Container for TF1/TF2 with CUDA support.
Note that the l4t-tensorflow containers are similar, with the addition of OpenCV and PyCUDA.
The TensorFlow wheels used in these are from https://docs.nvidia.com/deeplearning/frameworks/install-tf-jetson-platform
CONTAINERS
tensorflow |
|
|---|---|
| Builds |   |
| Requires | L4T >=32.6 |
| Dependencies | build-essential cuda cudnn python tensorrt numpy protobuf:cpp |
| Dependants | l4t-tensorflow:tf1 |
| Dockerfile | Dockerfile |
| Images | dustynv/tensorflow:r32.7.1 (2023-12-06, 0.8GB)dustynv/tensorflow:r35.2.1 (2023-12-05, 5.5GB)dustynv/tensorflow:r35.3.1 (2023-08-29, 5.5GB)dustynv/tensorflow:r35.4.1 (2023-12-06, 5.5GB) |
tensorflow2 |
|
|---|---|
| Builds |    |
| Requires | L4T >=32.6 |
| Dependencies | build-essential cuda cudnn python tensorrt numpy protobuf:cpp |
| Dependants | l4t-ml l4t-tensorflow:tf2 |
| Dockerfile | Dockerfile |
| Images | dustynv/tensorflow2:r32.7.1 (2023-12-06, 0.9GB)dustynv/tensorflow2:r35.2.1 (2023-12-06, 5.6GB)dustynv/tensorflow2:r35.3.1 (2023-12-05, 5.6GB)dustynv/tensorflow2:r35.4.1 (2023-10-07, 5.6GB)dustynv/tensorflow2:r36.2.0 (2023-12-05, 7.2GB) |
CONTAINER IMAGES
| Repository/Tag | Date | Arch | Size |
|---|---|---|---|
dustynv/tensorflow:r32.7.1 |
2023-12-06 |
arm64 |
0.8GB |
dustynv/tensorflow:r35.2.1 |
2023-12-05 |
arm64 |
5.5GB |
dustynv/tensorflow:r35.3.1 |
2023-08-29 |
arm64 |
5.5GB |
dustynv/tensorflow:r35.4.1 |
2023-12-06 |
arm64 |
5.5GB |
Container images are compatible with other minor versions of JetPack/L4T:
• L4T R32.7 containers can run on other versions of L4T R32.7 (JetPack 4.6+)
• L4T R35.x containers can run on other versions of L4T R35.x (JetPack 5.1+)
RUN CONTAINER
To start the container, you can use the run.sh/autotag helpers or manually put together a docker run command:
# automatically pull or build a compatible container image
./run.sh $(./autotag tensorflow)
# or explicitly specify one of the container images above
./run.sh dustynv/tensorflow:r32.7.1
# or if using 'docker run' (specify image and mounts/ect)
sudo docker run --runtime nvidia -it --rm --network=host dustynv/tensorflow:r32.7.1
run.shforwards arguments todocker runwith some defaults added (like--runtime nvidia, mounts a/datacache, and detects devices)
autotagfinds a container image that's compatible with your version of JetPack/L4T - either locally, pulled from a registry, or by building it.
To mount your own directories into the container, use the -v or --volume flags:
./run.sh -v /path/on/host:/path/in/container $(./autotag tensorflow)To launch the container running a command, as opposed to an interactive shell:
./run.sh $(./autotag tensorflow) my_app --abc xyzYou can pass any options to run.sh that you would to docker run, and it'll print out the full command that it constructs before executing it.
BUILD CONTAINER
If you use autotag as shown above, it'll ask to build the container for you if needed. To manually build it, first do the system setup, then run:
./build.sh tensorflowThe dependencies from above will be built into the container, and it'll be tested during. See ./build.sh --help for build options.