [`torchaudio:2.5.0`](/packages/pytorch/torchaudio) [`torchaudio:2.6.0`](/packages/pytorch/torchaudio) [`torchaudio:2.7.0`](/packages/pytorch/torchaudio) [`torchaudio:2.8.0`](/packages/pytorch/torchaudio) [`torchsde:0.2.1`](/packages/pytorch/torchsde) [`torchsde:0.2.2`](/packages/pytorch/torchsde) [`torchsde:0.2.3`](/packages/pytorch/torchsde) [`torchsde:0.2.4`](/packages/pytorch/torchsde) [`torchsde:0.2.5`](/packages/pytorch/torchsde) [`torchsde:0.2.6`](/packages/pytorch/torchsde) [`torchsde:0.2.7`](/packages/pytorch/torchsde) [`torchsde:0.2.8`](/packages/pytorch/torchsde) [`torchtext:0.18.0`](/packages/pytorch/torchtext) [`torchtext:0.19.0`](/packages/pytorch/torchtext) [`torchvision:0.16.2`](/packages/pytorch/torchvision) [`torchvision:0.17.2`](/packages/pytorch/torchvision) [`torchvision:0.18.0`](/packages/pytorch/torchvision) [`torchvision:0.19.1`](/packages/pytorch/torchvision) [`torchvision:0.20.0`](/packages/pytorch/torchvision) [`torchvision:0.21.0`](/packages/pytorch/torchvision) [`torchvision:0.22.0`](/packages/pytorch/torchvision) [`torchvision:0.23.0`](/packages/pytorch/torchvision) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`transformers:4.53.1`](/packages/llm/transformers) [`triton:3.0.0`](/packages/ml/triton) [`triton:3.1.0`](/packages/ml/triton) [`triton:3.2.0`](/packages/ml/triton) [`triton:3.3.0`](/packages/ml/triton) [`triton:3.3.1`](/packages/ml/triton) [`triton:3.4.0`](/packages/ml/triton) [`triton:3.5.0`](/packages/ml/triton) [`tritonserver`](/packages/ml/tritonserver) [`tvm`](/packages/ml/tvm) [`video-codec-sdk:12.2.72`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:12.2.72-samples`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19-samples`](/packages/multimedia/video-codec-sdk) [`videollama:1.0.0`](/packages/vlm/videollama) [`videomambasuite:1.0`](/packages/ml/mamba/videomambasuite) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vpi`](/packages/cv/vpi) [`vscode:cuda`](/packages/code/vscode) [`vscode:torch`](/packages/code/vscode) [`vscode:transformers`](/packages/code/vscode) [`warp:1.7.0`](/packages/numeric/warp) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.7.0-jax`](/packages/numeric/warp) [`warp:1.7.0-torch`](/packages/numeric/warp) [`warp:1.8.1`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`warp:1.8.1-jax`](/packages/numeric/warp) [`warp:1.8.1-torch`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`whisper_trt`](/packages/speech/whisper_trt) [`whisperx`](/packages/speech/whisperx) [`wyoming-piper:1.6.2`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-piper:master`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-whisper:2.5.0`](/packages/smart-home/wyoming/wyoming-whisper) [`wyoming-whisper:master`](/packages/smart-home/wyoming/wyoming-whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xformers:0.0.32`](/packages/attention/xformers) [`xformers:0.0.33`](/packages/attention/xformers) [`xgrammar:0.1.15`](/packages/llm/xgrammar) [`xgrammar:0.1.18`](/packages/llm/xgrammar) [`xgrammar:0.1.19`](/packages/llm/xgrammar) [`xgrammar:0.1.20`](/packages/llm/xgrammar) [`xgrammar:0.1.21`](/packages/llm/xgrammar) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtts`](/packages/speech/xtts) [`xtuner`](/packages/vlm/xtuner) [`zed:5.0`](/packages/hw/zed) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu126`** | | | :-- | :-- | |    Aliases | `pip_cache` `pip-cache` | |    Requires | `L4T ['==36.*']` | |    Dependencies | | |    Dependants | [`3d_diffusion_policy`](/packages/diffusion/3d_diffusion_policy) [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`4k4d:0.0.0`](/packages/3d/gaussian_splatting/4k4d) [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`aim`](/packages/ml/aim) [`apex:0.1`](/packages/pytorch/apex) [`arrow:12.0.1`](/packages/numeric/arrow) [`arrow:14.0.1`](/packages/numeric/arrow) [`arrow:19.0.1`](/packages/numeric/arrow) [`arrow:20.0.0`](/packages/numeric/arrow) [`arrow:5.0.0`](/packages/numeric/arrow) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`bitsandbytes:0.39.1`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.4`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.5`](/packages/llm/bitsandbytes) [`bitsandbytes:0.46.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.47.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.48.0`](/packages/llm/bitsandbytes) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`cache_dit:0.2.0`](/packages/diffusion/cache_edit) [`canbus`](/packages/hw/canbus) [`cangaroo`](/packages/hw/cangaroo) [`causalconv1d:1.6.0`](/packages/ml/mamba/causalconv1d) [`chatterbox-tts`](/packages/speech/chatterbox-tts) [`ciso8601:v2.3.2`](/packages/smart-home/dependencies/ciso8601) [`clip_trt`](/packages/vit/clip_trt) [`cmake:pip`](/packages/build/cmake/cmake_pip) [`cobra:0.0.1`](/packages/ml/mamba/cobra) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`crossformer`](/packages/physicalAI/vla/crossformer) [`ctranslate2:4.6.0`](/packages/ml/ctranslate2) [`ctranslate2:master`](/packages/ml/ctranslate2) [`cuda-python:12.6`](/packages/cuda/cuda-python) [`cuda:12.6`](/packages/cuda/cuda) [`cuda:12.6-samples`](/packages/cuda/cuda) [`cudf:23.10.03`](/packages/ml/rapids/cudf) [`cudf:25.08.00`](/packages/ml/rapids/cudf) [`cudnn:9.3`](/packages/cuda/cudnn) [`cudnn:9.4`](/packages/cuda/cudnn) [`cuml`](/packages/ml/rapids/cuml) [`cupy`](/packages/numeric/cupy) [`cutlass:4.1.0`](/packages/cuda/cutlass) [`cv-cuda:0.15`](/packages/cv/cv-cuda) [`cv-cuda:0.15-cpp`](/packages/cv/cv-cuda) [`decord2:1.0.0`](/packages/multimedia/decord) [`deepspeed-kernels:0.1.0`](/packages/llm/deepspeed/deepspeed-kernels) [`deepspeed:0.15.2`](/packages/llm/deepspeed) [`deepspeed:0.18.0`](/packages/llm/deepspeed) [`deepspeed:0.9.5`](/packages/llm/deepspeed) [`deepstream`](/packages/cv/deepstream) [`diffusers:0.35.0`](/packages/diffusion/diffusers) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dimba:1.0`](/packages/ml/mamba/dimba) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`docker`](/packages/build/docker) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`easyvolcap:0.0.0`](/packages/3d/gaussian_splatting/easyvolcap) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faiss:1.12.0`](/packages/vectordb/faiss) [`faiss_lite`](/packages/vectordb/faiss_lite) [`fast_gauss:1.0.0`](/packages/3d/gaussian_splatting/fast_gauss) [`faster-whisper`](/packages/speech/faster-whisper) [`ffmpeg:7.1`](/packages/multimedia/ffmpeg) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flashinfer:0.2.1.post2`](/packages/attention/flash-infer) [`flashinfer:0.2.2`](/packages/attention/flash-infer) [`flashinfer:0.2.2.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.6.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.7`](/packages/attention/flash-infer) [`flashinfer:0.2.8`](/packages/attention/flash-infer) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`gdrcopy`](/packages/cuda/gdrcopy) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`genesis-world:0.2.2`](/packages/sim/genesis) [`glomap:2.0.0`](/packages/3d/3dvision/glomap) [`gptqmodel:3.0.1`](/packages/llm/gptqmodel) [`graphiti`](/packages/rag/graphiti) [`gsplat:1.5.3`](/packages/3d/gaussian_splatting/gsplat) [`gstreamer`](/packages/multimedia/gstreamer) [`h5py`](/packages/build/h5py) [`habitat-sim:0.3.4`](/packages/sim/habitat-sim) [`hloc:1.4`](/packages/3d/3dvision/hloc) [`hloc:1.5`](/packages/3d/3dvision/hloc) [`holoscan`](/packages/cv/holoscan) [`homeassistant-core:2025.7.0`](/packages/smart-home/homeassistant-core) [`huggingface_hub`](/packages/llm/huggingface_hub) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaaclab:2.2.0`](/packages/sim/isaac-sim/isaac-lab) [`isaacsim:5.0.0`](/packages/sim/isaac-sim) [`jax:0.4.38`](/packages/ml/jax) [`jax:0.6.3`](/packages/ml/jax) [`jetcam`](/packages/hw/jetcam) [`jetson-copilot`](/packages/rag/jetson-copilot) [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`jetson-inference:main`](/packages/cv/jetson-inference) [`jetson-utils:python`](/packages/multimedia/jetson-utils) [`jetson-utils:v1`](/packages/multimedia/jetson-utils) [`jetson-utils:v2`](/packages/multimedia/jetson-utils) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`kai_scheduler:0.5.5`](/packages/llm/dynamo/kai-scheduler) [`kaolin:0.18.0`](/packages/3d/3dvision/kaolin) [`kat:1`](/packages/ml/kans/kat) [`kokoro-tts:fastapi`](/packages/speech/kokoro-tts/kokoro-tts-fastapi) [`kokoro-tts:hf`](/packages/speech/kokoro-tts/kokoro-tts-hf) [`kokoro-tts:onnx`](/packages/speech/kokoro-tts/kokoro-tts-onnx) [`ktransformers:0.3.3`](/packages/llm/ktransformers) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain`](/packages/rag/langchain) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`libcom:0.1.0`](/packages/multimedia/libcom) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-index`](/packages/rag/llama-index) [`llama-index:samples`](/packages/rag/llama-index) [`llama-vision`](/packages/vlm/llama-vision) [`llama_cpp:0.2.57`](/packages/llm/llama_cpp) [`llama_cpp:0.2.70`](/packages/llm/llama_cpp) [`llama_cpp:0.2.83`](/packages/llm/llama_cpp) [`llama_cpp:0.2.90`](/packages/llm/llama_cpp) [`llama_cpp:0.3.1`](/packages/llm/llama_cpp) [`llama_cpp:0.3.2`](/packages/llm/llama_cpp) [`llama_cpp:0.3.5`](/packages/llm/llama_cpp) [`llama_cpp:0.3.6`](/packages/llm/llama_cpp) [`llama_cpp:0.3.7`](/packages/llm/llama_cpp) [`llama_cpp:0.3.8`](/packages/llm/llama_cpp) [`llama_cpp:0.3.9`](/packages/llm/llama_cpp) [`llama_cpp:0.4.0`](/packages/llm/llama_cpp) [`llama_cpp:b5255`](/packages/llm/llama_cpp) [`llama_cpp:b5833`](/packages/llm/llama_cpp) [`llamaspeak`](/packages/llm/llamaspeak) [`llava`](/packages/vlm/llava) [`llvm:10`](/packages/build/llvm) [`llvm:11`](/packages/build/llvm) [`llvm:12`](/packages/build/llvm) [`llvm:13`](/packages/build/llvm) [`llvm:14`](/packages/build/llvm) [`llvm:15`](/packages/build/llvm) [`llvm:16`](/packages/build/llvm) [`llvm:17`](/packages/build/llvm) [`llvm:18`](/packages/build/llvm) [`llvm:19`](/packages/build/llvm) [`llvm:22`](/packages/build/llvm) [`llvm:22`](/packages/build/llvm) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`log-linear-attention:0.0.1`](/packages/attention/log-linear-attention) [`mamba:2.2.5`](/packages/ml/mamba/mamba) [`mambavision:1.0`](/packages/ml/mamba/mambavision) [`memvid:0.1.4`](/packages/rag/memvid) [`meshlab:MeshLab-2023.12`](/packages/3d/3dvision/meshlab) [`meshlab:MeshLab-2025.03`](/packages/3d/3dvision/meshlab) [`mimicgen`](/packages/sim/mimicgen) [`minference:0.1.7`](/packages/llm/minference) [`minigpt4`](/packages/vlm/minigpt4) [`mistral_rs`](/packages/llm/mistral_rs) [`mlc:0.1.0`](/packages/llm/mlc) [`mlc:0.1.1`](/packages/llm/mlc) [`mlc:0.1.2`](/packages/llm/mlc) [`mlc:0.1.3`](/packages/llm/mlc) [`mlc:0.1.4`](/packages/llm/mlc) [`mlc:0.19.0`](/packages/llm/mlc) [`mlc:0.20.0`](/packages/llm/mlc) [`mlc:0.21.0`](/packages/llm/mlc) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`mooncake:0.3.5`](/packages/llm/dynamo/mooncake) [`mujoco`](/packages/sim/mujoco) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nanoowl`](/packages/vit/nanoowl) [`nanosam`](/packages/vit/nanosam) [`nccl`](/packages/cuda/nccl) [`nemo`](/packages/llm/nemo) [`nerfacc:0.5.3`](/packages/3d/3dvision/nerfacc) [`nerfacc:0.5.4`](/packages/3d/3dvision/nerfacc) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`nerfview:0.1.4`](/packages/3d/gaussian_splatting/nerfview) [`ninja`](/packages/build/ninja) [`nixl:0.3.2`](/packages/llm/dynamo/nixl) [`numba`](/packages/numeric/numba) [`numpy`](/packages/numeric/numpy) [`nvdiffrast:0.3.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer/nvdiffrast) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`nvpl`](/packages/cuda/nvpl) [`octo`](/packages/physicalAI/vla/octo) [`ollama:0.10.0`](/packages/llm/ollama) [`ollama:0.4.0`](/packages/llm/ollama) [`ollama:0.5.1`](/packages/llm/ollama) [`ollama:0.5.5`](/packages/llm/ollama) [`ollama:0.5.7`](/packages/llm/ollama) [`ollama:0.6.7`](/packages/llm/ollama) [`ollama:0.7.0`](/packages/llm/ollama) [`ollama:0.8.0`](/packages/llm/ollama) [`ollama:0.9.6`](/packages/llm/ollama) [`onnx`](/packages/ml/onnx) [`onnxruntime:1.19.2`](/packages/ml/onnxruntime) [`onnxruntime:1.20`](/packages/ml/onnxruntime) [`onnxruntime:1.20.1`](/packages/ml/onnxruntime) [`onnxruntime:1.21`](/packages/ml/onnxruntime) [`onnxruntime:1.22`](/packages/ml/onnxruntime) [`onnxruntime_genai:0.8.5`](/packages/ml/onnxruntime_genai) [`open3d:1.19.0`](/packages/3d/3dvision/open3d) [`openai`](/packages/llm/openai) [`opencv:4.10.0`](/packages/cv/opencv) [`opencv:4.10.0-meta`](/packages/cv/opencv) [`opencv:4.11.0`](/packages/cv/opencv) [`opencv:4.11.0-meta`](/packages/cv/opencv) [`opencv:4.12.0`](/packages/cv/opencv) [`opencv:4.12.0-meta`](/packages/cv/opencv) [`opencv:4.8.1`](/packages/cv/opencv) [`opencv:4.8.1-deb`](/packages/cv/opencv) [`opencv:4.8.1-meta`](/packages/cv/opencv) [`opendronemap`](/packages/robots/opendronemap) [`opendronemap:node`](/packages/robots/opendronemap) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`optimum`](/packages/llm/optimum) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`physx`](/packages/sim/physx) [`piper-tts`](/packages/speech/piper-tts) [`piper1-tts:1.3.0`](/packages/speech/piper1-tts) [`pixsfm:1.0`](/packages/3d/3dvision/pixsfm) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`polyscope:2.5.0`](/packages/3d/3dvision/polyscope) [`prismatic`](/packages/vlm/prismatic) [`protobuf:apt`](/packages/build/protobuf/protobuf_apt) [`protobuf:cpp`](/packages/build/protobuf/protobuf_cpp) [`protomotions:2.5.0`](/packages/robots/protomotions) [`psutil-home-assistant:0.0.1`](/packages/smart-home/dependencies/psutil-home-assistant) [`pyav`](/packages/multimedia/pyav) [`pybind11`](/packages/build/pybind11) [`pyceres:2.5`](/packages/3d/3dvision/pyceres) [`pycolmap:3.12`](/packages/3d/3dvision/pycolmap) [`pycolmap:3.13`](/packages/3d/3dvision/pycolmap) [`pycuda`](/packages/cuda/pycuda) [`pykan:0.2.9`](/packages/ml/kans/pykan) [`pymeshlab:2023.12.post2`](/packages/3d/3dvision/pymeshlab) [`pymeshlab:2023.12.post3`](/packages/3d/3dvision/pymeshlab) [`pymeshlab:2025.6.23.dev0`](/packages/3d/3dvision/pymeshlab) [`python:3.10`](/packages/build/python) [`python:3.11`](/packages/build/python) [`python:3.12`](/packages/build/python) [`python:3.13`](/packages/build/python) [`python:3.14`](/packages/build/python) [`pytorch3d:0.7.8`](/packages/pytorch/torch3d) [`pytorch3d:0.7.9`](/packages/pytorch/torch3d) [`pytorch:2.1`](/packages/pytorch) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`raft`](/packages/ml/rapids/raft) [`realsense`](/packages/hw/realsense) [`riva-client:python`](/packages/speech/riva-client) [`robogen`](/packages/sim/robogen) [`robomimic`](/packages/sim/robomimic) [`robopoint`](/packages/physicalAI/vla/robopoint) [`robosuite`](/packages/sim/robosuite) [`ros:foxy-desktop`](/packages/robots/ros) [`ros:foxy-ros-base`](/packages/robots/ros) [`ros:foxy-ros-core`](/packages/robots/ros) [`ros:galactic-desktop`](/packages/robots/ros) [`ros:galactic-ros-base`](/packages/robots/ros) [`ros:galactic-ros-core`](/packages/robots/ros) [`ros:humble-desktop`](/packages/robots/ros) [`ros:humble-ros-base`](/packages/robots/ros) [`ros:humble-ros-core`](/packages/robots/ros) [`ros:iron-desktop`](/packages/robots/ros) [`ros:iron-ros-base`](/packages/robots/ros) [`ros:iron-ros-core`](/packages/robots/ros) [`ros:jazzy-desktop`](/packages/robots/ros) [`ros:jazzy-ros-base`](/packages/robots/ros) [`ros:jazzy-ros-core`](/packages/robots/ros) [`ros:noetic-desktop`](/packages/robots/ros) [`ros:noetic-ros-base`](/packages/robots/ros) [`ros:noetic-ros-core`](/packages/robots/ros) [`rust`](/packages/build/rust) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sapien`](/packages/sim/sapien) [`sapiens`](/packages/vit/sapiens) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sound-utils`](/packages/multimedia/sound-utils) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`spark-tts`](/packages/speech/spark-tts) [`speaches`](/packages/speech/speaches) [`splashsurf`](/packages/sim/genesis/splashSurf) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim`](/packages/llm/sudonim) [`sudonim:hf`](/packages/llm/sudonim) [`quadrants:1.8.0`](/packages/sim/genesis/quadrants) [`tam`](/packages/vit/tam) [`tensorboard`](/packages/ml/tensorflow/tensorboard) [`tensorflow2:2.16.1`](/packages/ml/tensorflow) [`tensorflow2:2.18.0`](/packages/ml/tensorflow) [`tensorflow2:2.19.0`](/packages/ml/tensorflow) [`tensorflow2:2.20.0`](/packages/ml/tensorflow) [`tensorflow2:2.21.0`](/packages/ml/tensorflow) [`tensorflow_graphics:2.18.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.19.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.20.0`](/packages/ml/tensorflow/graphics) [`tensorflow_text:2.18.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.19.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.20.0`](/packages/ml/tensorflow/text) [`tensorrt:10.3`](/packages/cuda/tensorrt) [`tensorrt:10.4`](/packages/cuda/tensorrt) [`tensorrt:10.5`](/packages/cuda/tensorrt) [`tensorrt:10.7`](/packages/cuda/tensorrt) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-inference`](/packages/llm/text-generation-inference) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`tinycudann:1.7`](/packages/3d/3dvision/tinycudann) [`torch-memory-saver:0.0.7`](/packages/pytorch/torchsaver) [`torch2trt`](/packages/pytorch/torch2trt) [`torch_tensorrt`](/packages/pytorch/torch_tensorrt) [`torchao:0.12.0`](/packages/pytorch/torchao) [`torchao:0.13.0`](/packages/pytorch/torchao) [`torchaudio:2.1.0`](/packages/pytorch/torchaudio) [`torchaudio:2.2.2`](/packages/pytorch/torchaudio) [`torchaudio:2.3.0`](/packages/pytorch/torchaudio) [`torchaudio:2.4.0`](/packages/pytorch/torchaudio) [`torchaudio:2.5.0`](/packages/pytorch/torchaudio) [`torchaudio:2.6.0`](/packages/pytorch/torchaudio) [`torchaudio:2.7.0`](/packages/pytorch/torchaudio) [`torchaudio:2.8.0`](/packages/pytorch/torchaudio) [`torchsde:0.2.1`](/packages/pytorch/torchsde) [`torchsde:0.2.2`](/packages/pytorch/torchsde) [`torchsde:0.2.3`](/packages/pytorch/torchsde) [`torchsde:0.2.4`](/packages/pytorch/torchsde) [`torchsde:0.2.5`](/packages/pytorch/torchsde) [`torchsde:0.2.6`](/packages/pytorch/torchsde) [`torchsde:0.2.7`](/packages/pytorch/torchsde) [`torchsde:0.2.8`](/packages/pytorch/torchsde) [`torchtext:0.18.0`](/packages/pytorch/torchtext) [`torchtext:0.19.0`](/packages/pytorch/torchtext) [`torchvision:0.16.2`](/packages/pytorch/torchvision) [`torchvision:0.17.2`](/packages/pytorch/torchvision) [`torchvision:0.18.0`](/packages/pytorch/torchvision) [`torchvision:0.19.1`](/packages/pytorch/torchvision) [`torchvision:0.20.0`](/packages/pytorch/torchvision) [`torchvision:0.21.0`](/packages/pytorch/torchvision) [`torchvision:0.22.0`](/packages/pytorch/torchvision) [`torchvision:0.23.0`](/packages/pytorch/torchvision) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`transformers:4.53.1`](/packages/llm/transformers) [`triton:3.0.0`](/packages/ml/triton) [`triton:3.1.0`](/packages/ml/triton) [`triton:3.2.0`](/packages/ml/triton) [`triton:3.3.0`](/packages/ml/triton) [`triton:3.3.1`](/packages/ml/triton) [`triton:3.4.0`](/packages/ml/triton) [`triton:3.5.0`](/packages/ml/triton) [`tritonserver`](/packages/ml/tritonserver) [`tvm`](/packages/ml/tvm) [`video-codec-sdk:12.2.72`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:12.2.72-samples`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19-samples`](/packages/multimedia/video-codec-sdk) [`videollama:1.0.0`](/packages/vlm/videollama) [`videomambasuite:1.0`](/packages/ml/mamba/videomambasuite) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vpi`](/packages/cv/vpi) [`vscode`](/packages/code/vscode) [`vscode:cuda`](/packages/code/vscode) [`vscode:torch`](/packages/code/vscode) [`vscode:transformers`](/packages/code/vscode) [`vtk:9.3.1`](/packages/sim/genesis/vtk) [`vtk:9.4.2`](/packages/sim/genesis/vtk) [`vtk:9.5.0`](/packages/sim/genesis/vtk) [`vulkan:1.3.275`](/packages/multimedia/vulkan) [`vulkan:1.3.296`](/packages/multimedia/vulkan) [`vulkan:1.4.309`](/packages/multimedia/vulkan) [`vulkan:1.4.313`](/packages/multimedia/vulkan) [`vulkan:1.4.314`](/packages/multimedia/vulkan) [`warp:1.7.0`](/packages/numeric/warp) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.7.0-jax`](/packages/numeric/warp) [`warp:1.7.0-torch`](/packages/numeric/warp) [`warp:1.8.1`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`warp:1.8.1-jax`](/packages/numeric/warp) [`warp:1.8.1-torch`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`whisper_trt`](/packages/speech/whisper_trt) [`whisperx`](/packages/speech/whisperx) [`wyoming-assist-microphone:1.4.1`](/packages/smart-home/wyoming/wyoming-assist-microphone) [`wyoming-assist-microphone:master`](/packages/smart-home/wyoming/wyoming-assist-microphone) [`wyoming-openwakeword:1.10.0`](/packages/smart-home/wyoming/wyoming-openwakeword) [`wyoming-openwakeword:master`](/packages/smart-home/wyoming/wyoming-openwakeword) [`wyoming-piper:1.6.2`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-piper:master`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-whisper:2.5.0`](/packages/smart-home/wyoming/wyoming-whisper) [`wyoming-whisper:master`](/packages/smart-home/wyoming/wyoming-whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xformers:0.0.32`](/packages/attention/xformers) [`xformers:0.0.33`](/packages/attention/xformers) [`xgrammar:0.1.15`](/packages/llm/xgrammar) [`xgrammar:0.1.18`](/packages/llm/xgrammar) [`xgrammar:0.1.19`](/packages/llm/xgrammar) [`xgrammar:0.1.20`](/packages/llm/xgrammar) [`xgrammar:0.1.21`](/packages/llm/xgrammar) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtts`](/packages/speech/xtts) [`xtuner`](/packages/vlm/xtuner) [`zed:5.0`](/packages/hw/zed) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.6-samples`** | | | :-- | :-- | |    Aliases | `cuda:samples` | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.8`** | | | :-- | :-- | |    Requires | `L4T ['>=36', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu128`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.8`](/packages/cuda/cuda-python) [`cuda:12.8-samples`](/packages/cuda/cuda) [`cudnn:9.8`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu128`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | | |    Dependants | [`cuda-python:12.8`](/packages/cuda/cuda-python) [`cuda:12.8`](/packages/cuda/cuda) [`cuda:12.8-samples`](/packages/cuda/cuda) [`cudnn:9.8`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.8-samples`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu128`](/packages/cuda/cuda) [`cuda:12.8`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.9`** | | | :-- | :-- | |    Requires | `L4T ['>=36', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu129`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.9`](/packages/cuda/cuda-python) [`cuda:12.9-samples`](/packages/cuda/cuda) [`cudnn:9.10`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/cuda:12.9-samples-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 5.7GB)`
