Collects data, trains, serves and distributes your Tensorflow models via REST API.
STATUS alpha; implementing swagger schema : https://app.swaggerhub.com/apis/kyr7/hivemind/0.0.1
- tested on ubuntu 18. CPU training should also run on Windows and Mac OS.
- install docker https://www.docker.com/
- install nvidia docker runtime https://github.com/NVIDIA/nvidia-docker
- recommended to
docker pull
:- tensorflow/serving:latest
- kyr7/emg-nn:nvidia
OR
kyr7/emg-nn:cpu if you set up CPU training - if you don't pull these images manually, they will have to be pulled before learning automatically, which is rather slow a.t.m.
- choose your profile from list
- if you want training model other than cpu-based then set system property
profile
or env variableHIVEMIND_PROFILE
with desired one - docker client must be set up to work without
sudo
https://docs.docker.com/install/linux/linux-postinstall/ - see examples folder for API reference.
- execute
./gradlew run
to start