search

sparks-baird / CrabNet

Predict materials properties using only the composition information!
https://crabnet.readthedocs.io/
MIT License
12 stars 4 forks source link

Problem about processing the data #71

Open yqdleiyi opened 1 year ago

yqdleiyi commented 1 year ago

I tried to fit the CrabNet model with a non-example dataset, which is saved as .csv file. Then, when I tried to fit it, I met an unexpected error. What's the problem behind it?

image

sgbaird commented 1 year ago

Hi, can you provide a minimal working example? My guess is that there's something non-standard in how one (or multiple) of your formulas is represented. Try this with a subset of formulas and provide the results and which formulas were used in the subset.

For example, I think you could use the following dummy data (taken from mat-discover docs):

train_df = pd.DataFrame(dict(formula=["Tc1V1", "Cu1Dy1", "Cd3N2"], target=[248.539, 66.8444, 91.5034]))
val_df = pd.DataFrame(dict(formula=["Al2O3", "SiO2"]))