[`dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 2.8GB)` | | **`pip_cache:cu129`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | | |    Dependants | [`cuda-python:12.9`](/packages/cuda/cuda-python) [`cuda:12.9`](/packages/cuda/cuda) [`cuda:12.9-samples`](/packages/cuda/cuda) [`cudnn:9.10`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.9-samples`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu129`](/packages/cuda/cuda) [`cuda:12.9`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Images | [`dustynv/cuda:12.9-samples-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 5.7GB)`
[`dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 2.8GB)` | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:13.0`** | | | :-- | :-- | |    Requires | `L4T ['>=36', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu130`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:13.0`](/packages/cuda/cuda-python) [`cuda:13.0-samples`](/packages/cuda/cuda) [`cudnn:10.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu130`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | | |    Dependants | [`cuda-python:13.0`](/packages/cuda/cuda-python) [`cuda:13.0`](/packages/cuda/cuda) [`cuda:13.0-samples`](/packages/cuda/cuda) [`cudnn:10.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:13.0-samples`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu130`](/packages/cuda/cuda) [`cuda:13.0`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](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`](https://hub.docker.com/r/dustynv/cuda/tags) | `2023-12-05` | `arm64` | `3.4GB` | |   [`dustynv/cuda:12.2-samples-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) | `2023-12-07` | `arm64` | `4.8GB` | |   [`dustynv/cuda:12.9-samples-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) | `2025-04-02` | `amd64` | `5.7GB` | |   [`dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) | `2025-04-02` | `arm64` | `2.8GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cuda) # or explicitly specify one of the container images above jetson-containers run dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cuda) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cuda) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cuda ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/cutlass/README.md`) # cutlass > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cutlass:4.1.0`** | | | :-- | :-- | |    Aliases | `cutlass` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`cudnn`](/packages/cuda/cudnn) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) | |    Dependants | [`awq:0.1.0`](/packages/llm/awq) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`exllama:0.1`](/packages/llm/exllama) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`minference:0.1.7`](/packages/llm/minference) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`sudonim:hf`](/packages/llm/sudonim) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`vscode:transformers`](/packages/code/vscode) [`xattention:0.0.1`](/packages/attention/xattention) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtuner`](/packages/vlm/xtuner) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cutlass) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host cutlass:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cutlass) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cutlass) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cutlass ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/gdrcopy/README.md`) # gdrcopy > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`gdrcopy`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) | |    Dependants | [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`nixl:0.3.2`](/packages/llm/dynamo/nixl) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag gdrcopy) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host gdrcopy:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag gdrcopy) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag gdrcopy) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build gdrcopy ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/nccl/README.md`) # nccl > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`nccl`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag nccl) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host nccl:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag nccl) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag nccl) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build nccl ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/nvpl/README.md`) # nvpl > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`nvpl`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag nvpl) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host nvpl:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag nvpl) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag nvpl) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build nvpl ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/pycuda/README.md`) # pycuda > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`pycuda`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dependants | [`audiocraft`](/packages/speech/audiocraft) [`cv-cuda:0.15`](/packages/cv/cv-cuda) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`efficientvit`](/packages/vit/efficientvit) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`llama-index:samples`](/packages/rag/llama-index) [`openpi`](/packages/robots/openpi) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`sam`](/packages/vit/sam) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`tam`](/packages/vit/tam) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/pycuda:r32.7.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-12-06, 0.4GB)`
[`dustynv/pycuda:r35.2.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-09-07, 5.0GB)`
[`dustynv/pycuda:r35.3.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-08-29, 5.0GB)`
[`dustynv/pycuda:r35.4.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-12-06, 5.0GB)`
[`dustynv/pycuda:r36.2.0`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-12-06, 3.5GB)`
[`dustynv/pycuda:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2025-03-03, 2.3GB)` |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/pycuda:r32.7.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-12-06` | `arm64` | `0.4GB` | |   [`dustynv/pycuda:r35.2.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-09-07` | `arm64` | `5.0GB` | |   [`dustynv/pycuda:r35.3.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-08-29` | `arm64` | `5.0GB` | |   [`dustynv/pycuda:r35.4.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-12-06` | `arm64` | `5.0GB` | |   [`dustynv/pycuda:r36.2.0`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-12-06` | `arm64` | `3.5GB` | |   [`dustynv/pycuda:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2025-03-03` | `arm64` | `2.3GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag pycuda) # or explicitly specify one of the container images above jetson-containers run dustynv/pycuda:r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/pycuda:r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag pycuda) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag pycuda) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build pycuda ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`deprecated/cuda/tensorrt/README.md`) # tensorrt > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`tensorrt:10.3`** | | | :-- | :-- | |    Aliases | `tensorrt` | |    Requires | `L4T ['==r36.*', '==cu126', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) | |    Dependants | [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`audiocraft`](/packages/speech/audiocraft) [`clip_trt`](/packages/vit/clip_trt) [`comfyui`](/packages/diffusion/comfyui) [`deepstream`](/packages/cv/deepstream) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`efficientvit`](/packages/vit/efficientvit) [`faster-whisper`](/packages/speech/faster-whisper) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`holoscan`](/packages/cv/holoscan) [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`jetson-inference:main`](/packages/cv/jetson-inference) [`jetson-utils:v1`](/packages/multimedia/jetson-utils) [`jetson-utils:v2`](/packages/multimedia/jetson-utils) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`kokoro-tts:onnx`](/packages/speech/kokoro-tts/kokoro-tts-onnx) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`llama-index:samples`](/packages/rag/llama-index) [`local_llm`](/packages/llm/local_llm) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nanoowl`](/packages/vit/nanoowl) [`nanosam`](/packages/vit/nanosam) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`onnxruntime:1.19.2`](/packages/ml/onnxruntime) [`onnxruntime:1.20`](/packages/ml/onnxruntime) [`onnxruntime:1.20.1`](/packages/ml/onnxruntime) [`onnxruntime:1.21`](/packages/ml/onnxruntime) [`onnxruntime:1.22`](/packages/ml/onnxruntime) [`onnxruntime_genai:0.8.5`](/packages/ml/onnxruntime_genai) [`opendronemap`](/packages/robots/opendronemap) [`opendronemap:node`](/packages/robots/opendronemap) [`openpi`](/packages/robots/openpi) [`optimum`](/packages/llm/optimum) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`piper-tts`](/packages/speech/piper-tts) [`piper1-tts:1.3.0`](/packages/speech/piper1-tts) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`ros:foxy-desktop`](/packages/robots/ros) [`ros:foxy-ros-base`](/packages/robots/ros) [`ros:foxy-ros-core`](/packages/robots/ros) [`ros:galactic-desktop`](/packages/robots/ros) [`ros:galactic-ros-base`](/packages/robots/ros) [`ros:galactic-ros-core`](/packages/robots/ros) [`ros:humble-desktop`](/packages/robots/ros) [`ros:humble-ros-base`](/packages/robots/ros) [`ros:humble-ros-core`](/packages/robots/ros) [`ros:iron-desktop`](/packages/robots/ros) [`ros:iron-ros-base`](/packages/robots/ros) [`ros:iron-ros-core`](/packages/robots/ros) [`ros:jazzy-desktop`](/packages/robots/ros) [`ros:jazzy-ros-base`](/packages/robots/ros) [`ros:jazzy-ros-core`](/packages/robots/ros) [`ros:noetic-desktop`](/packages/robots/ros) [`ros:noetic-ros-base`](/packages/robots/ros) [`ros:noetic-ros-core`](/packages/robots/ros) [`sam`](/packages/vit/sam) [`sapiens`](/packages/vit/sapiens) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`speaches`](/packages/speech/speaches) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`tam`](/packages/vit/tam) [`tensorflow2:2.16.1`](/packages/ml/tensorflow) [`tensorflow2:2.18.0`](/packages/ml/tensorflow) [`tensorflow2:2.19.0`](/packages/ml/tensorflow) [`tensorflow2:2.20.0`](/packages/ml/tensorflow) [`tensorflow2:2.21.0`](/packages/ml/tensorflow) [`tensorflow_graphics:2.18.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.19.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.20.0`](/packages/ml/tensorflow/graphics) [`tensorflow_text:2.18.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.19.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.20.0`](/packages/ml/tensorflow/text) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`torch2trt`](/packages/pytorch/torch2trt) [`torch_tensorrt`](/packages/pytorch/torch_tensorrt) [`tritonserver`](/packages/ml/tritonserver) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vscode:cuda`](/packages/code/vscode) [`vscode:torch`](/packages/code/vscode) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`whisper_trt`](/packages/speech/whisper_trt) [`whisperx`](/packages/speech/whisperx) [`wyoming-piper:1.6.2`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-piper:master`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-whisper:2.5.0`](/packages/smart-home/wyoming/wyoming-whisper) [`wyoming-whisper:master`](/packages/smart-home/wyoming/wyoming-whisper) [`xtts`](/packages/speech/xtts) [`zed:5.0`](/packages/hw/zed) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) | |    Dockerfile | [`Dockerfile.tar`](Dockerfile.tar) | | **`tensorrt:10.4`** | | | :-- | :-- | |    Requires | `L4T ['==r36.*', '==cu126', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) | |    Dockerfile | [`Dockerfile.tar`](Dockerfile.tar) | | **`tensorrt:10.5`** | | | :-- | :-- | |    Requires | `L4T ['==r36.*', '==cu126', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) | |    Dockerfile | [`Dockerfile.tar`](Dockerfile.tar) | | **`tensorrt:10.7`** | | | :-- | :-- | |    Requires | `L4T ['==r36.*', '==cu126', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) | |    Dockerfile | [`Dockerfile.tar`](Dockerfile.tar) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/tensorrt:8.6-r36.2.0`](https://hub.docker.com/r/dustynv/tensorrt/tags) | `2023-12-05` | `arm64` | `6.7GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag tensorrt) # or explicitly specify one of the container images above jetson-containers run dustynv/tensorrt:8.6-r36.2.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/tensorrt:8.6-r36.2.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag tensorrt) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag tensorrt) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build tensorrt ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/ParaAttention/README.md`) # ParaAttention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build) https://github.com/chengzeyi/ParaAttention
CONTAINERS
| **`paraattention:0.4.0`** | | | :-- | :-- | |    Aliases | `paraattention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) | |    Dependants | [`comfyui`](/packages/diffusion/comfyui) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`lobechat`](/packages/llm/lobe_chat) [`sdnext`](/packages/diffusion/sdnext) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag ParaAttention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host ParaAttention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag ParaAttention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag ParaAttention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build ParaAttention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/block-sparse-attention/README.md`) # block-sparse-attention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`block_sparse_attn:0.0.1`** | | | :-- | :-- | |    Aliases | `block_sparse_attn` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) | |    Dependants | [`xattention:0.0.1`](/packages/attention/xattention) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag block-sparse-attention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host block-sparse-attention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag block-sparse-attention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag block-sparse-attention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build block-sparse-attention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/flash-attention/README.md`) # flash-attention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`flash-attention:2.5.7`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flash-attention:2.6.3`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flash-attention:2.7.2.post1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flash-attention:2.7.4.post1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flash-attention:2.8.0.post2`** | | | :-- | :-- | |    Aliases | `flash-attention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) | |    Dependants | [`awq:0.1.0`](/packages/llm/awq) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`exllama:0.1`](/packages/llm/exllama) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`minference:0.1.7`](/packages/llm/minference) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`sudonim:hf`](/packages/llm/sudonim) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`vscode:transformers`](/packages/code/vscode) [`xattention:0.0.1`](/packages/attention/xattention) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtuner`](/packages/vlm/xtuner) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flash-attention:2.8.1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag flash-attention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host flash-attention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag flash-attention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag flash-attention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build flash-attention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/flashinfer/README.md`) # flash-infer > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`flashinfer:0.2.1.post2`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`triton`](/packages/ml/triton) | |    Dependants | [`vllm:0.8.4`](/packages/llm/vllm) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flashinfer:0.2.2`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flashinfer:0.2.2.post1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`triton`](/packages/ml/triton) | |    Dependants | [`vllm:v0.8.5.post1`](/packages/llm/vllm) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flashinfer:0.2.6.post1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`triton`](/packages/ml/triton) | |    Dependants | [`mlc:0.20.0`](/packages/llm/mlc) [`sglang:0.4.6`](/packages/llm/sglang) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flashinfer:0.2.7`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`flashinfer:0.2.8`** | | | :-- | :-- | |    Aliases | `flashinfer` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`triton`](/packages/ml/triton) | |    Dependants | [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`llama-factory`](/packages/llm/llama-factory) [`local_llm`](/packages/llm/local_llm) [`mlc:0.21.0`](/packages/llm/mlc) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag flash-infer) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host flash-infer:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag flash-infer) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag flash-infer) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build flash-infer ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/flexprefill/README.md`) # flexprefill > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`flexprefill:0.1.0`** | | | :-- | :-- | |    Aliases | `flexprefill` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) | |    Dependants | [`comfyui`](/packages/diffusion/comfyui) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`llama-factory`](/packages/llm/llama-factory) [`lobechat`](/packages/llm/lobe_chat) [`minference:0.1.7`](/packages/llm/minference) [`sdnext`](/packages/diffusion/sdnext) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag flexprefill) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host flexprefill:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag flexprefill) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag flexprefill) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build flexprefill ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/huggingface_kernels/README.md`) # huggingface_kernels > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`huggingface_kernels:0.7.0`** | | | :-- | :-- | |    Aliases | `huggingface_kernels` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`huggingface_hub`](/packages/llm/huggingface_hub) [`torchvision`](/packages/pytorch/torchvision) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag huggingface_kernels) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host huggingface_kernels:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag huggingface_kernels) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag huggingface_kernels) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build huggingface_kernels ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/log-linear-attention/README.md`) # log-linear-attention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`log-linear-attention:0.0.1`** | | | :-- | :-- | |    Aliases | `log-linear-attention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`ninja`](/packages/build/ninja) [`torchaudio`](/packages/pytorch/torchaudio) [`causalconv1d`](/packages/ml/mamba/causalconv1d) [`mamba`](/packages/ml/mamba/mamba) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag log-linear-attention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host log-linear-attention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag log-linear-attention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag log-linear-attention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build log-linear-attention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/radial-attention/README.md`) # radial-attention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`radial-attention:0.1.0`** | | | :-- | :-- | |    Aliases | `radial-attention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`flashinfer`](/packages/attention/flash-infer) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag radial-attention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host radial-attention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag radial-attention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag radial-attention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build radial-attention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/sage-attention/README.md`) # sage-attention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`sage-attention:3.0.0`** | | | :-- | :-- | |    Aliases | `sage-attention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) | |    Dependants | [`comfyui`](/packages/diffusion/comfyui) [`framepack`](/packages/diffusion/framepack) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`lobechat`](/packages/llm/lobe_chat) [`sdnext`](/packages/diffusion/sdnext) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag sage-attention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host sage-attention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag sage-attention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag sage-attention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build sage-attention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/sparge-attention/README.md`) # sparge-attention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`sparge-attention:0.1.0`** | | | :-- | :-- | |    Aliases | `sparge-attention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) | |    Dependants | [`comfyui`](/packages/diffusion/comfyui) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`lobechat`](/packages/llm/lobe_chat) [`sdnext`](/packages/diffusion/sdnext) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/thu-ml/SpargeAttn.git |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag sparge-attention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host sparge-attention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag sparge-attention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag sparge-attention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build sparge-attention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/xattention/README.md`) # xattention > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`xattention:0.0.1`** | | | :-- | :-- | |    Aliases | `xattention` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`block_sparse_attn`](/packages/attention/block-sparse-attention) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag xattention) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host xattention:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag xattention) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag xattention) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build xattention ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/attention/xformers/README.md`) # xformers > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`xformers:0.0.32`** | | | :-- | :-- | |    Aliases | `xformers` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) | |    Dependants | [`apex:0.1`](/packages/pytorch/apex) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`exllama:0.1`](/packages/llm/exllama) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`log-linear-attention:0.0.1`](/packages/attention/log-linear-attention) [`minference:0.1.7`](/packages/llm/minference) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim:hf`](/packages/llm/sudonim) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vscode:transformers`](/packages/code/vscode) [`xattention:0.0.1`](/packages/attention/xattention) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtuner`](/packages/vlm/xtuner) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/facebookresearch/xformers | | **`xformers:0.0.33`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/facebookresearch/xformers |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/xformers:r35.2.1`](https://hub.docker.com/r/dustynv/xformers/tags) | `2023-12-06` | `arm64` | `5.8GB` | |   [`dustynv/xformers:r35.3.1`](https://hub.docker.com/r/dustynv/xformers/tags) | `2023-12-14` | `arm64` | `5.8GB` | |   [`dustynv/xformers:r35.4.1`](https://hub.docker.com/r/dustynv/xformers/tags) | `2024-01-09` | `arm64` | `5.9GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag xformers) # or explicitly specify one of the container images above jetson-containers run dustynv/xformers:r35.4.1 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/xformers:r35.4.1 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag xformers) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag xformers) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build xformers ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/code/jupyterlab/README.md`) # jupyterlab > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build) Running the `jupyerlab` container will automatically start a JupyterLab server in the background on port 8888, with default login password `nvidia`. The JupyterLab server logs will be saved to `/data/logs/jupyter.log` if you need to inspect them (this location is automatically mounted under your `jetson-containers/data` directory) To change the default settings, you can set the `$JUPYTER_ROOT`, `$JUPYTER_PORT`, `$JUPYTER_PASSWORD`, and `$JUPYTER_LOG` environment variables when starting the container like so: ```bash jetson-containers run \ --env JUPYTER_ROOT=/home/user \ --env JUPYTER_PORT=8000 \ --env JUPYTER_PASSWORD=password \ --env JUPYTER_LOGS=/dev/null \ $(autotag jupyterlab) ``` The [`/start_jupyter`](./start_jupyter) script is the default CMD that the container runs when it starts - however, if you don't want the JupyterLab server started by default, you can either add a different CMD in your own Dockerfile, or override it at startup: ```bash # skip straight to the terminal instead of starting JupyterLab first jetson-containers run /bin/bash ``` You can then still manually run the [`/start_jupyter`](./start_jupyter) script later when desired.
CONTAINERS
| **`jupyterlab:latest`** | | | :-- | :-- | |    Aliases | `jupyterlab` | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`pycuda`](/packages/cuda/pycuda) [`rust`](/packages/build/rust) | |    Dependants | [`audiocraft`](/packages/speech/audiocraft) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`efficientvit`](/packages/vit/efficientvit) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`llama-index:samples`](/packages/rag/llama-index) [`openpi`](/packages/robots/openpi) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`sam`](/packages/vit/sam) [`tam`](/packages/vit/tam) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | will autostart Jupyter server on port 8888 unless container entry CMD is overridden | | **`jupyterlab:latest-myst`** | | | :-- | :-- | |    Aliases | `jupyterlab:myst` | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`pycuda`](/packages/cuda/pycuda) [`rust`](/packages/build/rust) [`jupyterlab:latest`](/packages/code/jupyterlab) | |    Dependants | [`lerobot`](/packages/robots/lerobot) [`llama-index:samples`](/packages/rag/llama-index) [`openpi`](/packages/robots/openpi) | |    Dockerfile | [`Dockerfile.myst`](Dockerfile.myst) | |    Notes | will autostart Jupyter server on port 8888 unless container entry CMD is overridden | | **`jupyterlab:4.2.0`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`pycuda`](/packages/cuda/pycuda) [`rust`](/packages/build/rust) | |    Dependants | [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | will autostart Jupyter server on port 8888 unless container entry CMD is overridden | | **`jupyterlab:4.2.0-myst`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`pycuda`](/packages/cuda/pycuda) [`rust`](/packages/build/rust) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) | |    Dockerfile | [`Dockerfile.myst`](Dockerfile.myst) | |    Notes | will autostart Jupyter server on port 8888 unless container entry CMD is overridden |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/jupyterlab:r32.7.1`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2024-03-07` | `arm64` | `0.7GB` | |   [`dustynv/jupyterlab:r35.2.1`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2023-12-06` | `arm64` | `5.3GB` | |   [`dustynv/jupyterlab:r35.3.1`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2024-03-07` | `arm64` | `5.4GB` | |   [`dustynv/jupyterlab:r35.4.1`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2023-10-07` | `arm64` | `5.3GB` | |   [`dustynv/jupyterlab:r36.2.0`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2024-03-07` | `arm64` | `0.6GB` | |   [`dustynv/jupyterlab:r36.4.0`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2025-03-10` | `arm64` | `5.8GB` | |   [`dustynv/jupyterlab:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/jupyterlab/tags) | `2025-03-03` | `arm64` | `5.1GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag jupyterlab) # or explicitly specify one of the container images above jetson-containers run dustynv/jupyterlab:r36.4.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/jupyterlab:r36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag jupyterlab) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag jupyterlab) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build jupyterlab ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/code/vscode/README.md`) # vscode > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`vscode`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache`](/packages/cuda/cuda) [`python`](/packages/build/python) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | installs portable build of VSCode inside container | | **`vscode:cuda`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`tensorrt`](/packages/cuda/tensorrt) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | installs portable build of VSCode inside container | | **`vscode:torch`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`torchao`](/packages/pytorch/torchao) [`tensorrt`](/packages/cuda/tensorrt) [`torch2trt`](/packages/pytorch/torch2trt) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | installs portable build of VSCode inside container | | **`vscode:transformers`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`bitsandbytes`](/packages/llm/bitsandbytes) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | installs portable build of VSCode inside container |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag vscode) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host vscode:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag vscode) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag vscode) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build vscode ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cuda/cuda-python/README.md`) # cuda-python > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cuda-python:12.2`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) [`cuda:12.2`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.4`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu124`](/packages/cuda/cuda) [`cuda:12.4`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.6`** | | | :-- | :-- | |    Aliases | `cuda-python` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dependants | [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`cudf:23.10.03`](/packages/ml/rapids/cudf) [`cudf:25.08.00`](/packages/ml/rapids/cudf) [`cuml`](/packages/ml/rapids/cuml) [`cutlass:4.1.0`](/packages/cuda/cutlass) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faiss_lite`](/packages/vectordb/faiss_lite) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-index:samples`](/packages/rag/llama-index) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`minference:0.1.7`](/packages/llm/minference) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`raft`](/packages/ml/rapids/raft) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`sudonim:hf`](/packages/llm/sudonim) [`tam`](/packages/vit/tam) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vscode:transformers`](/packages/code/vscode) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtuner`](/packages/vlm/xtuner) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.8`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu128`](/packages/cuda/cuda) [`cuda:12.8`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:12.9`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu129`](/packages/cuda/cuda) [`cuda:12.9`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cuda-python:13.0`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu130`](/packages/cuda/cuda) [`cuda:13.0`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/cuda-python:builder-r35.4.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2024-03-26` | `arm64` | `5.1GB` | |   [`dustynv/cuda-python:builder-r36.2.0`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2024-03-11` | `arm64` | `3.6GB` | |   [`dustynv/cuda-python:r35.2.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-12-06` | `arm64` | `5.0GB` | |   [`dustynv/cuda-python:r35.3.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-08-29` | `arm64` | `5.0GB` | |   [`dustynv/cuda-python:r35.4.1`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-12-06` | `arm64` | `5.0GB` | |   [`dustynv/cuda-python:r36.2.0`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2023-12-06` | `arm64` | `3.5GB` | |   [`dustynv/cuda-python:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda-python/tags) | `2025-03-03` | `arm64` | `2.3GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cuda-python) # or explicitly specify one of the container images above jetson-containers run dustynv/cuda-python:r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/cuda-python:r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cuda-python) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cuda-python) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cuda-python ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cuda/cuda/README.md`) # cuda > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cuda:12.2`** | | | :-- | :-- | |    Requires | `L4T ['==36.*', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.2`](/packages/cuda/cuda-python) [`cuda:12.2-samples`](/packages/cuda/cuda) [`cudnn:8.9`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/cuda:12.2-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) `(2023-12-05, 3.4GB)`
