CODEWITHSUNDEEP
deep-learningfast-ai
Rajan Lagah
Rajan Lagah

Reputation: 2528

How to get predictions on dataframe in fastai

I have defined databunch as

data = (TabularList.from_df(train_df, path='./', cont_names=cont_names, procs=procs)
                           .split_by_idx(list(range(500,3000)))
                           .label_from_df(cols=dep_var)
                           .add_test(test, label=0)
                           .databunch())

After training i am

p = learn3.get_preds()
len(p[1])

o/p is 2500 which is correct as

                       .split_by_idx(list(range(500,3000)))

this is 2500 numbers


But now i have other data of around few thousand inputs and if i do 
for index in range(len(test_df)):
    predictions = learn3.predict(test_df.iloc[index])
    predictions = predictions[1].tolist()
    print(index)

It will take really huge amount of time. I want to pass test_df and get the predictions as was as

p = learn3.get_preds()

How can i do that ?

                       .split_by_idx(list(range(500,3000)))

Do we have way to replace this and pass data from other file ?

Upvotes: 5

Views: 2493

Answers (2)

User2321
User2321

Reputation: 3062

I have struggled with this for a while. The best method I have found is to do the following:

test_dl = learn.dls.test_dl(test_data) # Create a test dataloader
learn.get_preds(dl = test_dl) # Make predictions on it

Upvotes: 4

LucyDrops
LucyDrops

Reputation: 609

you can use fastai learner's add_test() method to replace the learner's original test set with your data. Alternatively you can also use learn.data.test_dl = new_df to achieve the same and then use get_preds() to get the predictions

Upvotes: 1

Related Questions