Hello
I have created a text classifier learner and would like to use it for inference now. The model was trained with databunch
data_clas = TextClasDataBunch.from_df(path='.', train_df=train_df, valid_df=valid_df, vocab=data_lm.train_ds.vocab, bs=32, text_cols=['col1', 'col2', 'col3'], label_cols='label', mark_fields=True)
so it takes text from three separate columns and also marks fields.
How can I apply the same preprocessing step to the input columns for inference as in the training procedure?
Another question is whether the learn.predict
should be given as argument a list with three non preprocessed items or one preprocessed string only? I think these details are missing in the documentation and it would be useful to clarify these.
Thank you in advance.