[`dustynv/cuda:12.2-samples-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) `(2023-12-07, 4.8GB)` | | **`pip_cache:cu122`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | | |    Dependants | [`cuda-python:12.2`](/packages/cuda/cuda-python) [`cuda:12.2`](/packages/cuda/cuda) [`cuda:12.2-samples`](/packages/cuda/cuda) [`cudnn:8.9`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.2-samples`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu122`](/packages/cuda/cuda) [`cuda:12.2`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Images | [`dustynv/cuda:12.2-samples-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) `(2023-12-07, 4.8GB)` | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.4`** | | | :-- | :-- | |    Requires | `L4T ['==36.*', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu124`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.4`](/packages/cuda/cuda-python) [`cuda:12.4-samples`](/packages/cuda/cuda) [`cudnn:9.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu124`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | | |    Dependants | [`cuda-python:12.4`](/packages/cuda/cuda-python) [`cuda:12.4`](/packages/cuda/cuda) [`cuda:12.4-samples`](/packages/cuda/cuda) [`cudnn:9.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.4-samples`** | | | :-- | :-- | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu124`](/packages/cuda/cuda) [`cuda:12.4`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.6`** | | | :-- | :-- | |    Aliases | `cuda` | |    Requires | `L4T ['==36.*', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) | |    Dependants | [`3d_diffusion_policy`](/packages/diffusion/3d_diffusion_policy) [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`4k4d:0.0.0`](/packages/3d/gaussian_splatting/4k4d) [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`apex:0.1`](/packages/pytorch/apex) [`arrow:12.0.1`](/packages/numeric/arrow) [`arrow:14.0.1`](/packages/numeric/arrow) [`arrow:19.0.1`](/packages/numeric/arrow) [`arrow:20.0.0`](/packages/numeric/arrow) [`arrow:5.0.0`](/packages/numeric/arrow) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`bitsandbytes:0.39.1`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.4`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.5`](/packages/llm/bitsandbytes) [`bitsandbytes:0.46.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.47.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.48.0`](/packages/llm/bitsandbytes) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`cache_dit:0.2.0`](/packages/diffusion/cache_edit) [`causalconv1d:1.6.0`](/packages/ml/mamba/causalconv1d) [`chatterbox-tts`](/packages/speech/chatterbox-tts) [`clip_trt`](/packages/vit/clip_trt) [`cobra:0.0.1`](/packages/ml/mamba/cobra) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`crossformer`](/packages/physicalAI/vla/crossformer) [`ctranslate2:4.6.0`](/packages/ml/ctranslate2) [`ctranslate2:master`](/packages/ml/ctranslate2) [`cuda-python:12.6`](/packages/cuda/cuda-python) [`cuda:12.6-samples`](/packages/cuda/cuda) [`cudf:23.10.03`](/packages/ml/rapids/cudf) [`cudf:25.08.00`](/packages/ml/rapids/cudf) [`cudnn:9.3`](/packages/cuda/cudnn) [`cudnn:9.4`](/packages/cuda/cudnn) [`cuml`](/packages/ml/rapids/cuml) [`cupy`](/packages/numeric/cupy) [`cutlass:4.1.0`](/packages/cuda/cutlass) [`cv-cuda:0.15`](/packages/cv/cv-cuda) [`cv-cuda:0.15-cpp`](/packages/cv/cv-cuda) [`decord2:1.0.0`](/packages/multimedia/decord) [`deepspeed-kernels:0.1.0`](/packages/llm/deepspeed/deepspeed-kernels) [`deepspeed:0.15.2`](/packages/llm/deepspeed) [`deepspeed:0.18.0`](/packages/llm/deepspeed) [`deepspeed:0.9.5`](/packages/llm/deepspeed) [`deepstream`](/packages/cv/deepstream) [`diffusers:0.35.0`](/packages/diffusion/diffusers) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dimba:1.0`](/packages/ml/mamba/dimba) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`easyvolcap:0.0.0`](/packages/3d/gaussian_splatting/easyvolcap) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faiss:1.12.0`](/packages/vectordb/faiss) [`faiss_lite`](/packages/vectordb/faiss_lite) [`fast_gauss:1.0.0`](/packages/3d/gaussian_splatting/fast_gauss) [`faster-whisper`](/packages/speech/faster-whisper) [`ffmpeg:7.1`](/packages/multimedia/ffmpeg) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flashinfer:0.2.1.post2`](/packages/attention/flash-infer) [`flashinfer:0.2.2`](/packages/attention/flash-infer) [`flashinfer:0.2.2.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.6.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.7`](/packages/attention/flash-infer) [`flashinfer:0.2.8`](/packages/attention/flash-infer) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`gdrcopy`](/packages/cuda/gdrcopy) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`genesis-world:0.2.2`](/packages/sim/genesis) [`glomap:2.0.0`](/packages/3d/3dvision/glomap) [`gptqmodel:3.0.1`](/packages/llm/gptqmodel) [`gsplat:1.5.3`](/packages/3d/gaussian_splatting/gsplat) [`gstreamer`](/packages/multimedia/gstreamer) [`habitat-sim:0.3.4`](/packages/sim/habitat-sim) [`hloc:1.4`](/packages/3d/3dvision/hloc) [`hloc:1.5`](/packages/3d/3dvision/hloc) [`holoscan`](/packages/cv/holoscan) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaaclab:2.2.0`](/packages/sim/isaac-sim/isaac-lab) [`isaacsim:5.0.0`](/packages/sim/isaac-sim) [`jax:0.4.38`](/packages/ml/jax) [`jax:0.6.3`](/packages/ml/jax) [`jetcam`](/packages/hw/jetcam) [`jetson-copilot`](/packages/rag/jetson-copilot) [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`jetson-inference:main`](/packages/cv/jetson-inference) [`jetson-utils:v1`](/packages/multimedia/jetson-utils) [`jetson-utils:v2`](/packages/multimedia/jetson-utils) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`kai_scheduler:0.5.5`](/packages/llm/dynamo/kai-scheduler) [`kaolin:0.18.0`](/packages/3d/3dvision/kaolin) [`kat:1`](/packages/ml/kans/kat) [`kokoro-tts:fastapi`](/packages/speech/kokoro-tts/kokoro-tts-fastapi) [`kokoro-tts:hf`](/packages/speech/kokoro-tts/kokoro-tts-hf) [`kokoro-tts:onnx`](/packages/speech/kokoro-tts/kokoro-tts-onnx) [`ktransformers:0.3.3`](/packages/llm/ktransformers) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain`](/packages/rag/langchain) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`libcom:0.1.0`](/packages/multimedia/libcom) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-index`](/packages/rag/llama-index) [`llama-index:samples`](/packages/rag/llama-index) [`llama-vision`](/packages/vlm/llama-vision) [`llama_cpp:0.2.57`](/packages/llm/llama_cpp) [`llama_cpp:0.2.70`](/packages/llm/llama_cpp) [`llama_cpp:0.2.83`](/packages/llm/llama_cpp) [`llama_cpp:0.2.90`](/packages/llm/llama_cpp) [`llama_cpp:0.3.1`](/packages/llm/llama_cpp) [`llama_cpp:0.3.2`](/packages/llm/llama_cpp) [`llama_cpp:0.3.5`](/packages/llm/llama_cpp) [`llama_cpp:0.3.6`](/packages/llm/llama_cpp) [`llama_cpp:0.3.7`](/packages/llm/llama_cpp) [`llama_cpp:0.3.8`](/packages/llm/llama_cpp) [`llama_cpp:0.3.9`](/packages/llm/llama_cpp) [`llama_cpp:0.4.0`](/packages/llm/llama_cpp) [`llama_cpp:b5255`](/packages/llm/llama_cpp) [`llama_cpp:b5833`](/packages/llm/llama_cpp) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`log-linear-attention:0.0.1`](/packages/attention/log-linear-attention) [`mamba:2.2.5`](/packages/ml/mamba/mamba) [`mambavision:1.0`](/packages/ml/mamba/mambavision) [`memvid:0.1.4`](/packages/rag/memvid) [`mimicgen`](/packages/sim/mimicgen) [`minference:0.1.7`](/packages/llm/minference) [`minigpt4`](/packages/vlm/minigpt4) [`mistral_rs`](/packages/llm/mistral_rs) [`mlc:0.1.0`](/packages/llm/mlc) [`mlc:0.1.1`](/packages/llm/mlc) [`mlc:0.1.2`](/packages/llm/mlc) [`mlc:0.1.3`](/packages/llm/mlc) [`mlc:0.1.4`](/packages/llm/mlc) [`mlc:0.19.0`](/packages/llm/mlc) [`mlc:0.20.0`](/packages/llm/mlc) [`mlc:0.21.0`](/packages/llm/mlc) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`mooncake:0.3.5`](/packages/llm/dynamo/mooncake) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nanoowl`](/packages/vit/nanoowl) [`nanosam`](/packages/vit/nanosam) [`nccl`](/packages/cuda/nccl) [`nemo`](/packages/llm/nemo) [`nerfacc:0.5.3`](/packages/3d/3dvision/nerfacc) [`nerfacc:0.5.4`](/packages/3d/3dvision/nerfacc) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`nerfview:0.1.4`](/packages/3d/gaussian_splatting/nerfview) [`nixl:0.3.2`](/packages/llm/dynamo/nixl) [`numba`](/packages/numeric/numba) [`nvdiffrast:0.3.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer/nvdiffrast) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`nvpl`](/packages/cuda/nvpl) [`octo`](/packages/physicalAI/vla/octo) [`ollama:0.10.0`](/packages/llm/ollama) [`ollama:0.4.0`](/packages/llm/ollama) [`ollama:0.5.1`](/packages/llm/ollama) [`ollama:0.5.5`](/packages/llm/ollama) [`ollama:0.5.7`](/packages/llm/ollama) [`ollama:0.6.7`](/packages/llm/ollama) [`ollama:0.7.0`](/packages/llm/ollama) [`ollama:0.8.0`](/packages/llm/ollama) [`ollama:0.9.6`](/packages/llm/ollama) [`onnxruntime:1.19.2`](/packages/ml/onnxruntime) [`onnxruntime:1.20`](/packages/ml/onnxruntime) [`onnxruntime:1.20.1`](/packages/ml/onnxruntime) [`onnxruntime:1.21`](/packages/ml/onnxruntime) [`onnxruntime:1.22`](/packages/ml/onnxruntime) [`onnxruntime_genai:0.8.5`](/packages/ml/onnxruntime_genai) [`open3d:1.19.0`](/packages/3d/3dvision/open3d) [`opencv:4.10.0`](/packages/cv/opencv) [`opencv:4.10.0-meta`](/packages/cv/opencv) [`opencv:4.11.0`](/packages/cv/opencv) [`opencv:4.11.0-meta`](/packages/cv/opencv) [`opencv:4.12.0`](/packages/cv/opencv) [`opencv:4.12.0-meta`](/packages/cv/opencv) [`opencv:4.8.1`](/packages/cv/opencv) [`opencv:4.8.1-deb`](/packages/cv/opencv) [`opencv:4.8.1-meta`](/packages/cv/opencv) [`opendronemap`](/packages/robots/opendronemap) [`opendronemap:node`](/packages/robots/opendronemap) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`optimum`](/packages/llm/optimum) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`physx`](/packages/sim/physx) [`piper-tts`](/packages/speech/piper-tts) [`piper1-tts:1.3.0`](/packages/speech/piper1-tts) [`pixsfm:1.0`](/packages/3d/3dvision/pixsfm) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`protomotions:2.5.0`](/packages/robots/protomotions) [`pyav`](/packages/multimedia/pyav) [`pycolmap:3.12`](/packages/3d/3dvision/pycolmap) [`pycolmap:3.13`](/packages/3d/3dvision/pycolmap) [`pycuda`](/packages/cuda/pycuda) [`pykan:0.2.9`](/packages/ml/kans/pykan) [`pytorch3d:0.7.8`](/packages/pytorch/torch3d) [`pytorch3d:0.7.9`](/packages/pytorch/torch3d) [`pytorch:2.1`](/packages/pytorch) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`raft`](/packages/ml/rapids/raft) [`realsense`](/packages/hw/realsense) [`robogen`](/packages/sim/robogen) [`robomimic`](/packages/sim/robomimic) [`robopoint`](/packages/physicalAI/vla/robopoint) [`robosuite`](/packages/sim/robosuite) [`ros:foxy-desktop`](/packages/robots/ros) [`ros:foxy-ros-base`](/packages/robots/ros) [`ros:foxy-ros-core`](/packages/robots/ros) [`ros:galactic-desktop`](/packages/robots/ros) [`ros:galactic-ros-base`](/packages/robots/ros) [`ros:galactic-ros-core`](/packages/robots/ros) [`ros:humble-desktop`](/packages/robots/ros) [`ros:humble-ros-base`](/packages/robots/ros) [`ros:humble-ros-core`](/packages/robots/ros) [`ros:iron-desktop`](/packages/robots/ros) [`ros:iron-ros-base`](/packages/robots/ros) [`ros:iron-ros-core`](/packages/robots/ros) [`ros:jazzy-desktop`](/packages/robots/ros) [`ros:jazzy-ros-base`](/packages/robots/ros) [`ros:jazzy-ros-core`](/packages/robots/ros) [`ros:noetic-desktop`](/packages/robots/ros) [`ros:noetic-ros-base`](/packages/robots/ros) [`ros:noetic-ros-core`](/packages/robots/ros) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sapien`](/packages/sim/sapien) [`sapiens`](/packages/vit/sapiens) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sound-utils`](/packages/multimedia/sound-utils) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`spark-tts`](/packages/speech/spark-tts) [`speaches`](/packages/speech/speaches) [`splashsurf`](/packages/sim/genesis/splashSurf) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim:hf`](/packages/llm/sudonim) [`quadrants:1.8.0`](/packages/sim/genesis/quadrants) [`tam`](/packages/vit/tam) [`tensorflow2:2.16.1`](/packages/ml/tensorflow) [`tensorflow2:2.18.0`](/packages/ml/tensorflow) [`tensorflow2:2.19.0`](/packages/ml/tensorflow) [`tensorflow2:2.20.0`](/packages/ml/tensorflow) [`tensorflow2:2.21.0`](/packages/ml/tensorflow) [`tensorflow_graphics:2.18.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.19.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.20.0`](/packages/ml/tensorflow/graphics) [`tensorflow_text:2.18.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.19.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.20.0`](/packages/ml/tensorflow/text) [`tensorrt:10.3`](/packages/cuda/tensorrt) [`tensorrt:10.4`](/packages/cuda/tensorrt) [`tensorrt:10.5`](/packages/cuda/tensorrt) [`tensorrt:10.7`](/packages/cuda/tensorrt) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-inference`](/packages/llm/text-generation-inference) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`tinycudann:1.7`](/packages/3d/3dvision/tinycudann) [`torch-memory-saver:0.0.7`](/packages/pytorch/torchsaver) [`torch2trt`](/packages/pytorch/torch2trt) [`torch_tensorrt`](/packages/pytorch/torch_tensorrt) [`torchao:0.12.0`](/packages/pytorch/torchao) [`torchao:0.13.0`](/packages/pytorch/torchao) [`torchaudio:2.1.0`](/packages/pytorch/torchaudio) [`torchaudio:2.2.2`](/packages/pytorch/torchaudio) [`torchaudio:2.3.0`](/packages/pytorch/torchaudio) [`torchaudio:2.4.0`](/packages/pytorch/torchaudio) [`torchaudio:2.5.0`](/packages/pytorch/torchaudio) [`torchaudio:2.6.0`](/packages/pytorch/torchaudio) [`torchaudio:2.7.0`](/packages/pytorch/torchaudio) [`torchaudio:2.8.0`](/packages/pytorch/torchaudio) [`torchsde:0.2.1`](/packages/pytorch/torchsde) [`torchsde:0.2.2`](/packages/pytorch/torchsde) [`torchsde:0.2.3`](/packages/pytorch/torchsde) [`torchsde:0.2.4`](/packages/pytorch/torchsde) [`torchsde:0.2.5`](/packages/pytorch/torchsde) [`torchsde:0.2.6`](/packages/pytorch/torchsde) [`torchsde:0.2.7`](/packages/pytorch/torchsde) [`torchsde:0.2.8`](/packages/pytorch/torchsde) [`torchtext:0.18.0`](/packages/pytorch/torchtext) [`torchtext:0.19.0`](/packages/pytorch/torchtext) [`torchvision:0.16.2`](/packages/pytorch/torchvision) [`torchvision:0.17.2`](/packages/pytorch/torchvision) [`torchvision:0.18.0`](/packages/pytorch/torchvision) [`torchvision:0.19.1`](/packages/pytorch/torchvision) [`torchvision:0.20.0`](/packages/pytorch/torchvision) [`torchvision:0.21.0`](/packages/pytorch/torchvision) [`torchvision:0.22.0`](/packages/pytorch/torchvision) [`torchvision:0.23.0`](/packages/pytorch/torchvision) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`transformers:4.53.1`](/packages/llm/transformers) [`triton:3.0.0`](/packages/ml/triton) [`triton:3.1.0`](/packages/ml/triton) [`triton:3.2.0`](/packages/ml/triton) [`triton:3.3.0`](/packages/ml/triton) [`triton:3.3.1`](/packages/ml/triton) [`triton:3.4.0`](/packages/ml/triton) [`triton:3.5.0`](/packages/ml/triton) [`tritonserver`](/packages/ml/tritonserver) [`tvm`](/packages/ml/tvm) [`video-codec-sdk:12.2.72`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:12.2.72-samples`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19-samples`](/packages/multimedia/video-codec-sdk) [`videollama:1.0.0`](/packages/vlm/videollama) [`videomambasuite:1.0`](/packages/ml/mamba/videomambasuite) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vpi`](/packages/cv/vpi) [`vscode:cuda`](/packages/code/vscode) [`vscode:torch`](/packages/code/vscode) [`vscode:transformers`](/packages/code/vscode) [`warp:1.7.0`](/packages/numeric/warp) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.7.0-jax`](/packages/numeric/warp) [`warp:1.7.0-torch`](/packages/numeric/warp) [`warp:1.8.1`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`warp:1.8.1-jax`](/packages/numeric/warp) [`warp:1.8.1-torch`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`whisper_trt`](/packages/speech/whisper_trt) [`whisperx`](/packages/speech/whisperx) [`wyoming-piper:1.6.2`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-piper:master`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-whisper:2.5.0`](/packages/smart-home/wyoming/wyoming-whisper) [`wyoming-whisper:master`](/packages/smart-home/wyoming/wyoming-whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xformers:0.0.32`](/packages/attention/xformers) [`xformers:0.0.33`](/packages/attention/xformers) [`xgrammar:0.1.15`](/packages/llm/xgrammar) [`xgrammar:0.1.18`](/packages/llm/xgrammar) [`xgrammar:0.1.19`](/packages/llm/xgrammar) [`xgrammar:0.1.20`](/packages/llm/xgrammar) [`xgrammar:0.1.21`](/packages/llm/xgrammar) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtts`](/packages/speech/xtts) [`xtuner`](/packages/vlm/xtuner) [`zed:5.0`](/packages/hw/zed) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu126`** | | | :-- | :-- | |    Aliases | `pip_cache` `pip-cache` | |    Requires | `L4T ['==36.*']` | |    Dependencies | | |    Dependants | [`3d_diffusion_policy`](/packages/diffusion/3d_diffusion_policy) [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`4k4d:0.0.0`](/packages/3d/gaussian_splatting/4k4d) [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`aim`](/packages/ml/aim) [`apex:0.1`](/packages/pytorch/apex) [`arrow:12.0.1`](/packages/numeric/arrow) [`arrow:14.0.1`](/packages/numeric/arrow) [`arrow:19.0.1`](/packages/numeric/arrow) [`arrow:20.0.0`](/packages/numeric/arrow) [`arrow:5.0.0`](/packages/numeric/arrow) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`bitsandbytes:0.39.1`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.4`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.5`](/packages/llm/bitsandbytes) [`bitsandbytes:0.46.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.47.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.48.0`](/packages/llm/bitsandbytes) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`cache_dit:0.2.0`](/packages/diffusion/cache_edit) [`canbus`](/packages/hw/canbus) [`cangaroo`](/packages/hw/cangaroo) [`causalconv1d:1.6.0`](/packages/ml/mamba/causalconv1d) [`chatterbox-tts`](/packages/speech/chatterbox-tts) [`ciso8601:v2.3.2`](/packages/smart-home/dependencies/ciso8601) [`clip_trt`](/packages/vit/clip_trt) [`cmake:pip`](/packages/build/cmake/cmake_pip) [`cobra:0.0.1`](/packages/ml/mamba/cobra) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`crossformer`](/packages/physicalAI/vla/crossformer) [`ctranslate2:4.6.0`](/packages/ml/ctranslate2) [`ctranslate2:master`](/packages/ml/ctranslate2) [`cuda-python:12.6`](/packages/cuda/cuda-python) [`cuda:12.6`](/packages/cuda/cuda) [`cuda:12.6-samples`](/packages/cuda/cuda) [`cudf:23.10.03`](/packages/ml/rapids/cudf) [`cudf:25.08.00`](/packages/ml/rapids/cudf) [`cudnn:9.3`](/packages/cuda/cudnn) [`cudnn:9.4`](/packages/cuda/cudnn) [`cuml`](/packages/ml/rapids/cuml) [`cupy`](/packages/numeric/cupy) [`cutlass:4.1.0`](/packages/cuda/cutlass) [`cv-cuda:0.15`](/packages/cv/cv-cuda) [`cv-cuda:0.15-cpp`](/packages/cv/cv-cuda) [`decord2:1.0.0`](/packages/multimedia/decord) [`deepspeed-kernels:0.1.0`](/packages/llm/deepspeed/deepspeed-kernels) [`deepspeed:0.15.2`](/packages/llm/deepspeed) [`deepspeed:0.18.0`](/packages/llm/deepspeed) [`deepspeed:0.9.5`](/packages/llm/deepspeed) [`deepstream`](/packages/cv/deepstream) [`diffusers:0.35.0`](/packages/diffusion/diffusers) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dimba:1.0`](/packages/ml/mamba/dimba) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`docker`](/packages/build/docker) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`easyvolcap:0.0.0`](/packages/3d/gaussian_splatting/easyvolcap) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faiss:1.12.0`](/packages/vectordb/faiss) [`faiss_lite`](/packages/vectordb/faiss_lite) [`fast_gauss:1.0.0`](/packages/3d/gaussian_splatting/fast_gauss) [`faster-whisper`](/packages/speech/faster-whisper) [`ffmpeg:7.1`](/packages/multimedia/ffmpeg) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flashinfer:0.2.1.post2`](/packages/attention/flash-infer) [`flashinfer:0.2.2`](/packages/attention/flash-infer) [`flashinfer:0.2.2.