Hello,
I am trying to run on more datasets from GNS.
- Did you run on SandRamps or WaterRamps which have ramps? How did you handle that and enforce the static local boundary conditions?
- Do you have the saved weights for more of these models? Mainly,
Sand2D_T and MultiMaterial_T, for which there are configs in the code, but no weights are provided. If you have weights for other models in case you trained on the other GNS datasets, it would be appreciated. Specifically, the weights for Goop, for which you obtain such a low MSE along with its config would be nice.
Also, what is average training time (such as for Sand2D or MultiMaterial) that it required? You specify 5 million update steps on an RTX 3060, but you do not specify how long these are. For 5 million steps, the runs I've started so far for GIOROM are planned to take about a week. Is this correct? If yes, then saved weights would be even more appreciated.
Thanks in advance,
Sacha Lewin
Hello,
I am trying to run on more datasets from GNS.
Sand2D_TandMultiMaterial_T, for which there are configs in the code, but no weights are provided. If you have weights for other models in case you trained on the other GNS datasets, it would be appreciated. Specifically, the weights for Goop, for which you obtain such a low MSE along with its config would be nice.Also, what is average training time (such as for Sand2D or MultiMaterial) that it required? You specify 5 million update steps on an RTX 3060, but you do not specify how long these are. For 5 million steps, the runs I've started so far for GIOROM are planned to take about a week. Is this correct? If yes, then saved weights would be even more appreciated.
Thanks in advance,
Sacha Lewin