Installation

Vortex consists of 2 packages:

• Vortex development package (vortex.development), which contains all the necessary components to develop a model.

• Vortex runtime package (vortex.runtime), which contains only minimal dependencies to run a model for inference with the option for: * all * torchscript * onnxruntime * onnxruntime-tensorrt

Currently vortex is only work using python3.6, but support for other versions are to come.

There are two ways to use vortex, using docker or install from source. But we generallly recommend to use the provided docker images.

Docker (recommended)

We have provided the pre-built docker images in our official docker hub repo:

https://hub.docker.com/r/nodefluxio/vortex

For the vortex development the tag name is <version>, for vortex runtime the tag is runtime-<runtime_type>-<version>.

If you want to build the images yourself, run these command in the root directory of the vortex source:

• vortex development

  docker build --target=development -t vortex:dev .
  

• vortex runtime

  docker build --target=runtime --build-arg RUNTIME_TYPE=<runtime_type> -t vortex:runtime .
  

  you need to change the <runtime_type> option. The available options are: all, torchscript, and onnxruntime. So if you want to only build onnxruntime image:

  docker build --target=runtime --build-arg RUNTIME_TYPE=onnxruntime -t vortex:runtime .
  

  if you want to build the onnxruntime TensorRT, you can follow the steps in https://github.com/nodefluxio/vortex/blob/master/dockerfiles/README.md

From Source

Vortex currently only tested in python 3.6, so if you want to build it from source make sure to build it for python3.6. You also need to make sure that you have pip installed for the python3.6.

Before installing vortex, you need to first install the opencv dependencies:

sudo apt-get update
sudo apt-get install libsm6 libxrender-dev

Then, clone the vortex repository.

• vortex development

  It should be noted that vortex.development package also depends on vortex.runtime package, so you need to install both packages:

  pip install ./src/runtime[all] ./src/development
  

  Or if you want to install with additional optuna visualization support:

  pip install 'src/development[optuna_vis]'
  

• vortex runtime

  To install, you can simply

  pip install ./src/runtime[all]
  

  you can also change the option all to the desired runtime type. The available options are all, torchscript, and onnxruntime.

It should be noted that for TensorRT runtime type, you need to build the onnxruntime from the source. If you want to build vortex from source with TensorRT runtime support, you can follow the steps in dockerfile here.