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.6.post1`](/packages/attention/flash-infer) [`flashinfer:0.2.7`](/packages/attention/flash-infer) [`flashinfer:0.2.8`](/packages/attention/flash-infer) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`gdrcopy`](/packages/cuda/gdrcopy) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`genesis-world:0.2.2`](/packages/sim/genesis) [`glomap:2.0.0`](/packages/3d/3dvision/glomap) [`gptqmodel:3.0.1`](/packages/llm/gptqmodel) [`graphiti`](/packages/rag/graphiti) [`gsplat:1.5.3`](/packages/3d/gaussian_splatting/gsplat) [`gstreamer`](/packages/multimedia/gstreamer) [`h5py`](/packages/build/h5py) [`habitat-sim:0.3.4`](/packages/sim/habitat-sim) [`hloc:1.4`](/packages/3d/3dvision/hloc) [`hloc:1.5`](/packages/3d/3dvision/hloc) [`holoscan`](/packages/cv/holoscan) [`homeassistant-core:2025.7.0`](/packages/smart-home/homeassistant-core) [`huggingface_hub`](/packages/llm/huggingface_hub) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaaclab:2.2.0`](/packages/sim/isaac-sim/isaac-lab) [`isaacsim:5.0.0`](/packages/sim/isaac-sim) [`jax:0.4.38`](/packages/ml/jax) [`jax:0.6.3`](/packages/ml/jax) [`jetcam`](/packages/hw/jetcam) [`jetson-copilot`](/packages/rag/jetson-copilot) [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`jetson-inference:main`](/packages/cv/jetson-inference) [`jetson-utils:python`](/packages/multimedia/jetson-utils) [`jetson-utils:v1`](/packages/multimedia/jetson-utils) [`jetson-utils:v2`](/packages/multimedia/jetson-utils) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`kai_scheduler:0.5.5`](/packages/llm/dynamo/kai-scheduler) [`kaolin:0.18.0`](/packages/3d/3dvision/kaolin) [`kat:1`](/packages/ml/kans/kat) [`kokoro-tts:fastapi`](/packages/speech/kokoro-tts/kokoro-tts-fastapi) [`kokoro-tts:hf`](/packages/speech/kokoro-tts/kokoro-tts-hf) [`kokoro-tts:onnx`](/packages/speech/kokoro-tts/kokoro-tts-onnx) [`ktransformers:0.3.3`](/packages/llm/ktransformers) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain`](/packages/rag/langchain) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`libcom:0.1.0`](/packages/multimedia/libcom) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-index`](/packages/rag/llama-index) [`llama-index:samples`](/packages/rag/llama-index) [`llama-vision`](/packages/vlm/llama-vision) [`llama_cpp:0.2.57`](/packages/llm/llama_cpp) [`llama_cpp:0.2.70`](/packages/llm/llama_cpp) [`llama_cpp:0.2.83`](/packages/llm/llama_cpp) [`llama_cpp:0.2.90`](/packages/llm/llama_cpp) [`llama_cpp:0.3.1`](/packages/llm/llama_cpp) [`llama_cpp:0.3.2`](/packages/llm/llama_cpp) [`llama_cpp:0.3.5`](/packages/llm/llama_cpp) [`llama_cpp:0.3.6`](/packages/llm/llama_cpp) [`llama_cpp:0.3.7`](/packages/llm/llama_cpp) [`llama_cpp:0.3.8`](/packages/llm/llama_cpp) [`llama_cpp:0.3.9`](/packages/llm/llama_cpp) [`llama_cpp:0.4.0`](/packages/llm/llama_cpp) [`llama_cpp:b5255`](/packages/llm/llama_cpp) [`llama_cpp:b5833`](/packages/llm/llama_cpp) [`llamaspeak`](/packages/llm/llamaspeak) [`llava`](/packages/vlm/llava) [`llvm:10`](/packages/build/llvm) [`llvm:11`](/packages/build/llvm) [`llvm:12`](/packages/build/llvm) [`llvm:13`](/packages/build/llvm) [`llvm:14`](/packages/build/llvm) [`llvm:15`](/packages/build/llvm) [`llvm:16`](/packages/build/llvm) [`llvm:17`](/packages/build/llvm) [`llvm:18`](/packages/build/llvm) [`llvm:19`](/packages/build/llvm) [`llvm:22`](/packages/build/llvm) [`llvm:22`](/packages/build/llvm) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`log-linear-attention:0.0.1`](/packages/attention/log-linear-attention) [`mamba:2.2.5`](/packages/ml/mamba/mamba) [`mambavision:1.0`](/packages/ml/mamba/mambavision) [`memvid:0.1.4`](/packages/rag/memvid) [`meshlab:MeshLab-2023.12`](/packages/3d/3dvision/meshlab) [`meshlab:MeshLab-2025.03`](/packages/3d/3dvision/meshlab) [`mimicgen`](/packages/sim/mimicgen) [`minference:0.1.7`](/packages/llm/minference) [`minigpt4`](/packages/vlm/minigpt4) [`mistral_rs`](/packages/llm/mistral_rs) [`mlc:0.1.0`](/packages/llm/mlc) [`mlc:0.1.1`](/packages/llm/mlc) [`mlc:0.1.2`](/packages/llm/mlc) [`mlc:0.1.3`](/packages/llm/mlc) [`mlc:0.1.4`](/packages/llm/mlc) [`mlc:0.19.0`](/packages/llm/mlc) [`mlc:0.20.0`](/packages/llm/mlc) [`mlc:0.21.0`](/packages/llm/mlc) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`mooncake:0.3.5`](/packages/llm/dynamo/mooncake) [`mujoco`](/packages/sim/mujoco) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nanoowl`](/packages/vit/nanoowl) [`nanosam`](/packages/vit/nanosam) [`nccl`](/packages/cuda/nccl) [`nemo`](/packages/llm/nemo) [`nerfacc:0.5.3`](/packages/3d/3dvision/nerfacc) [`nerfacc:0.5.4`](/packages/3d/3dvision/nerfacc) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`nerfview:0.1.4`](/packages/3d/gaussian_splatting/nerfview) [`ninja`](/packages/build/ninja) [`nixl:0.3.2`](/packages/llm/dynamo/nixl) [`numba`](/packages/numeric/numba) [`numpy`](/packages/numeric/numpy) [`nvdiffrast:0.3.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer/nvdiffrast) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`nvpl`](/packages/cuda/nvpl) [`octo`](/packages/physicalAI/vla/octo) [`ollama:0.10.0`](/packages/llm/ollama) [`ollama:0.4.0`](/packages/llm/ollama) [`ollama:0.5.1`](/packages/llm/ollama) [`ollama:0.5.5`](/packages/llm/ollama) [`ollama:0.5.7`](/packages/llm/ollama) [`ollama:0.6.7`](/packages/llm/ollama) [`ollama:0.7.0`](/packages/llm/ollama) [`ollama:0.8.0`](/packages/llm/ollama) [`ollama:0.9.6`](/packages/llm/ollama) [`onnx`](/packages/ml/onnx) [`onnxruntime:1.19.2`](/packages/ml/onnxruntime) [`onnxruntime:1.20`](/packages/ml/onnxruntime) [`onnxruntime:1.20.1`](/packages/ml/onnxruntime) [`onnxruntime:1.21`](/packages/ml/onnxruntime) [`onnxruntime:1.22`](/packages/ml/onnxruntime) [`onnxruntime_genai:0.8.5`](/packages/ml/onnxruntime_genai) [`open3d:1.19.0`](/packages/3d/3dvision/open3d) [`openai`](/packages/llm/openai) [`opencv:4.10.0`](/packages/cv/opencv) [`opencv:4.10.0-meta`](/packages/cv/opencv) [`opencv:4.11.0`](/packages/cv/opencv) [`opencv:4.11.0-meta`](/packages/cv/opencv) [`opencv:4.12.0`](/packages/cv/opencv) [`opencv:4.12.0-meta`](/packages/cv/opencv) [`opencv:4.8.1`](/packages/cv/opencv) [`opencv:4.8.1-deb`](/packages/cv/opencv) [`opencv:4.8.1-meta`](/packages/cv/opencv) [`opendronemap`](/packages/robots/opendronemap) [`opendronemap:node`](/packages/robots/opendronemap) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`optimum`](/packages/llm/optimum) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`physx`](/packages/sim/physx) [`piper-tts`](/packages/speech/piper-tts) [`piper1-tts:1.3.0`](/packages/speech/piper1-tts) [`pixsfm:1.0`](/packages/3d/3dvision/pixsfm) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`polyscope:2.5.0`](/packages/3d/3dvision/polyscope) [`prismatic`](/packages/vlm/prismatic) [`protobuf:apt`](/packages/build/protobuf/protobuf_apt) [`protobuf:cpp`](/packages/build/protobuf/protobuf_cpp) [`protomotions:2.5.0`](/packages/robots/protomotions) [`psutil-home-assistant:0.0.1`](/packages/smart-home/dependencies/psutil-home-assistant) [`pyav`](/packages/multimedia/pyav) [`pybind11`](/packages/build/pybind11) [`pyceres:2.5`](/packages/3d/3dvision/pyceres) [`pycolmap:3.12`](/packages/3d/3dvision/pycolmap) [`pycolmap:3.13`](/packages/3d/3dvision/pycolmap) [`pycuda`](/packages/cuda/pycuda) [`pykan:0.2.9`](/packages/ml/kans/pykan) [`pymeshlab:2023.12.post2`](/packages/3d/3dvision/pymeshlab) [`pymeshlab:2023.12.post3`](/packages/3d/3dvision/pymeshlab) [`pymeshlab:2025.6.23.dev0`](/packages/3d/3dvision/pymeshlab) [`python:3.10`](/packages/build/python) [`python:3.11`](/packages/build/python) [`python:3.12`](/packages/build/python) [`python:3.13`](/packages/build/python) [`python:3.14`](/packages/build/python) [`pytorch3d:0.7.8`](/packages/pytorch/torch3d) [`pytorch3d:0.7.9`](/packages/pytorch/torch3d) [`pytorch:2.1`](/packages/pytorch) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`raft`](/packages/ml/rapids/raft) [`realsense`](/packages/hw/realsense) [`riva-client:python`](/packages/speech/riva-client) [`robogen`](/packages/sim/robogen) [`robomimic`](/packages/sim/robomimic) [`robopoint`](/packages/physicalAI/vla/robopoint) [`robosuite`](/packages/sim/robosuite) [`ros:foxy-desktop`](/packages/robots/ros) [`ros:foxy-ros-base`](/packages/robots/ros) [`ros:foxy-ros-core`](/packages/robots/ros) [`ros:galactic-desktop`](/packages/robots/ros) [`ros:galactic-ros-base`](/packages/robots/ros) [`ros:galactic-ros-core`](/packages/robots/ros) [`ros:humble-desktop`](/packages/robots/ros) [`ros:humble-ros-base`](/packages/robots/ros) [`ros:humble-ros-core`](/packages/robots/ros) [`ros:iron-desktop`](/packages/robots/ros) [`ros:iron-ros-base`](/packages/robots/ros) [`ros:iron-ros-core`](/packages/robots/ros) [`ros:jazzy-desktop`](/packages/robots/ros) [`ros:jazzy-ros-base`](/packages/robots/ros) [`ros:jazzy-ros-core`](/packages/robots/ros) [`ros:noetic-desktop`](/packages/robots/ros) [`ros:noetic-ros-base`](/packages/robots/ros) [`ros:noetic-ros-core`](/packages/robots/ros) [`rust`](/packages/build/rust) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sapien`](/packages/sim/sapien) [`sapiens`](/packages/vit/sapiens) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sound-utils`](/packages/multimedia/sound-utils) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`spark-tts`](/packages/speech/spark-tts) [`speaches`](/packages/speech/speaches) [`splashsurf`](/packages/sim/genesis/splashSurf) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim`](/packages/llm/sudonim) [`sudonim:hf`](/packages/llm/sudonim) [`quadrants:1.8.0`](/packages/sim/genesis/quadrants) [`tam`](/packages/vit/tam) [`tensorboard`](/packages/ml/tensorflow/tensorboard) [`tensorflow2:2.16.1`](/packages/ml/tensorflow) [`tensorflow2:2.18.0`](/packages/ml/tensorflow) [`tensorflow2:2.19.0`](/packages/ml/tensorflow) [`tensorflow2:2.20.0`](/packages/ml/tensorflow) [`tensorflow2:2.21.0`](/packages/ml/tensorflow) [`tensorflow_graphics:2.18.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.19.0`](/packages/ml/tensorflow/graphics) [`tensorflow_graphics:2.20.0`](/packages/ml/tensorflow/graphics) [`tensorflow_text:2.18.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.19.0`](/packages/ml/tensorflow/text) [`tensorflow_text:2.20.0`](/packages/ml/tensorflow/text) [`tensorrt:10.3`](/packages/cuda/tensorrt) [`tensorrt:10.4`](/packages/cuda/tensorrt) [`tensorrt:10.5`](/packages/cuda/tensorrt) [`tensorrt:10.7`](/packages/cuda/tensorrt) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-inference`](/packages/llm/text-generation-inference) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`tinycudann:1.7`](/packages/3d/3dvision/tinycudann) [`torch-memory-saver:0.0.7`](/packages/pytorch/torchsaver) [`torch2trt`](/packages/pytorch/torch2trt) [`torch_tensorrt`](/packages/pytorch/torch_tensorrt) [`torchao:0.12.0`](/packages/pytorch/torchao) [`torchao:0.13.0`](/packages/pytorch/torchao) [`torchaudio:2.1.0`](/packages/pytorch/torchaudio) [`torchaudio:2.2.2`](/packages/pytorch/torchaudio) [`torchaudio:2.3.0`](/packages/pytorch/torchaudio) [`torchaudio:2.4.0`](/packages/pytorch/torchaudio) [`torchaudio:2.5.0`](/packages/pytorch/torchaudio) [`torchaudio:2.6.0`](/packages/pytorch/torchaudio) [`torchaudio:2.7.0`](/packages/pytorch/torchaudio) [`torchaudio:2.8.0`](/packages/pytorch/torchaudio) [`torchsde:0.2.1`](/packages/pytorch/torchsde) [`torchsde:0.2.2`](/packages/pytorch/torchsde) [`torchsde:0.2.3`](/packages/pytorch/torchsde) [`torchsde:0.2.4`](/packages/pytorch/torchsde) [`torchsde:0.2.5`](/packages/pytorch/torchsde) [`torchsde:0.2.6`](/packages/pytorch/torchsde) [`torchsde:0.2.7`](/packages/pytorch/torchsde) [`torchsde:0.2.8`](/packages/pytorch/torchsde) [`torchtext:0.18.0`](/packages/pytorch/torchtext) [`torchtext:0.19.0`](/packages/pytorch/torchtext) [`torchvision:0.16.2`](/packages/pytorch/torchvision) [`torchvision:0.17.2`](/packages/pytorch/torchvision) [`torchvision:0.18.0`](/packages/pytorch/torchvision) [`torchvision:0.19.1`](/packages/pytorch/torchvision) [`torchvision:0.20.0`](/packages/pytorch/torchvision) [`torchvision:0.21.0`](/packages/pytorch/torchvision) [`torchvision:0.22.0`](/packages/pytorch/torchvision) [`torchvision:0.23.0`](/packages/pytorch/torchvision) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`transformers:4.53.1`](/packages/llm/transformers) [`triton:3.0.0`](/packages/ml/triton) [`triton:3.1.0`](/packages/ml/triton) [`triton:3.2.0`](/packages/ml/triton) [`triton:3.3.0`](/packages/ml/triton) [`triton:3.3.1`](/packages/ml/triton) [`triton:3.4.0`](/packages/ml/triton) [`triton:3.5.0`](/packages/ml/triton) [`tritonserver`](/packages/ml/tritonserver) [`tvm`](/packages/ml/tvm) [`video-codec-sdk:12.2.72`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:12.2.72-samples`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19`](/packages/multimedia/video-codec-sdk) [`video-codec-sdk:13.0.19-samples`](/packages/multimedia/video-codec-sdk) [`videollama:1.0.0`](/packages/vlm/videollama) [`videomambasuite:1.0`](/packages/ml/mamba/videomambasuite) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vpi`](/packages/cv/vpi) [`vscode`](/packages/code/vscode) [`vscode:cuda`](/packages/code/vscode) [`vscode:torch`](/packages/code/vscode) [`vscode:transformers`](/packages/code/vscode) [`vtk:9.3.1`](/packages/sim/genesis/vtk) [`vtk:9.4.2`](/packages/sim/genesis/vtk) [`vtk:9.5.0`](/packages/sim/genesis/vtk) [`vulkan:1.3.275`](/packages/multimedia/vulkan) [`vulkan:1.3.296`](/packages/multimedia/vulkan) [`vulkan:1.4.309`](/packages/multimedia/vulkan) [`vulkan:1.4.313`](/packages/multimedia/vulkan) [`vulkan:1.4.314`](/packages/multimedia/vulkan) [`warp:1.7.0`](/packages/numeric/warp) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.7.0-jax`](/packages/numeric/warp) [`warp:1.7.0-torch`](/packages/numeric/warp) [`warp:1.8.1`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`warp:1.8.1-jax`](/packages/numeric/warp) [`warp:1.8.1-torch`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) [`whisper_trt`](/packages/speech/whisper_trt) [`whisperx`](/packages/speech/whisperx) [`wyoming-assist-microphone:1.4.1`](/packages/smart-home/wyoming/wyoming-assist-microphone) [`wyoming-assist-microphone:master`](/packages/smart-home/wyoming/wyoming-assist-microphone) [`wyoming-openwakeword:1.10.0`](/packages/smart-home/wyoming/wyoming-openwakeword) [`wyoming-openwakeword:master`](/packages/smart-home/wyoming/wyoming-openwakeword) [`wyoming-piper:1.6.2`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-piper:master`](/packages/smart-home/wyoming/wyoming-piper) [`wyoming-whisper:2.5.0`](/packages/smart-home/wyoming/wyoming-whisper) [`wyoming-whisper:master`](/packages/smart-home/wyoming/wyoming-whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xformers:0.0.32`](/packages/attention/xformers) [`xformers:0.0.33`](/packages/attention/xformers) [`xgrammar:0.1.15`](/packages/llm/xgrammar) [`xgrammar:0.1.18`](/packages/llm/xgrammar) [`xgrammar:0.1.19`](/packages/llm/xgrammar) [`xgrammar:0.1.20`](/packages/llm/xgrammar) [`xgrammar:0.1.21`](/packages/llm/xgrammar) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtts`](/packages/speech/xtts) [`xtuner`](/packages/vlm/xtuner) [`zed:5.0`](/packages/hw/zed) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.6-samples`** | | | :-- | :-- | |    Aliases | `cuda:samples` | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.8`** | | | :-- | :-- | |    Requires | `L4T ['>=36', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu128`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.8`](/packages/cuda/cuda-python) [`cuda:12.8-samples`](/packages/cuda/cuda) [`cudnn:9.8`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu128`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | | |    Dependants | [`cuda-python:12.8`](/packages/cuda/cuda-python) [`cuda:12.8`](/packages/cuda/cuda) [`cuda:12.8-samples`](/packages/cuda/cuda) [`cudnn:9.8`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.8-samples`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu128`](/packages/cuda/cuda) [`cuda:12.8`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:12.9`** | | | :-- | :-- | |    Requires | `L4T ['>=36', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu129`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:12.9`](/packages/cuda/cuda-python) [`cuda:12.9-samples`](/packages/cuda/cuda) [`cudnn:9.10`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/cuda:12.9-samples-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 5.7GB)`
[`dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 2.8GB)` | | **`pip_cache:cu129`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | | |    Dependants | [`cuda-python:12.9`](/packages/cuda/cuda-python) [`cuda:12.9`](/packages/cuda/cuda) [`cuda:12.9-samples`](/packages/cuda/cuda) [`cudnn:9.10`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:12.9-samples`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu129`](/packages/cuda/cuda) [`cuda:12.9`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](Dockerfile.samples) | |    Images | [`dustynv/cuda:12.9-samples-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 5.7GB)`
[`dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) `(2025-04-02, 2.8GB)` | |    Notes | CUDA samples from https://github.com/NVIDIA/cuda-samples installed under /opt/cuda-samples | | **`cuda:13.0`** | | | :-- | :-- | |    Requires | `L4T ['>=36', 'aarch64']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu130`](/packages/cuda/cuda) | |    Dependants | [`cuda-python:13.0`](/packages/cuda/cuda-python) [`cuda:13.0-samples`](/packages/cuda/cuda) [`cudnn:10.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`pip_cache:cu130`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | | |    Dependants | [`cuda-python:13.0`](/packages/cuda/cuda-python) [`cuda:13.0`](/packages/cuda/cuda) [`cuda:13.0-samples`](/packages/cuda/cuda) [`cudnn:10.0`](/packages/cuda/cudnn) | |    Dockerfile | [`Dockerfile.pip`](Dockerfile.pip) | | **`cuda:13.0-samples`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu130`](/packages/cuda/cuda) [`cuda:13.0`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) | |    Dockerfile | [`Dockerfile.samples`](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`](https://hub.docker.com/r/dustynv/cuda/tags) | `2023-12-05` | `arm64` | `3.4GB` | |   [`dustynv/cuda:12.2-samples-r36.2.0`](https://hub.docker.com/r/dustynv/cuda/tags) | `2023-12-07` | `arm64` | `4.8GB` | |   [`dustynv/cuda:12.9-samples-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) | `2025-04-02` | `amd64` | `5.7GB` | |   [`dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/cuda/tags) | `2025-04-02` | `arm64` | `2.8GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cuda) # or explicitly specify one of the container images above jetson-containers run dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/cuda:12.9-samples-r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cuda) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cuda) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cuda ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cuda/cudastack/README.md`) # CUDA Stack - Consolidated Library Installation ## What is cudastack? `cudastack` is a **consolidation package** that installs multiple CUDA libraries (cuDNN, NCCL, TensorRT, CUTLASS, etc.) in **ONE RUN command** to avoid Docker's "max depth exceeded" error. ## Important: cudastack vs cuda ``` cuda/ cudastack/ ├── Dockerfile ├── Dockerfile (ONE RUN for all libs) ├── Dockerfile.pip ├── config.py (variants) ├── Dockerfile.builtin └── install/ & build/ scripts ├── Dockerfile.samples └── config.py ├── cuda_package() cudastack DEPENDS ON cuda ├── cuda_builtin() (doesn't replace it!) ├── cuda_samples() └── pip_cache() ``` **Key Points:** 1. ✅ **`cuda` package** - Installs CUDA toolkit, provides pip_cache, samples, builtin variants 2. ✅ **`cudastack` package** - Builds ON TOP of `cuda`, adds cuDNN/NCCL/TensorRT/etc. in ONE layer 3. ✅ **pip_cache, cuda_samples, cuda_builtin** - Still work! They're part of `cuda`, not `cudastack` ## The Problem It Solves ### Before (Separate Packages): ``` Layer 1: cuda Layer 2: cudnn Layer 3: tensorrt Layer 4: nccl Layer 5: cutlass ... Layer 50+: "max depth exceeded" ❌ ``` ### After (cudastack): ``` Layer 1: cuda (with pip_cache, samples, etc.) Layer 2: cudastack (cudnn + nccl + tensorrt + cutlass in ONE RUN) Layer 3: your-app Result: Works! ✅ ``` ## Architecture ``` BASE_IMAGE ↓ cuda:12.4 (from cuda/Dockerfile) ├── pip_cache:cu124 (from cuda/Dockerfile.pip) └── cuda:12.4-samples (from cuda/Dockerfile.samples) ↓ cudastack:minimal (from cudastack/Dockerfile) ├── cuDNN ├── NCCL └── (all installed in ONE RUN) ↓ your-app ``` ## Variants | Variant | Includes | Use Case | |---------|----------|----------| | `cudastack:minimal` | cuDNN + NCCL | Training, inference | | `cudastack:standard` | + TensorRT | Optimized inference | | `cudastack:full` | + CUTLASS + GDRCopy | Research, development | ## Usage ### Build with jetson-containers: ```bash # Build minimal stack ./build.sh cudastack:minimal # Build standard (with TensorRT) ./build.sh cudastack:standard # Build full (with CUTLASS) ./build.sh cudastack:full # Use as base for your app ./build.sh --base cudastack:minimal pytorch ``` ### The cudastack Dockerfile does ONE RUN: ```dockerfile # Copy all install/build scripts COPY install/*.sh /tmp/cudastack/install/ COPY build/*.sh /tmp/cudastack/build/ # ONE RUN command calls all scripts conditionally RUN set -ex && \ if [ "$WITH_CUDNN" = "1" ]; then \ ./install/install_cudnn.sh; \ fi && \ if [ "$WITH_NCCL" = "1" ]; then \ ./install/install_nccl.sh; \ fi && \ if [ "$WITH_TENSORRT" = "1" ]; then \ ./install/install_tensorrt.sh; \ fi && \ # ... more components ... rm -rf /tmp/cudastack ``` **Key**: All library installations happen in this ONE RUN = ONE Docker layer! ## Setup The install scripts need to be populated from existing packages: ```bash cd packages/cuda/cudastack ./setup-scripts.sh ``` Then review and complete the scripts in `install/` and `build/`. ## What About pip_cache, samples, builtin? These are **still available** from the `cuda` package: ```bash # pip_cache is automatically included by cuda package ./build.sh cuda:12.4 # This creates both: # - cuda:12.4 (main CUDA toolkit) # - pip_cache:cu124 (pip caching) # CUDA samples work as before ./build.sh cuda:12.4-samples # Builtin works as before (for JetPack 4-5) ./build.sh cuda:11.4 # Uses Dockerfile.builtin when CUDA pre-installed ``` The `cuda` package's `config.py` returns multiple packages: ```python # In