My data is normalized before fit:
db = (tabular.TabularList.from_df(data, path='.', cat_names=cats, cont_names=cons, procs=[Categorify, Normalize])
.split_by_rand_pct(valid_pct = 0.1, seed = 42)
.label_from_df(cols=target)
.databunch())
learn = tabular_learner(data, layers=[2000,500,50], metrics=rmse)
learn.lr_find()
learn.recorder.plot()
learn.fit_one_cycle(15, max_lr =1e-01)
When I do predict, shall I just simply call predict
or shall I do the normalization first? If the latter, how? Especially, where to get the mean/stdev of the training data?