Deep Lagrangian Fluids

Back to Index

Source code: GitHub

This network was built for a POSIX system, that is, a Mac or a Linux. There are three dependencies confirming that. Two of them can be removed (the resource python package, and the signal.SIGUSR1), because they only measure performance. But the third dependency is of Open3D-ML with Tensorflow or PyTorch, which is disabled for Windows due to several complaints of the users. If you try to compile it from the source on Windows, you'll get a warning and the compilation will stop. So, don't try to use this net on Windows, until this issue is solved. 

Also, respect the versions needed, that are Open3D-ML 0.11, which requires PyTorch 1.6.0 or Tensorflow 2.3.0, both with CUDA 10.1, SPlisH SPlasH 2.4.0, and OpenVDB with Python 3.7.

But not everything is lost. We have WSL. Follow me:

Index

• Virtualization
• Nvidia Drivers and CUDA for Windows
• Windows Insider Program and WSL - Windows Subsystem Linux
• Setting up WSL
• CUDA on WSL
• CMake 
• Miniconda
• Deep Learning libs (choose one)
• PyTorch
• Tensorflow
• Open3D-ML 
• SPlisH SPlasH
• Partio
• Deep Lagrangian Fluids
• OpenVDB
• Visualizing the results on Blender
• Training the Network in WSL