cuda/config.py def cuda_package(version, url, ...): cuda = package.copy() cuda['name'] = f'cuda:{version}' cuda_pip = pip_cache(version, requires) # Creates pip_cache package return cuda, cuda_pip # Returns TWO packages! ``` **cudastack doesn't need to duplicate this** - it just depends on `cuda` being installed first. ## Component Toggles Controlled by `WITH_*` build args (set in config.py): - ✅ `WITH_CUDNN=1` - cuDNN (always on) - ✅ `WITH_NCCL=1` - NCCL (always on) - 🎛️ `WITH_TENSORRT` - TensorRT (on in standard/full) - 🎛️ `WITH_CUTLASS` - CUTLASS (on in full only) - 🎛️ `WITH_GDRCOPY` - GDRCopy (on for Tegra) - ⚙️ `WITH_CUDSS=0` - cuDSS (off by default) - ⚙️ `WITH_CUSPARSELT=0` - cuSPARSELt (off by default) - ⚙️ `WITH_CUTENSOR=0` - cuTENSOR (off by default) - ⚙️ `WITH_NVPL=0` - NVPL (off by default) - ⚙️ `WITH_NVSHMEM=0` - NVSHMEM (off by default) ## How config.py Works Following the jetson-containers pattern: ```python def cuda_stack(name, with_tensorrt=False, with_cutlass=False, requires=None): """Generate a cudastack package variant""" pkg = package.copy() pkg['name'] = name pkg['depends'] = ['cuda'] # Depends ON cuda, doesn't replace it pkg['build_args'] = { 'WITH_CUDNN': '1', 'WITH_TENSORRT': '1' if with_tensorrt else '0', 'WITH_NCCL': '1', # ... etc } return pkg # Define variants package = [ cuda_stack('cudastack:minimal', with_tensorrt=False), cuda_stack('cudastack:standard', with_tensorrt=True), cuda_stack('cudastack:full', with_tensorrt=True, with_cutlass=True), ] ``` ## Comparison with Individual Packages | Approach | Layers | Build Time | Complexity | |----------|--------|------------|------------| | **Individual packages** | 50+ | Longer | Simple per-package | | **cudastack (consolidated)** | 2-3 | Faster | One complex layer | Both approaches work! Use `cudastack` when you need multiple libraries and want to avoid layer limits. ## See Also - [cuda package](../cuda/) - CUDA toolkit, pip_cache, samples, builtin - [cudnn package](../cudnn/) - Individual cuDNN package - [tensorrt package](../tensorrt/) - Individual TensorRT package (has Dockerfile.deb and Dockerfile.tar) - [nccl package](../nccl/) - Individual NCCL package ## Summary ✅ **cudastack is a consolidation layer** on top of cuda ✅ **pip_cache, samples, builtin still work** from the cuda package ✅ **ONE RUN command** installs all libraries to avoid layer limits ✅ **Follows jetson-containers patterns** with config.py and variants --- ### Overview (`packages/cuda/cutlass/README.md`) # cutlass > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cutlass:4.1.0`** | | | :-- | :-- | |    Aliases | `cutlass` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`cudnn`](/packages/cuda/cudnn) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) | |    Dependants | [`awq:0.1.0`](/packages/llm/awq) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`exllama:0.1`](/packages/llm/exllama) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`minference:0.1.7`](/packages/llm/minference) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`sudonim:hf`](/packages/llm/sudonim) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`vscode:transformers`](/packages/code/vscode) [`xattention:0.0.1`](/packages/attention/xattention) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtuner`](/packages/vlm/xtuner) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cutlass) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host cutlass:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cutlass) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cutlass) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cutlass ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cuda/pycuda/README.md`) # pycuda > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`pycuda`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dependants | [`audiocraft`](/packages/speech/audiocraft) [`cv-cuda:0.15`](/packages/cv/cv-cuda) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`efficientvit`](/packages/vit/efficientvit) [`jupyter_clickable_image_widget`](/packages/hw/jupyter_clickable_image_widget) [`jupyterlab:4.2.0`](/packages/code/jupyterlab) [`jupyterlab:4.2.0-myst`](/packages/code/jupyterlab) [`jupyterlab:latest`](/packages/code/jupyterlab) [`jupyterlab:latest-myst`](/packages/code/jupyterlab) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`llama-index:samples`](/packages/rag/llama-index) [`openpi`](/packages/robots/openpi) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`sam`](/packages/vit/sam) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`tam`](/packages/vit/tam) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`warp:1.7.0-all`](/packages/numeric/warp) [`warp:1.8.1-all`](/packages/numeric/warp) [`whisper`](/packages/speech/whisper) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/pycuda:r32.7.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-12-06, 0.4GB)`
[`dustynv/pycuda:r35.2.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-09-07, 5.0GB)`
[`dustynv/pycuda:r35.3.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-08-29, 5.0GB)`
[`dustynv/pycuda:r35.4.1`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-12-06, 5.0GB)`
[`dustynv/pycuda:r36.2.0`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2023-12-06, 3.5GB)`
[`dustynv/pycuda:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/pycuda/tags) `(2025-03-03, 2.3GB)` |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/pycuda:r32.7.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-12-06` | `arm64` | `0.4GB` | |   [`dustynv/pycuda:r35.2.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-09-07` | `arm64` | `5.0GB` | |   [`dustynv/pycuda:r35.3.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-08-29` | `arm64` | `5.0GB` | |   [`dustynv/pycuda:r35.4.1`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-12-06` | `arm64` | `5.0GB` | |   [`dustynv/pycuda:r36.2.0`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2023-12-06` | `arm64` | `3.5GB` | |   [`dustynv/pycuda:r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/pycuda/tags) | `2025-03-03` | `arm64` | `2.3GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag pycuda) # or explicitly specify one of the container images above jetson-containers run dustynv/pycuda:r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/pycuda:r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag pycuda) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag pycuda) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build pycuda ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cv/cv-cuda/README.md`) # cv-cuda > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cv-cuda:0.15`** | | | :-- | :-- | |    Aliases | `cvcuda:0.15` `cv-cuda` `cvcuda` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg:git`](/packages/multimedia/ffmpeg) [`pyav`](/packages/multimedia/pyav) [`pycuda`](/packages/cuda/pycuda) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`cv-cuda:0.15-cpp`** | | | :-- | :-- | |    Aliases | `cvcuda:0.15-cpp` `cv-cuda:cpp` `cvcuda:cpp` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) | |    Dependants | [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cv-cuda) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host cv-cuda:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cv-cuda) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cv-cuda) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cv-cuda ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cv/deepstream/README.md`) # deepstream > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`deepstream`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`cmake`](/packages/build/cmake/cmake_pip) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`gstreamer`](/packages/multimedia/gstreamer) [`tritonserver`](/packages/ml/tritonserver) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/deepstream:r32.7.1`](https://hub.docker.com/r/dustynv/deepstream/tags) `(2024-03-06, 2.3GB)`
[`dustynv/deepstream:r35.2.1`](https://hub.docker.com/r/dustynv/deepstream/tags) `(2023-12-22, 6.8GB)`
[`dustynv/deepstream:r35.3.1`](https://hub.docker.com/r/dustynv/deepstream/tags) `(2024-03-05, 6.8GB)`
[`dustynv/deepstream:r35.4.1`](https://hub.docker.com/r/dustynv/deepstream/tags) `(2023-12-06, 6.7GB)`
[`dustynv/deepstream:r36.2.0`](https://hub.docker.com/r/dustynv/deepstream/tags) `(2024-03-05, 9.8GB)` | |    Notes | https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_Quickstart.html |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/deepstream:r32.7.1`](https://hub.docker.com/r/dustynv/deepstream/tags) | `2024-03-06` | `arm64` | `2.3GB` | |   [`dustynv/deepstream:r35.2.1`](https://hub.docker.com/r/dustynv/deepstream/tags) | `2023-12-22` | `arm64` | `6.8GB` | |   [`dustynv/deepstream:r35.3.1`](https://hub.docker.com/r/dustynv/deepstream/tags) | `2024-03-05` | `arm64` | `6.8GB` | |   [`dustynv/deepstream:r35.4.1`](https://hub.docker.com/r/dustynv/deepstream/tags) | `2023-12-06` | `arm64` | `6.7GB` | |   [`dustynv/deepstream:r36.2.0`](https://hub.docker.com/r/dustynv/deepstream/tags) | `2024-03-05` | `arm64` | `9.8GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag deepstream) # or explicitly specify one of the container images above jetson-containers run dustynv/deepstream:r32.7.1 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/deepstream:r32.7.1 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag deepstream) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag deepstream) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build deepstream ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cv/holoscan/README.md`) # holoscan > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`holoscan`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch`](/packages/pytorch) [`cupy`](/packages/numeric/cupy) [`tensorrt`](/packages/cuda/tensorrt) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/holoscan:r36.3.0`](https://hub.docker.com/r/dustynv/holoscan/tags) `(2024-07-03, 6.9GB)` | |    Notes | https://docs.nvidia.com/holoscan/sdk-user-guide/overview.html |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/holoscan:r36.3.0`](https://hub.docker.com/r/dustynv/holoscan/tags) | `2024-07-03` | `arm64` | `6.9GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag holoscan) # or explicitly specify one of the container images above jetson-containers run dustynv/holoscan:r36.3.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/holoscan:r36.3.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag holoscan) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag holoscan) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build holoscan ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cv/jetson-inference/README.md`) # jetson-inference > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`jetson-inference:main`** | | | :-- | :-- | |    Aliases | `jetson-inference` | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`tensorrt`](/packages/cuda/tensorrt) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`gstreamer`](/packages/multimedia/gstreamer) | |    Dependants | [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`local_llm`](/packages/llm/local_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`jetson-inference:foxy`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:foxy-desktop`](/packages/robots/ros) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`gstreamer`](/packages/multimedia/gstreamer) [`jetson-inference:main`](/packages/cv/jetson-inference) | |    Dependants | [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) | |    Dockerfile | [`Dockerfile.ros`](Dockerfile.ros) | | **`jetson-inference:galactic`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:galactic-desktop`](/packages/robots/ros) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`gstreamer`](/packages/multimedia/gstreamer) [`jetson-inference:main`](/packages/cv/jetson-inference) | |    Dependants | [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) | |    Dockerfile | [`Dockerfile.ros`](Dockerfile.ros) | | **`jetson-inference:humble`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:humble-desktop`](/packages/robots/ros) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`gstreamer`](/packages/multimedia/gstreamer) [`jetson-inference:main`](/packages/cv/jetson-inference) | |    Dependants | [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) | |    Dockerfile | [`Dockerfile.ros`](Dockerfile.ros) | | **`jetson-inference:iron`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:iron-desktop`](/packages/robots/ros) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`gstreamer`](/packages/multimedia/gstreamer) [`jetson-inference:main`](/packages/cv/jetson-inference) | |    Dependants | [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) | |    Dockerfile | [`Dockerfile.ros`](Dockerfile.ros) | | **`jetson-inference:jazzy`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:jazzy-desktop`](/packages/robots/ros) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`gstreamer`](/packages/multimedia/gstreamer) [`jetson-inference:main`](/packages/cv/jetson-inference) | |    Dockerfile | [`Dockerfile.ros`](Dockerfile.ros) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/jetson-inference:22.06`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2022-09-30` | `amd64` | `6.5GB` | |   [`dustynv/jetson-inference:r32.4.3`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2020-10-27` | `arm64` | `0.9GB` | |   [`dustynv/jetson-inference:r32.4.4`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2021-11-16` | `arm64` | `0.9GB` | |   [`dustynv/jetson-inference:r32.5.0`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2021-08-09` | `arm64` | `0.9GB` | |   [`dustynv/jetson-inference:r32.6.1`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2021-08-24` | `arm64` | `0.9GB` | |   [`dustynv/jetson-inference:r32.7.1`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2023-05-15` | `arm64` | `1.1GB` | |   [`dustynv/jetson-inference:r34.1.0`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2022-04-08` | `arm64` | `5.9GB` | |   [`dustynv/jetson-inference:r34.1.1`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2023-03-18` | `arm64` | `6.1GB` | |   [`dustynv/jetson-inference:r35.1.0`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2023-05-15` | `arm64` | `6.1GB` | |   [`dustynv/jetson-inference:r35.2.1`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2023-05-15` | `arm64` | `6.0GB` | |   [`dustynv/jetson-inference:r35.3.1`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2023-05-15` | `arm64` | `5.6GB` | |   [`dustynv/jetson-inference:r35.4.1`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2023-08-30` | `arm64` | `5.7GB` | |   [`dustynv/jetson-inference:r36.2.0`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2024-06-20` | `arm64` | `7.2GB` | |   [`dustynv/jetson-inference:r36.3.0`](https://hub.docker.com/r/dustynv/jetson-inference/tags) | `2024-06-20` | `arm64` | `7.2GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag jetson-inference) # or explicitly specify one of the container images above jetson-containers run dustynv/jetson-inference:r36.2.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/jetson-inference:r36.2.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag jetson-inference) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag jetson-inference) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build jetson-inference ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cv/opencv/README.md`) # opencv > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`opencv:4.8.1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/opencv:4.8.1-r36.2.0`](https://hub.docker.com/r/dustynv/opencv/tags) `(2023-12-07, 5.1GB)` | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.8.1-meta`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile.meta`](Dockerfile.meta) | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.10.0`** | | | :-- | :-- | |    Aliases | `opencv` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dependants | [`3dgrut:2.0.0`](/packages/3d/gaussian_splatting/3dgrut) [`4k4d:0.0.0`](/packages/3d/gaussian_splatting/4k4d) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`crossformer`](/packages/physicalAI/vla/crossformer) [`deepstream`](/packages/cv/deepstream) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) [`easyvolcap:0.0.0`](/packages/3d/gaussian_splatting/easyvolcap) [`efficientvit`](/packages/vit/efficientvit) [`fast_gauss:1.0.0`](/packages/3d/gaussian_splatting/fast_gauss) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`genesis-world:0.2.2`](/packages/sim/genesis) [`gstreamer`](/packages/multimedia/gstreamer) [`hloc:1.4`](/packages/3d/3dvision/hloc) [`hloc:1.5`](/packages/3d/3dvision/hloc) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`jetcam`](/packages/hw/jetcam) [`jetson-inference:foxy`](/packages/cv/jetson-inference) [`jetson-inference:galactic`](/packages/cv/jetson-inference) [`jetson-inference:humble`](/packages/cv/jetson-inference) [`jetson-inference:iron`](/packages/cv/jetson-inference) [`jetson-inference:jazzy`](/packages/cv/jetson-inference) [`jetson-inference:main`](/packages/cv/jetson-inference) [`jetson-utils:v1`](/packages/multimedia/jetson-utils) [`jetson-utils:v2`](/packages/multimedia/jetson-utils) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-ml`](/packages/ml/l4t/l4t-ml) [`l4t-pytorch`](/packages/ml/l4t/l4t-pytorch) [`l4t-tensorflow:tf1`](/packages/ml/l4t/l4t-tensorflow) [`l4t-tensorflow:tf2`](/packages/ml/l4t/l4t-tensorflow) [`lerobot`](/packages/robots/lerobot) [`libcom:0.1.0`](/packages/multimedia/libcom) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`memvid:0.1.4`](/packages/rag/memvid) [`mimicgen`](/packages/sim/mimicgen) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanoowl`](/packages/vit/nanoowl) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`octo`](/packages/physicalAI/vla/octo) [`open3d:1.19.0`](/packages/3d/3dvision/open3d) [`openpi`](/packages/robots/openpi) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`pixsfm:1.0`](/packages/3d/3dvision/pixsfm) [`protomotions:2.5.0`](/packages/robots/protomotions) [`robogen`](/packages/sim/robogen) [`robomimic`](/packages/sim/robomimic) [`robopoint`](/packages/physicalAI/vla/robopoint) [`robosuite`](/packages/sim/robosuite) [`ros:foxy-desktop`](/packages/robots/ros) [`ros:foxy-ros-base`](/packages/robots/ros) [`ros:foxy-ros-core`](/packages/robots/ros) [`ros:galactic-desktop`](/packages/robots/ros) [`ros:galactic-ros-base`](/packages/robots/ros) [`ros:galactic-ros-core`](/packages/robots/ros) [`ros:humble-desktop`](/packages/robots/ros) [`ros:humble-ros-base`](/packages/robots/ros) [`ros:humble-ros-core`](/packages/robots/ros) [`ros:iron-desktop`](/packages/robots/ros) [`ros:iron-ros-base`](/packages/robots/ros) [`ros:iron-ros-core`](/packages/robots/ros) [`ros:jazzy-desktop`](/packages/robots/ros) [`ros:jazzy-ros-base`](/packages/robots/ros) [`ros:jazzy-ros-core`](/packages/robots/ros) [`ros:noetic-desktop`](/packages/robots/ros) [`ros:noetic-ros-base`](/packages/robots/ros) [`ros:noetic-ros-core`](/packages/robots/ros) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sam`](/packages/vit/sam) [`sapiens`](/packages/vit/sapiens) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim:hf`](/packages/llm/sudonim) [`tam`](/packages/vit/tam) [`videollama:1.0.0`](/packages/vlm/videollama) [`vila`](/packages/vlm/vila) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`xattention:0.0.1`](/packages/attention/xattention) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.10.0-meta`** | | | :-- | :-- | |    Aliases | `opencv:meta` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile.meta`](Dockerfile.meta) | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.11.0`** | | | :-- | :-- | |    Aliases | `opencv` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/opencv:4.11.0-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/opencv/tags) `(2025-04-21, 6.9GB)`
