zabaras / transformer-physx Goto Github PK

Hi, I got the error where these functions can't be found.
from trphysx.data_utils import AutoDataset, AutoPredictionDataset
from trphysx.data_utils.data_utils import EvalDataCollator
from trphysx.data_utils.dataset_lorenz import LorenzPredictDataset

is there any updated code?
Please advice
thanks

Hi, i have an issue with the input of ConvLstms yu presented it in the benchamrk for the paper. Wht is th input is it a mask or initial conditions. Thanks a lot.

Maybe mask0 should be a tensor not a bool value. So we need to delete "is True".

Hello,

Thank you very much for this amazing work.
I am trying to understand a bit better differentiable physics for fluid dynamics and this would help me a lot.

However when I am trying to run your code there is an error on cell 14 that I was unable to solve. Maybe due to an update of HfArgumentParser? Everything ran well on the previous cells (except for a small typo on cell 5 where there were some spaces after the '%' sign but that's straight forward to fix)

Do you have an idea on how to fix it and run your code?

Thank you very much

Dear authors:

Thanks for your interesting and inspirational work! But I think there may be a minor mistake in the provided codes. In the file embedding_lorenz.py, the following codes are included

        # Test accuracy of one time-step
        for i in range(xInput.size(1)):
            xInput0 = xInput[:,i].to(device)
            g0 = self.embedding_model.embed(xInput0)
            yPred0 = self.embedding_model.recover(g0)
            yPred[:,i] = yPred0.squeeze().detach()

        test_loss = mseLoss(yTarget, yPred)

I think it should be

        # Test accuracy of one time-step
        for i in range(xInput.size(1)):
            xInput0 = xInput[:,i].to(device)
            g0 = self.embedding_model.embed(xInput0)
            g0 = self.embedding_model.koopmanOperation(g0)
            yPred0 = self.embedding_model.recover(g0)
            yPred[:,i] = yPred0.squeeze().detach()

        test_loss = mseLoss(yTarget, yPred)

for its practical effect.

Hi @NickGeneva very nice work!

do you plan to release the embedding modules (from ..embedding.embedding_model import EmbeddingModel) and one example of test?

thanks!

Hello!

Thank you for the interesting paper.

Could you please share the data you used for the Lorenz benchmark?

Thank you

Hello!

Thanks for your exciting work!

I'm confused about the result of the Lorenz system. The paper shows that the Lorenz system's prediction results are perfect, but I can't reproduce it based on the currently provided code. Moreover, the run results provided in Colab are also not great enough (states error stay 18). I really want to know what caused this.

Thank you very much!

What should be the input to the embedding and transformer colab? How are the HDF5 files structured? What is the meaning of the following naming in the hdf5 files ['0', '1', '10', '100', '1000', '1001', '1002', '1003', '1004', '1005', '1006', '1007', '1008', '1009', '101', '1010'... ?