search

mariusmerkle / TL-PINNs

Boosting the training of physics informed neural networks with transfer learning
25 stars 3 forks source link

Load data issue #4

Open iqbalkaleem opened 3 years ago

iqbalkaleem commented 3 years ago

Hi, marius hope you are fine. Very nice work done by you and your group. When I try to run your code the following error occurs "epochs_source = np.loadtxt("E:\kaleem\TL-PINNs-main\TL-PINNs-main\3 - Kovasznay Flow\Source/Neural_Networks", dtype = 'int32', delimiter = ',')" not found. This is my code downloads directory. Your actual directory or statement is "epochs_source = np.loadtxt("Neural_Networks/epochs.csv", dtype = 'int32', delimiter = ',')" but shows the same error. So kindly guide me how to adjust or remove this error.

mariusmerkle commented 3 years ago

Hi iqbalkaleem,

thank you very much. Indeed, PINNs are a very exciting research field.

In epochs.csv, the number of epochs for all neural networks are stored. This line serves to reload the saved numbers because DeepXDE stores a neural network with file name that contains the number of epochs.

You have two options to solve this issue:

  1. Save epochs.csv in the correct directory. From the python file where you run the code, it should be in Neural_Networks/epochs.csv.
  2. Open the file epochs.csv and transfer the number of epochs manually into the python file. You would store them in an array, similar to epochs_source = [10001, 15032, etc.]. Of course, the numbers need to correspond to the ones contained in the file epochs.csv.