I have a multilabel problem. My data is store in the df and two of the columns contain information about the categories that I want to predict (like in the planet dataset). I tried many ways and there was always a problem. Maybe someone knows what should I enter into “label_from_df (???)” ?
(all my variables are continuous)
procs = [FillMissing, Normalize]
valid_idx=np.random.randint(0,len(data),int(len(data)*0.15))
data_to_model = (TabularList.from_df (data, path = ‘.’, procs = procs)
.split_by_idx (valid_idx = valid_idx)
.label_from_df (???)
.databunch ())
Grzegorz,