Tabular data prediction

I’m attempting to obtain the tabular learner instance for a saved model for prediction without having to load the entire training data. While i find there is a easy way in image classification to do so
" data2 = ImageDataBunch.single_from_classes(path, classes, tfms=get_transforms(), size=224).normalize(imagenet_stats)"
I couldn’t find a corresponding mechanism for tabular data. Any help would be highly appreciated. Thanks

Found the answer. The docs are pretty helpful. DataBunch has helpful api for doing this


I have couple of questions around tabular data pre-processing. My problem is similar to rossmann (continuous variable (spend) prediction). But I am not taking Log of the target variable. My question is - do I have to still normalize my data. Because when I do so my Lag variable (spend last week) is also normalized and my results also suffer.

Also if I chose to take log and also Normalize then can I pick R2-Square or MSE as my metrics=r2_score. If I do this I get very good R2 Square/MSE but when I do exp(target variable) and calculate the MPAE or MSE ; results look terrible.

Please advice…Thanks a ton