Setting up Anaconda and Houdini for 3D Deep Learning on Linux

Last updated Jun 4, 2022 Edit Source

• Linux
• Houdini
• Anaconda
• Machinelearning
• Pytorch
• Pytorch3d

To get started install Anaconda and run the following commands in your terminal:

( hou is the name of the new enviroment and python 3.7.4 is required because this is the version used in Houdini )

1

conda create -n hou python=3.7.4

1

conda activate hou

Then install libraries like for example PyTorch 3D. ( For simplicities sake we will go forward with vanilla PyTorch. )

1

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

As jpparkeramnh pointed out in this SideFX forum post you have to export the LD_PRELOAD variable.

1

export LD_PRELOAD=$CONDA_PREFIX/lib/libpython3.7m.so

I also recommend sourcing Houdini in the terminal.

To do so first open the .bashrc in the terminal

1

nano .bashrc

and add the following lines to your .bashrc

1
2
3
4
5
6

cd /opt/hfs19.0/
source ./houdini_setup
cd ~
conda activate hou
export LD_PRELOAD=$CONDA_PREFIX/lib/libpython3.7m.so
alias expenv='LD_PRELOAD=$CONDA_PREFIX/lib/libpython3.7m.so'

• The first 3 lines run the houdini setup bash script
• Line 4 activates a default environment ( for me “hou” )
• Line 5 exports the LD_PRELOAD varible of the currently activated environment ( hou )
• Line 6 creates an alias to export the currently activated environment from the terminal

Now you should be able to just activate your respecitve conda environment, export the correct LD_PRELOAD variable by typing “expenv” and then run houdini from there by typing “houdini“.

To check if everything is running as expected open the Houdini Python shell and type:

1

import torch 

If it doesn’t give you any errors you should be good to go.