The error occurs asy_train contains non-numeric data, which is not compatible with torch.tensor.

Check the contents of y_train and ensure that the is only numeric values so you can safely convert to a tensor.

Also, based on the error it seems the incorrect type in y_train is a numpy object.