[`dustynv/opencv:4.11.0-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/opencv/tags) `(2025-03-02, 3.3GB)` | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.11.0-meta`** | | | :-- | :-- | |    Aliases | `opencv:meta` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile.meta`](Dockerfile.meta) | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.12.0`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.12.0-meta`** | | | :-- | :-- | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile.meta`](Dockerfile.meta) | |    Notes | https://github.com/opencv/opencv-python | | **`opencv:4.8.1-deb`** | | | :-- | :-- | |    Aliases | `opencv:deb` | |    Requires | `L4T ['==36.*']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/opencv/opencv-python |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/opencv:4.11.0-amd64-cu128-24.04`](https://hub.docker.com/r/dustynv/opencv/tags) | `2025-04-21` | `amd64` | `6.9GB` | |   [`dustynv/opencv:4.11.0-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/opencv/tags) | `2025-03-02` | `arm64` | `3.3GB` | |   [`dustynv/opencv:4.8.1-r36.2.0`](https://hub.docker.com/r/dustynv/opencv/tags) | `2023-12-07` | `arm64` | `5.1GB` | |   [`dustynv/opencv:r32.7.1`](https://hub.docker.com/r/dustynv/opencv/tags) | `2023-12-06` | `arm64` | `0.5GB` | |   [`dustynv/opencv:r35.2.1`](https://hub.docker.com/r/dustynv/opencv/tags) | `2023-12-05` | `arm64` | `5.0GB` | |   [`dustynv/opencv:r35.3.1`](https://hub.docker.com/r/dustynv/opencv/tags) | `2023-08-29` | `arm64` | `5.1GB` | |   [`dustynv/opencv:r35.4.1`](https://hub.docker.com/r/dustynv/opencv/tags) | `2023-10-07` | `arm64` | `5.0GB` | |   [`dustynv/opencv:r36.3.0`](https://hub.docker.com/r/dustynv/opencv/tags) | `2024-10-15` | `arm64` | `5.4GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag opencv) # or explicitly specify one of the container images above jetson-containers run dustynv/opencv:4.11.0-amd64-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/opencv:4.11.0-amd64-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag opencv) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag opencv) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build opencv ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/cv/vpi/README.md`) # vpi > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`vpi`** | | | :-- | :-- | |    Requires | `L4T ['>=36.4']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) [`opengl`](/packages/multimedia/opengl) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) | |    Dependants | [`isaac-ros:common-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:common-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:compression-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:data-tools-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:dnn-inference-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:image-pipeline-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:manipulator-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nitros-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:nvblox-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:pose-estimation-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-humble-desktop`](/packages/robots/isaac-ros) [`isaac-ros:visual-slam-3.2-jazzy-desktop`](/packages/robots/isaac-ros) [`zed:5.0-humble`](/packages/hw/zed) [`zed:5.0-jazzy`](/packages/hw/zed) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag vpi) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host vpi:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag vpi) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag vpi) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build vpi ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/canable/README.md`) # canable > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`canable`** | | | :-- | :-- | |    Aliases | `cannable` | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | Kernel module builder for CANable USB-to-CAN bus adapter (canable.io) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag canable) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host canable:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag canable) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag canable) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build canable ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/canbus/README.md`) # canbus > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`canbus`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache`](/packages/cuda/cuda) [`python`](/packages/build/python) | |    Dependants | [`cangaroo`](/packages/hw/cangaroo) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | CAN bus utilities for Linux and Python |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag canbus) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host canbus:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag canbus) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag canbus) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build canbus ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/cangaroo/README.md`) # cangaroo > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`cangaroo`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache`](/packages/cuda/cuda) [`python`](/packages/build/python) [`canbus`](/packages/hw/canbus) [`cmake`](/packages/build/cmake/cmake_pip) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | CAN bus analyzer (github.com/normaldotcom/cangaroo) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag cangaroo) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host cangaroo:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag cangaroo) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag cangaroo) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build cangaroo ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/jetcam/README.md`) # jetcam > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`jetcam`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`gstreamer`](/packages/multimedia/gstreamer) | |    Dependants | [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/NVIDIA-AI-IOT/jetcam |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag jetcam) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host jetcam:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag jetcam) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag jetcam) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build jetcam ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/jupyter_clickable_image_widget/README.md`) # jupyter_clickable_image_widget > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`jupyter_clickable_image_widget`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`cuda-python`](/packages/cuda/cuda-python) [`pycuda`](/packages/cuda/pycuda) [`rust`](/packages/build/rust) [`jupyterlab`](/packages/code/jupyterlab) | |    Dependants | [`dli-nano-ai`](/packages/ml/dli/dli-nano-ai) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/jaybdub/jupyter_clickable_image_widget |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag jupyter_clickable_image_widget) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host jupyter_clickable_image_widget:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag jupyter_clickable_image_widget) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag jupyter_clickable_image_widget) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build jupyter_clickable_image_widget ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/librealsense/README.md`) # realsense > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`realsense`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda`](/packages/cuda/cuda) [`python`](/packages/build/python) [`cmake`](/packages/build/cmake/cmake_pip) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/realsense:r32.7.1`](https://hub.docker.com/r/dustynv/realsense/tags) `(2024-02-22, 0.8GB)`
[`dustynv/realsense:r35.2.1`](https://hub.docker.com/r/dustynv/realsense/tags) `(2023-08-29, 5.5GB)`
[`dustynv/realsense:r35.3.1`](https://hub.docker.com/r/dustynv/realsense/tags) `(2024-02-22, 5.4GB)`
[`dustynv/realsense:r35.4.1`](https://hub.docker.com/r/dustynv/realsense/tags) `(2023-10-07, 5.5GB)`
[`dustynv/realsense:r36.2.0`](https://hub.docker.com/r/dustynv/realsense/tags) `(2024-02-22, 4.0GB)` | |    Notes | https://github.com/IntelRealSense/librealsense/blob/master/doc/installation_jetson.md |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/realsense:r32.7.1`](https://hub.docker.com/r/dustynv/realsense/tags) | `2024-02-22` | `arm64` | `0.8GB` | |   [`dustynv/realsense:r35.2.1`](https://hub.docker.com/r/dustynv/realsense/tags) | `2023-08-29` | `arm64` | `5.5GB` | |   [`dustynv/realsense:r35.3.1`](https://hub.docker.com/r/dustynv/realsense/tags) | `2024-02-22` | `arm64` | `5.4GB` | |   [`dustynv/realsense:r35.4.1`](https://hub.docker.com/r/dustynv/realsense/tags) | `2023-10-07` | `arm64` | `5.5GB` | |   [`dustynv/realsense:r36.2.0`](https://hub.docker.com/r/dustynv/realsense/tags) | `2024-02-22` | `arm64` | `4.0GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag realsense) # or explicitly specify one of the container images above jetson-containers run dustynv/realsense:r36.2.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/realsense:r36.2.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag realsense) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag realsense) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build realsense ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/oled/README.md`) # oled > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`oled`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/NVIDIA-AI-IOT/jetcard/ |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag oled) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host oled:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag oled) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag oled) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build oled ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/pl2303/README.md`) # pl2303 > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`pl2303`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | Kernel module builder for PL2303 serial-to-USB converter |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag pl2303) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host pl2303:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag pl2303) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag pl2303) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build pl2303 ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/hw/zed/README.md`) # zed > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`zed:5.0`** | | | :-- | :-- | |    Aliases | `zed` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`opengl`](/packages/multimedia/opengl) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | Stereolabs ZED SDK | | **`zed:5.0-humble`** | | | :-- | :-- | |    Aliases | `zed-humble` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:humble-desktop`](/packages/robots/ros) [`vpi`](/packages/cv/vpi) | |    Dockerfile | [`/home/narandill/Projects/fph/Internal-AI-Base-Station-Server/third_party/jetson-containers/packages/robots/ros/Dockerfile.ros2.extras`](/home/narandill/Projects/fph/Internal-AI-Base-Station-Server/third_party/jetson-containers/packages/robots/ros/Dockerfile.ros2.extras) | |    Notes | Stereolabs ZED SDK | | **`zed:5.0-jazzy`** | | | :-- | :-- | |    Aliases | `zed-jazzy` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn:9.3`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`tensorrt`](/packages/cuda/tensorrt) [`numpy`](/packages/numeric/numpy) [`opengl`](/packages/multimedia/opengl) [`cmake`](/packages/build/cmake/cmake_pip) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`pybind11`](/packages/build/pybind11) [`ros:jazzy-desktop`](/packages/robots/ros) [`vpi`](/packages/cv/vpi) | |    Dockerfile | [`/home/narandill/Projects/fph/Internal-AI-Base-Station-Server/third_party/jetson-containers/packages/robots/ros/Dockerfile.ros2.extras`](/home/narandill/Projects/fph/Internal-AI-Base-Station-Server/third_party/jetson-containers/packages/robots/ros/Dockerfile.ros2.extras) | |    Notes | Stereolabs ZED SDK |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/zed:r32.7.1`](https://hub.docker.com/r/dustynv/zed/tags) | `2023-09-07` | `arm64` | `0.6GB` | |   [`dustynv/zed:r35.2.1`](https://hub.docker.com/r/dustynv/zed/tags) | `2023-12-11` | `arm64` | `5.2GB` | |   [`dustynv/zed:r35.3.1`](https://hub.docker.com/r/dustynv/zed/tags) | `2023-08-29` | `arm64` | `5.2GB` | |   [`dustynv/zed:r35.4.1`](https://hub.docker.com/r/dustynv/zed/tags) | `2023-10-07` | `arm64` | `5.1GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag zed) # or explicitly specify one of the container images above jetson-containers run dustynv/zed:r35.2.1 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/zed:r35.2.1 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag zed) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag zed) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build zed ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/awq/README.md`) # awq > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build) * AWQ from https://github.com/mit-han-lab/llm-awq (installed under `/opt/awq`) * AWQ's CUDA kernels require a GPU with `sm_75` or newer (so for Jetson, Orin only) ### Quantization Follow the instructions from https://github.com/mit-han-lab/llm-awq#usage to quantize your model of choice. Or use [`awq/quantize.py`](/packages/llm/awq/quantize.py) ```bash ./run.sh --env HUGGINGFACE_TOKEN= $(./autotag awq) /bin/bash -c \ '/opt/awq/quantize.py --model=$(huggingface-downloader meta-llama/Llama-2-7b-hf) \ --output=/data/models/awq/Llama-2-7b' ``` If you downloaded a model from the [AWQ Model Zoo](https://huggingface.co/datasets/mit-han-lab/awq-model-zoo) that already has the AWQ search results applied, you can load that with `--load_awq` and skip the search step (which can take a while and use lots of memory) ```bash ./run.sh --env HUGGINGFACE_TOKEN= $(./autotag awq) /bin/bash -c \ '/opt/awq/quantize.py --model=$(huggingface-downloader meta-llama/Llama-2-7b-hf) \ --output=/data/models/awq/Llama-2-7b \ --load_awq=/data/models/awq/Llama-2-7b/llama-2-7b-w4-g128.pt' ``` This process will save the model with the real quantized weights (to a file like `$OUTPUT/w4-g128-awq.pt`) ### Inference Benchmark You can use the [`awq/benchmark.py`](/packages/llm/awq/quantize.py) tool to gather performance and memory measurements: ```bash ./run.sh --env HUGGINGFACE_TOKEN= $(./autotag awq) /bin/bash -c \ '/opt/awq/benchmark.py --model=$(huggingface-downloader meta-llama/Llama-2-7b-hf) \ --quant=/data/models/awq/Llama-2-7b/w4-g128-awq.pt' ``` Make sure that you load the output from the quantization steps above with `--quant` (use the model that ends with `-awq.pt`)
CONTAINERS
| **`awq:0.1.0`** | | | :-- | :-- | |    Aliases | `awq` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`opengl`](/packages/multimedia/opengl) [`llvm`](/packages/build/llvm) [`vulkan`](/packages/multimedia/vulkan) [`video-codec-sdk`](/packages/multimedia/video-codec-sdk) [`ffmpeg`](/packages/multimedia/ffmpeg) [`opencv`](/packages/cv/opencv) [`deepspeed-kernels`](/packages/llm/deepspeed/deepspeed-kernels) [`deepspeed:0.9.5`](/packages/llm/deepspeed) [`decord2`](/packages/multimedia/decord) [`vila`](/packages/vlm/vila) [`sudonim`](/packages/llm/sudonim) | |    Dependants | [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/awq:r35.2.1`](https://hub.docker.com/r/dustynv/awq/tags) | `2023-12-14` | `arm64` | `6.1GB` | |   [`dustynv/awq:r35.3.1`](https://hub.docker.com/r/dustynv/awq/tags) | `2023-12-15` | `arm64` | `6.1GB` | |   [`dustynv/awq:r35.4.1`](https://hub.docker.com/r/dustynv/awq/tags) | `2023-12-12` | `arm64` | `6.1GB` | |   [`dustynv/awq:r36.2.0`](https://hub.docker.com/r/dustynv/awq/tags) | `2023-12-15` | `arm64` | `7.8GB` | |   [`dustynv/awq:r36.4.0`](https://hub.docker.com/r/dustynv/awq/tags) | `2025-02-02` | `arm64` | `5.3GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag awq) # or explicitly specify one of the container images above jetson-containers run dustynv/awq:r36.4.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/awq:r36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag awq) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag awq) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build awq ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/bitsandbytes/README.md`) # bitsandbytes > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`bitsandbytes:0.39.1`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/bitsandbytes-foundation/bitsandbytes | | **`bitsandbytes:0.45.4`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/bitsandbytes:0.45.4-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/bitsandbytes/tags) `(2025-03-03, 4.4GB)` | |    Notes | https://github.com/bitsandbytes-foundation/bitsandbytes | | **`bitsandbytes:0.45.5`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/bitsandbytes-foundation/bitsandbytes | | **`bitsandbytes:0.46.0`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/bitsandbytes-foundation/bitsandbytes | | **`bitsandbytes:0.47.0`** | | | :-- | :-- | |    Aliases | `bitsandbytes` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) | |    Dependants | [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`dimba:1.0`](/packages/ml/mamba/dimba) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-vision`](/packages/vlm/llama-vision) [`lobechat`](/packages/llm/lobe_chat) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`prismatic`](/packages/vlm/prismatic) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sudonim:hf`](/packages/llm/sudonim) [`text-generation-inference`](/packages/llm/text-generation-inference) [`videollama:1.0.0`](/packages/vlm/videollama) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`vscode:transformers`](/packages/code/vscode) [`xtuner`](/packages/vlm/xtuner) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/bitsandbytes-foundation/bitsandbytes | | **`bitsandbytes:0.48.0`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/bitsandbytes-foundation/bitsandbytes |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/bitsandbytes:0.45.4-r36.4.0-cu128-24.04`](https://hub.docker.com/r/dustynv/bitsandbytes/tags) | `2025-03-03` | `arm64` | `4.4GB` | |   [`dustynv/bitsandbytes:r35.2.1`](https://hub.docker.com/r/dustynv/bitsandbytes/tags) | `2023-12-15` | `arm64` | `5.9GB` | |   [`dustynv/bitsandbytes:r35.3.1`](https://hub.docker.com/r/dustynv/bitsandbytes/tags) | `2023-12-11` | `arm64` | `6.0GB` | |   [`dustynv/bitsandbytes:r35.4.1`](https://hub.docker.com/r/dustynv/bitsandbytes/tags) | `2023-12-14` | `arm64` | `5.9GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag bitsandbytes) # or explicitly specify one of the container images above jetson-containers run dustynv/bitsandbytes:0.45.4-r36.4.0-cu128-24.04 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/bitsandbytes:0.45.4-r36.4.0-cu128-24.04 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag bitsandbytes) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag bitsandbytes) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build bitsandbytes ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/deepspeed/README.md`) # deepspeed > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`deepspeed:0.9.5`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`deepspeed-kernels`](/packages/llm/deepspeed/deepspeed-kernels) [`triton`](/packages/ml/triton) | |    Dependants | [`awq:0.1.0`](/packages/llm/awq) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`vila`](/packages/vlm/vila) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`deepspeed:0.15.2`** | | | :-- | :-- | |    Aliases | `deepspeed` | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`deepspeed-kernels`](/packages/llm/deepspeed/deepspeed-kernels) [`triton`](/packages/ml/triton) | |    Dependants | [`llama-factory`](/packages/llm/llama-factory) | |    Dockerfile | [`Dockerfile`](Dockerfile) | | **`deepspeed:0.18.0`** | | | :-- | :-- | |    Requires | `L4T ['>=35']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`deepspeed-kernels`](/packages/llm/deepspeed/deepspeed-kernels) [`triton`](/packages/ml/triton) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag deepspeed) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host deepspeed:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag deepspeed) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag deepspeed) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build deepspeed ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/exllama/README.md`) # exllama > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build) This package provides containers for both ExLlama and exllamav3: * `exllama` container uses the https://github.com/jllllll/exllama fork of https://github.com/turboderp/exllama (installed under `/opt/exllama`) * `exllama:v2` container uses https://github.com/turboderp/exllamav3 (installed under `/opt/exllamav3`) Both loaders are also supported in the oobabooga [`text-generation-webui`](/packages/llm/text-generation-webui) container. ### Inference Benchmark Substitute the GPTQ model from [HuggingFace Hub](https://huggingface.co/models?search=gptq) that you want to run (see [exllama compatible models](https://github.com/turboderp/exllama/blob/master/doc/model_compatibility.md)) ```bash ./run.sh --workdir=/opt/exllama $(./autotag exllama) /bin/bash -c \ 'python3 test_benchmark_inference.py --perf --validate -d $(huggingface-downloader TheBloke/Llama-2-7B-GPTQ)' ``` > If the model repository is private or requires authentication, add `--env HUGGINGFACE_TOKEN=` ### Memory Usage | Model | Memory (MB) | |---------------------------------------------------------------------------------|:-----------:| | [`TheBloke/Llama-2-7B-GPTQ`](https://huggingface.co/TheBloke/Llama-2-7B-GPTQ) | 5,200 | | [`TheBloke/Llama-2-13B-GPTQ`](https://huggingface.co/TheBloke/Llama-2-13B-GPTQ) | 9,135 | | [`TheBloke/LLaMA-30b-GPTQ`](https://huggingface.co/TheBloke/LLaMA-30b-GPTQ) | 20,206 | | [`TheBloke/Llama-2-70B-GPTQ`](https://huggingface.co/TheBloke/Llama-2-70B-GPTQ) | 35,462 |
