Can't get procs/transform to work on TabularPandas or TabularDataLoaders

fastai
1

Hello,

I am trying to experiment with fastai on Tabular datas, and i want to normalize/scale some columns but i can’t get the transforms specified in procs to work on my dataset.
I tried reproducing the tutorial from the docs on Kaggle, but the transforms aren’t applied to the data.
See here if you want to reproduce at home:

from fastai.tabular.all import *

path = untar_data(URLs.ADULT_SAMPLE)
path.ls()
df = pd.read_csv(path/'adult.csv')
splits = RandomSplitter(valid_pct=0.2)(range_of(df))

to = TabularPandas(df, procs=[Categorify, FillMissing,Normalize],
                   cat_names = ['workclass', 'education', 'marital-status', 'occupation', 'relationship', 'race'],
                   cont_names = ['age', 'fnlwgt', 'education-num'],
                   y_names='salary',
                   splits=splits)
dls = to.dataloaders(bs=64)
dls.show_batch()

dls = TabularDataLoaders.from_csv(path/'adult.csv', path=path, y_names="salary",
    cat_names = ['workclass', 'education', 'marital-status', 'occupation', 'relationship', 'race'],
    cont_names = ['age', 'fnlwgt', 'education-num'],
    procs = [Categorify, FillMissing, Normalize])
dls.show_batch()
workclass education marital-status occupation relationship race education-num_na age fnlwgt education-num salary
0 Private HS-grad Divorced Priv-house-serv Unmarried Black False 62.000000 272209.000628 9.0 <50k
1 Local-gov HS-grad Married-civ-spouse Transport-moving Husband White False 49.000000 67001.004943 9.0 <50k
2 Private Bachelors Divorced Prof-specialty Unmarried White False 35.000000 225749.998765 13.0 <50k
3 Local-gov HS-grad Divorced Adm-clerical Unmarried White False 37.000000 272471.002093 9.0 <50k
4 ? HS-grad Divorced ? Not-in-family White False 46.000000 37671.997994 9.0 <50k
5 State-gov HS-grad Never-married Sales Unmarried Black False 30.000000 312767.002367 9.0 <50k
6 Private HS-grad Divorced Other-service Not-in-family White False 63.000001 174826.000719 9.0 <50k
7 Federal-gov Some-college Married-civ-spouse Adm-clerical Husband White False 61.000000 229062.000296 10.0 >=50k
8 State-gov Assoc-voc Divorced Tech-support Unmarried White False 45.000000 37671.997994 11.0 <50k
9 Private HS-grad Married-civ-spouse Adm-clerical Wife White False 20.000000 221661.000241 9.0 <50k
workclass education marital-status occupation relationship race education-num_na age fnlwgt education-num salary
0 Private Some-college Married-civ-spouse Adm-clerical Wife White False 36.000000 461336.999879 10.0 <50k
1 Federal-gov Some-college Married-civ-spouse Tech-support Husband White False 29.000000 309777.997350 10.0 >=50k
2 Self-emp-not-inc Some-college Divorced Craft-repair Not-in-family White False 58.000001 195835.000087 10.0 <50k
3 Private HS-grad Never-married Craft-repair Not-in-family White False 39.000000 248693.996852 9.0 <50k
4 Private Doctorate Married-civ-spouse #na# Husband White False 64.000000 47298.001898 16.0 >=50k
5 Local-gov 11th Married-civ-spouse Craft-repair Husband Black False 64.000000 244903.001857 7.0 >=50k
6 Private Some-college Divorced Exec-managerial Not-in-family White False 35.000000 187119.000058 10.0 <50k
7 Private Assoc-acdm Never-married Other-service Own-child White False 25.000000 82559.995684 12.0 <50k
8 Private Assoc-acdm Divorced Prof-specialty Unmarried White False 42.000000 341204.005738 12.0 >=50k
9 ? HS-grad Never-married ? Not-in-family White False 40.000000 84232.001317 9.0 <50k
2

Alright, i’ve still finished my pipeline and trained my model, it seems the procs are happening during the training process, as i see them with learn.show_results(), but don’t show when i call show_batch() on my dataloaders.