Installation

pip install tensorflow==2.8.3 -f https://diyor28.github.io/wheels

Supported versions:

Python: 3.7, 3.8, 3.9, 3.10, 3.11
Tensorflow: 2.7, 2.8, 2.9, 2.10, 2.11, 2.12
Tensorflow data validation: 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 1.10, 1.12, 1.13

NOTE: Other versions can be built and uploaded upon request.

You can also use prebuilt images from dockerhub

FROM diyor28/tensorflow:2.7.3-py37
RUN python -c "import tensorflow; print(tensorflow.__version__)"

All the images are based on official python docker images.

TFDV images are coming soon

Running webserver using docker

Make htpasswd for basic auth

$ mkdir basic_auth
$ htpasswd -c basic_auth/.htpasswd admin

Make directory for workbench

We use /tmp/tf_aarch64/ for workbench for this build tool.

$ mkdir /tmp/tf_aarch64

Configuration

If you would like to disable caching or do more than one build at a time

vim .env

# do not set this higher than 2 if you have less then 32GB of RAM. Espeically for building tensorflow
BUILDER_THREADS=1
# disable if you are using buildkit or do not want caching in your builds
USE_CACHE=True

Now in the root of the project run:

docker compose up -d --build

Now visit http://localhost you should see the web interface

In the web UI there are two tabs: Tensorflow and TFX. Each page has a version dropdown at the bottom of the page for selecting python version, the library version. A version number that looks like 2.7.x means that it will build the latest changes in the 2.7 branch.

Development setup

Running python server

cd back/
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

Running ui server

cd front/
npm i
npm run serve

Adding and building custom versions

Coming soon...

Building wheels manually

Generate the dockerfile with build instructions (tensorflow in this case)

python gen.py tensorflow -v 2.7.3 -py 3.7 ./Dockerfile_tf

Now build the actual image

docker build -t tensorflow:2.7.3-py3.7 -f ./Dockerfile_tf ./build_templates/context/

Copy wheels from the resulting image to host machine using:

docker run -v /host/machine/path:/builds tensorflow:2.7.3-py3.7 cp -a /wheels/. /builds