- clone this repo
- get python + Tensorflow(>=2.1)
- clone mantflow
- add the content of NeuralTurbulence/plugin into mantaflow/source/plugin and to the CMakeLists.txt
- build manta
To train a network, just update the parameters and architecture in manta_trainGenerator.py, then run
manta manta_trainGenerator.py
If you want to evaluate a network or get validation errors during training, you first need to generate a suitable dataset with the option writeData=True. Afterwards,
python runmodel.py <models> <params>
provides a varity of options to compute different error measures and visualize the outputs.