Install

It is recommended to read the Packages section first to understand the packages organisation.

The CORE package DeepPhysX.Core is always installed by default.

Using a SIMULATION package is recommended but not required.

At least one AI package is mandatory (default one is DeepPhysX.Torch).

Prerequisites

DeepPhysX packages have the following dependencies :

Package	Dependency	Type	Install
DeepPhysX_Core	Numpy	Required	pip install numpy
	Vedo	Required	pip install vedo
	Tensorboard	Required	pip install tensorboard
	SSD	Required	pip install SimulationSimpleDatabase
DeepPhysX_Sofa	SOFA Framework	Required	Follow instructions
	SofaPython3	Required	Follow instructions
	Caribou	Optional	Follow instructions
DeepPhysX_Torch	PyTorch	Required	pip install torch

Note

DeepPhysX.Sofa has a dependency to Caribou to run the demo scripts from Examples/SOFA/Demo since implemented simulations involve some of its components.

Install

Install with pip

DeepPhysX packages are all registered on PyPi. They can easily be installed with pip:

$ pip3 install DeepPhysX
$ pip3 install DeepPhysX.Sofa
$ pip3 install DeepPhysX.Torch

Then, you should be able to run:

$ pip3 show DeepPhysX
$ pip3 show DeepPhysX.Sofa
$ pip3 show DeepPhysX.Torch

import DeepPhysX.Core
import DeepPhysX.Sofa
import DeepPhysX.Torch

Install from sources

DeepPhysX packages must have the following architecture. These detailed instructions will build such an installation:

DeepPhysX
 ├── Core
 ├── Sofa
 └── Torch

Start by cloning the DeepPhysX source code from its Github repository in a directory named DeepPhysX/Core:

$ mkdir DeepPhysX
$ cd DeepPhysX
$ git clone https://github.com/mimesis-inria/DeepPhysX.git Core

Then, you can add compatibility layers to your DeepPhysX environment and install packages:

• Option 1 (recommended): run one of the setup_.py scripts that handle the installation of all packages

  • Use setup_user.py to install and manage packages with pip as non-editable.

    $ python3 setup_user.py
    

  • Use setup_dev.py to link packages in the site-packages.

    $ python3 setup_dev.py set
    

  Note

  Both scripts will asks the packages to install and will automatically clone missing packages.

• Option 2: clone the corresponding Github repositories in the created DeepPhysX directory, then install packages manually.

  # Clone compatibility layers
  $ git clone https://github.com/mimesis-inria/DeepPhysX.Sofa.git Sofa
  $ git clone https://github.com/mimesis-inria/DeepPhysX.Torch.git Torch
  
  # Install packages manually
  $ cd Core ; pip3 install .
  $ cd ../Sofa ; pip3 install .
  $ cd ../Torch ; pip3 install .
  

Finally, you should be able to run:

# If installed with pip
$ pip3 show DeepPhysX
$ pip3 show DeepPhysX.Sofa
$ pip3 show DeepPhysX.Torch

# In both cases
import DeepPhysX.Core
import DeepPhysX.Sofa
import DeepPhysX.Torch