CONTAINERS
| **`exllama:0.1`** | | | :-- | :-- | |    Aliases | `exllama` | |    Requires | `L4T ['>=36']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`huggingface_hub`](/packages/llm/huggingface_hub) [`triton`](/packages/ml/triton) [`torchvision`](/packages/pytorch/torchvision) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) [`flash-attention`](/packages/attention/flash-attention) | |    Dependants | [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/exllama:r35.2.1`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-15` | `arm64` | `5.5GB` | |   [`dustynv/exllama:r35.3.1`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-11` | `arm64` | `5.5GB` | |   [`dustynv/exllama:r35.4.1`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-14` | `arm64` | `5.4GB` | |   [`dustynv/exllama:v1-r36.2.0`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-15` | `arm64` | `7.2GB` | |   [`dustynv/exllama:v2-r35.2.1`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-15` | `arm64` | `5.5GB` | |   [`dustynv/exllama:v2-r35.3.1`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-14` | `arm64` | `5.5GB` | |   [`dustynv/exllama:v2-r35.4.1`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-12` | `arm64` | `5.5GB` | |   [`dustynv/exllama:v2-r36.2.0`](https://hub.docker.com/r/dustynv/exllama/tags) | `2023-12-15` | `arm64` | `7.2GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag exllama) # or explicitly specify one of the container images above jetson-containers run dustynv/exllama:v1-r36.2.0 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/exllama:v1-r36.2.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag exllama) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag exllama) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build exllama ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/gptqmodel/README.md`) # gptqmodel > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build) GPTQModel from https://github.com/ModelCloud/GPTQModel (installed under `/opt/gptqmodel`) ### Inference Benchmark Substitute the GPTQ model from [HuggingFace Hub](https://huggingface.co/models?search=gptq) (or model path) that you want to run: ```bash ./run.sh --workdir=/opt/gptqmodel/examples/benchmark/ $(./autotag auto_gptq) \ python3 generation_speed.py --model_name_or_path TheBloke/LLaMA-7b-GPTQ --use_safetensors --max_new_tokens=128 ``` If you get the error `Exllama kernel does not support query/key/value fusion with act-order`, try adding `--no_inject_fused_attention`
CONTAINERS
| **`gptqmodel:3.0.1`** | | | :-- | :-- | |    Aliases | `gptqmodel` | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) | |    Dependants | [`llama-factory`](/packages/llm/llama-factory) | |    Dockerfile | [`Dockerfile`](Dockerfile) |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag gptqmodel) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host gptqmodel:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag gptqmodel) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag gptqmodel) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build gptqmodel ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/huggingface_hub/README.md`) # huggingface_hub > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`huggingface_hub`** | | | :-- | :-- | |    Requires | `L4T ['>=32.6']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache`](/packages/cuda/cuda) [`python`](/packages/build/python) | |    Dependants | [`3d_diffusion_policy`](/packages/diffusion/3d_diffusion_policy) [`ai-toolkit`](/packages/diffusion/ai-toolkit) [`apex:0.1`](/packages/pytorch/apex) [`audiocraft`](/packages/speech/audiocraft) [`awq:0.1.0`](/packages/llm/awq) [`bitsandbytes:0.39.1`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.4`](/packages/llm/bitsandbytes) [`bitsandbytes:0.45.5`](/packages/llm/bitsandbytes) [`bitsandbytes:0.46.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.47.0`](/packages/llm/bitsandbytes) [`bitsandbytes:0.48.0`](/packages/llm/bitsandbytes) [`block_sparse_attn:0.0.1`](/packages/attention/block-sparse-attention) [`cache_dit:0.2.0`](/packages/diffusion/cache_edit) [`chatterbox-tts`](/packages/speech/chatterbox-tts) [`clip_trt`](/packages/vit/clip_trt) [`cobra:0.0.1`](/packages/ml/mamba/cobra) [`comfyui`](/packages/diffusion/comfyui) [`cosmos-predict2`](/packages/diffusion/cosmos/cosmos-predict2) [`cosmos-reason1`](/packages/diffusion/cosmos/cosmos-reason1) [`cosmos-transfer1`](/packages/diffusion/cosmos/cosmos-transfer1) [`cosmos1-diffusion-renderer:1.0.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer) [`crossformer`](/packages/physicalAI/vla/crossformer) [`deepspeed:0.15.2`](/packages/llm/deepspeed) [`deepspeed:0.18.0`](/packages/llm/deepspeed) [`deepspeed:0.9.5`](/packages/llm/deepspeed) [`diffusers:0.35.0`](/packages/diffusion/diffusers) [`diffusion_policy`](/packages/diffusion/diffusion_policy) [`dimba:1.0`](/packages/ml/mamba/dimba) [`dynamo:0.3.2`](/packages/llm/dynamo/dynamo) [`efficientvit`](/packages/vit/efficientvit) [`exllama:0.1`](/packages/llm/exllama) [`faster-whisper`](/packages/speech/faster-whisper) [`flash-attention:2.5.7`](/packages/attention/flash-attention) [`flash-attention:2.6.3`](/packages/attention/flash-attention) [`flash-attention:2.7.2.post1`](/packages/attention/flash-attention) [`flash-attention:2.7.4.post1`](/packages/attention/flash-attention) [`flash-attention:2.8.0.post2`](/packages/attention/flash-attention) [`flash-attention:2.8.1`](/packages/attention/flash-attention) [`flexprefill:0.1.0`](/packages/attention/flexprefill) [`framepack`](/packages/diffusion/framepack) [`fruitnerf:1.0`](/packages/3d/nerf/fruitnerf) [`genai-bench:0.1.0`](/packages/llm/sglang/genai-bench) [`gptqmodel:3.0.1`](/packages/llm/gptqmodel) [`huggingface_kernels:0.7.0`](/packages/attention/huggingface_kernels) [`hymba`](/packages/llm/hymba) [`isaac-gr00t`](/packages/physicalAI/vla/isaac-gr00t) [`isaaclab:2.2.0`](/packages/sim/isaac-sim/isaac-lab) [`isaacsim:5.0.0`](/packages/sim/isaac-sim) [`jetson-copilot`](/packages/rag/jetson-copilot) [`kat:1`](/packages/ml/kans/kat) [`kokoro-tts:fastapi`](/packages/speech/kokoro-tts/kokoro-tts-fastapi) [`kokoro-tts:hf`](/packages/speech/kokoro-tts/kokoro-tts-hf) [`ktransformers:0.3.3`](/packages/llm/ktransformers) [`l4t-diffusion`](/packages/ml/l4t/l4t-diffusion) [`l4t-dynamo`](/packages/ml/l4t/l4t-dynamo) [`l4t-text-generation`](/packages/ml/l4t/l4t-text-generation) [`langchain`](/packages/rag/langchain) [`langchain:samples`](/packages/rag/langchain) [`lerobot`](/packages/robots/lerobot) [`lita`](/packages/vlm/lita) [`llama-factory`](/packages/llm/llama-factory) [`llama-vision`](/packages/vlm/llama-vision) [`llama_cpp:0.2.57`](/packages/llm/llama_cpp) [`llama_cpp:0.2.70`](/packages/llm/llama_cpp) [`llama_cpp:0.2.83`](/packages/llm/llama_cpp) [`llama_cpp:0.2.90`](/packages/llm/llama_cpp) [`llama_cpp:0.3.1`](/packages/llm/llama_cpp) [`llama_cpp:0.3.2`](/packages/llm/llama_cpp) [`llama_cpp:0.3.5`](/packages/llm/llama_cpp) [`llama_cpp:0.3.6`](/packages/llm/llama_cpp) [`llama_cpp:0.3.7`](/packages/llm/llama_cpp) [`llama_cpp:0.3.8`](/packages/llm/llama_cpp) [`llama_cpp:0.3.9`](/packages/llm/llama_cpp) [`llama_cpp:0.4.0`](/packages/llm/llama_cpp) [`llama_cpp:b5255`](/packages/llm/llama_cpp) [`llama_cpp:b5833`](/packages/llm/llama_cpp) [`llava`](/packages/vlm/llava) [`lobechat`](/packages/llm/lobe_chat) [`local_llm`](/packages/llm/local_llm) [`log-linear-attention:0.0.1`](/packages/attention/log-linear-attention) [`mamba:2.2.5`](/packages/ml/mamba/mamba) [`mambavision:1.0`](/packages/ml/mamba/mambavision) [`minference:0.1.7`](/packages/llm/minference) [`minigpt4`](/packages/vlm/minigpt4) [`mlc:0.1.0`](/packages/llm/mlc) [`mlc:0.1.1`](/packages/llm/mlc) [`mlc:0.1.2`](/packages/llm/mlc) [`mlc:0.1.3`](/packages/llm/mlc) [`mlc:0.1.4`](/packages/llm/mlc) [`mlc:0.19.0`](/packages/llm/mlc) [`mlc:0.20.0`](/packages/llm/mlc) [`mlc:0.21.0`](/packages/llm/mlc) [`mlstm_kernels:2.0.1`](/packages/ml/xlstm/mlstm_kernels) [`mooncake:0.3.5`](/packages/llm/dynamo/mooncake) [`nano_llm:24.4`](/packages/llm/nano_llm) [`nano_llm:24.4-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4-humble`](/packages/llm/nano_llm) [`nano_llm:24.4-iron`](/packages/llm/nano_llm) [`nano_llm:24.4.1`](/packages/llm/nano_llm) [`nano_llm:24.4.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.4.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.4.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.4.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.5`](/packages/llm/nano_llm) [`nano_llm:24.5-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5-humble`](/packages/llm/nano_llm) [`nano_llm:24.5-iron`](/packages/llm/nano_llm) [`nano_llm:24.5.1`](/packages/llm/nano_llm) [`nano_llm:24.5.1-foxy`](/packages/llm/nano_llm) [`nano_llm:24.5.1-galactic`](/packages/llm/nano_llm) [`nano_llm:24.5.1-humble`](/packages/llm/nano_llm) [`nano_llm:24.5.1-iron`](/packages/llm/nano_llm) [`nano_llm:24.6`](/packages/llm/nano_llm) [`nano_llm:24.6-foxy`](/packages/llm/nano_llm) [`nano_llm:24.6-galactic`](/packages/llm/nano_llm) [`nano_llm:24.6-humble`](/packages/llm/nano_llm) [`nano_llm:24.6-iron`](/packages/llm/nano_llm) [`nano_llm:24.7`](/packages/llm/nano_llm) [`nano_llm:24.7-foxy`](/packages/llm/nano_llm) [`nano_llm:24.7-galactic`](/packages/llm/nano_llm) [`nano_llm:24.7-humble`](/packages/llm/nano_llm) [`nano_llm:24.7-iron`](/packages/llm/nano_llm) [`nano_llm:main`](/packages/llm/nano_llm) [`nano_llm:main-foxy`](/packages/llm/nano_llm) [`nano_llm:main-galactic`](/packages/llm/nano_llm) [`nano_llm:main-humble`](/packages/llm/nano_llm) [`nano_llm:main-iron`](/packages/llm/nano_llm) [`nanodb`](/packages/vectordb/nanodb) [`nanoowl`](/packages/vit/nanoowl) [`nanosam`](/packages/vit/nanosam) [`nemo`](/packages/llm/nemo) [`nerfstudio:1.1.7`](/packages/3d/nerf/nerfstudio) [`nixl:0.3.2`](/packages/llm/dynamo/nixl) [`nvdiffrast:0.3.4`](/packages/diffusion/cosmos/cosmos_diffusion_renderer/nvdiffrast) [`nvidia_modelopt:0.32.0`](/packages/llm/tensorrt_optimizer/nvidia-modelopt) [`octo`](/packages/physicalAI/vla/octo) [`onnxruntime_genai:0.8.5`](/packages/ml/onnxruntime_genai) [`openpi`](/packages/robots/openpi) [`openvla`](/packages/physicalAI/vla/openvla) [`openvla:mimicgen`](/packages/physicalAI/vla/openvla) [`optimum`](/packages/llm/optimum) [`paraattention:0.4.0`](/packages/attention/ParaAttention) [`partpacker:0.1.0`](/packages/3d/3dobjects/partpacker) [`plstm:0.1.0`](/packages/ml/xlstm/pltsm) [`prismatic`](/packages/vlm/prismatic) [`pytorch:2.1-all`](/packages/pytorch) [`pytorch:2.2-all`](/packages/pytorch) [`pytorch:2.3-all`](/packages/pytorch) [`pytorch:2.3.1-all`](/packages/pytorch) [`pytorch:2.4-all`](/packages/pytorch) [`pytorch:2.5-all`](/packages/pytorch) [`pytorch:2.6-all`](/packages/pytorch) [`pytorch:2.7-all`](/packages/pytorch) [`pytorch:2.8-all`](/packages/pytorch) [`radial-attention:0.1.0`](/packages/attention/radial-attention) [`robopoint`](/packages/physicalAI/vla/robopoint) [`sage-attention:3.0.0`](/packages/attention/sage-attention) [`sdnext`](/packages/diffusion/sdnext) [`self-forcing`](/packages/diffusion/self-forcing) [`sgl-kernel:0.2.3`](/packages/llm/sglang/sgl-kernel) [`sglang:0.4.4`](/packages/llm/sglang) [`sglang:0.4.6`](/packages/llm/sglang) [`sglang:0.4.9`](/packages/llm/sglang) [`sparc3d:0.1.0`](/packages/3d/3dobjects/sparc3d) [`sparge-attention:0.1.0`](/packages/attention/sparge-attention) [`spark-tts`](/packages/speech/spark-tts) [`speaches`](/packages/speech/speaches) [`stable-diffusion-webui`](/packages/diffusion/stable-diffusion-webui) [`sudonim:hf`](/packages/llm/sudonim) [`tensorrt_llm:0.12`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`tensorrt_llm:0.22.0`](/packages/llm/tensorrt_optimizer/tensorrt_llm) [`text-generation-inference`](/packages/llm/text-generation-inference) [`text-generation-webui:1.7`](/packages/llm/text-generation-webui) [`text-generation-webui:6a7cd01`](/packages/llm/text-generation-webui) [`text-generation-webui:main`](/packages/llm/text-generation-webui) [`transformer-engine:2.7`](/packages/ml/transformer-engine) [`transformers:4.53.1`](/packages/llm/transformers) [`videollama:1.0.0`](/packages/vlm/videollama) [`videomambasuite:1.0`](/packages/ml/mamba/videomambasuite) [`vila`](/packages/vlm/vila) [`vllm:0.7.4`](/packages/llm/vllm) [`vllm:0.8.4`](/packages/llm/vllm) [`vllm:0.9.0`](/packages/llm/vllm) [`vllm:0.9.2`](/packages/llm/vllm) [`vllm:0.9.3`](/packages/llm/vllm) [`vllm:v0.8.5.post1`](/packages/llm/vllm) [`voice-pro`](/packages/speech/voice-pro) [`voicecraft`](/packages/speech/voicecraft) [`vscode:transformers`](/packages/code/vscode) [`whisperx`](/packages/speech/whisperx) [`wyoming-whisper:2.5.0`](/packages/smart-home/wyoming/wyoming-whisper) [`wyoming-whisper:master`](/packages/smart-home/wyoming/wyoming-whisper) [`xattention:0.0.1`](/packages/attention/xattention) [`xformers:0.0.32`](/packages/attention/xformers) [`xformers:0.0.33`](/packages/attention/xformers) [`xgrammar:0.1.15`](/packages/llm/xgrammar) [`xgrammar:0.1.18`](/packages/llm/xgrammar) [`xgrammar:0.1.19`](/packages/llm/xgrammar) [`xgrammar:0.1.20`](/packages/llm/xgrammar) [`xgrammar:0.1.21`](/packages/llm/xgrammar) [`xlstm:2.0.5`](/packages/ml/xlstm/xlstm) [`xtts`](/packages/speech/xtts) [`xtuner`](/packages/vlm/xtuner) [`zigma:1.0`](/packages/ml/mamba/zigma) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Images | [`dustynv/huggingface_hub:r32.7.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) `(2023-12-15, 0.4GB)`
[`dustynv/huggingface_hub:r35.2.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) `(2023-12-15, 5.0GB)`
[`dustynv/huggingface_hub:r35.3.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) `(2023-09-07, 5.0GB)`
[`dustynv/huggingface_hub:r35.4.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) `(2023-10-07, 5.0GB)` | |    Notes | https://github.com/huggingface/huggingface_hub |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size | | :-- | :--: | :--: | :--: | |   [`dustynv/huggingface_hub:r32.7.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) | `2023-12-15` | `arm64` | `0.4GB` | |   [`dustynv/huggingface_hub:r35.2.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) | `2023-12-15` | `arm64` | `5.0GB` | |   [`dustynv/huggingface_hub:r35.3.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) | `2023-09-07` | `arm64` | `5.0GB` | |   [`dustynv/huggingface_hub:r35.4.1`](https://hub.docker.com/r/dustynv/huggingface_hub/tags) | `2023-10-07` | `arm64` | `5.0GB` | > 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`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag huggingface_hub) # or explicitly specify one of the container images above jetson-containers run dustynv/huggingface_hub:r35.2.1 # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host dustynv/huggingface_hub:r35.2.1 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag huggingface_hub) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag huggingface_hub) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build huggingface_hub ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- ### Overview (`packages/llm/hymba/README.md`) # hymba > [`CONTAINERS`](#user-content-containers) [`IMAGES`](#user-content-images) [`RUN`](#user-content-run) [`BUILD`](#user-content-build)
CONTAINERS
| **`hymba`** | | | :-- | :-- | |    Requires | `L4T ['>=34.1.0']` | |    Dependencies | [`build-essential`](/packages/build/build-essential) [`pip_cache:cu126`](/packages/cuda/cuda) [`cuda:12.6`](/packages/cuda/cuda) [`cudnn`](/packages/cuda/cudnn) [`python`](/packages/build/python) [`numpy`](/packages/numeric/numpy) [`cmake`](/packages/build/cmake/cmake_pip) [`onnx`](/packages/ml/onnx) [`pytorch:2.8`](/packages/pytorch) [`torchvision`](/packages/pytorch/torchvision) [`huggingface_hub`](/packages/llm/huggingface_hub) [`rust`](/packages/build/rust) [`transformers`](/packages/llm/transformers) [`ninja`](/packages/build/ninja) [`torchaudio`](/packages/pytorch/torchaudio) [`triton`](/packages/ml/triton) [`causalconv1d`](/packages/ml/mamba/causalconv1d) [`mamba`](/packages/ml/mamba/mamba) [`diffusers`](/packages/diffusion/diffusers) [`xformers`](/packages/attention/xformers) [`cuda-python`](/packages/cuda/cuda-python) [`cutlass`](/packages/cuda/cutlass) [`flash-attention`](/packages/attention/flash-attention) | |    Dockerfile | [`Dockerfile`](Dockerfile) | |    Notes | https://github.com/NVlabs/hymba |
RUN CONTAINER
To start the container, you can use [`jetson-containers run`](/docs/run.md) and [`autotag`](/docs/run.md#autotag), or manually put together a [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) command: ```bash # automatically pull or build a compatible container image jetson-containers run $(autotag hymba) # or if using 'docker run' (specify image and mounts/ect) sudo docker run --runtime nvidia -it --rm --network=host hymba:36.4.0 ``` > [`jetson-containers run`](/docs/run.md) forwards arguments to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/) with some defaults added (like `--runtime nvidia`, mounts a `/data` cache, and detects devices)
> [`autotag`](/docs/run.md#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`](https://docs.docker.com/engine/reference/commandline/run/#volume) or [`--volume`](https://docs.docker.com/engine/reference/commandline/run/#volume) flags: ```bash jetson-containers run -v /path/on/host:/path/in/container $(autotag hymba) ``` To launch the container running a command, as opposed to an interactive shell: ```bash jetson-containers run $(autotag hymba) my_app --abc xyz ``` You can pass any options to it that you would to [`docker run`](https://docs.docker.com/engine/reference/commandline/run/), and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use [`autotag`](/docs/run.md#autotag) as shown above, it'll ask to build the container for you if needed. To manually build it, first do the [system setup](/docs/setup.md), then run: ```bash jetson-containers build hymba ``` The dependencies from above will be built into the container, and it'll be tested during. Run it with [`--help`](/jetson_containers/build.py) for build options.
--- --- METRICS --- - Files Extracted: 57 - Estimated Token Budget: ~125617 tokens - Canonical Reference: https://codewiki.google/github.com/dusty-nv/jetson-containers