About test_dl

does anyone know how to access test_dl data? why the n_subsets is still 2 after I added test_imgs using the dls.test_dl(test_imgs, with_label=True)?

if I have a vision learner, and I want to do prediction with probabilities using the test_dl instead of validation set or other image, what should I do? learner.predict(img) works. But since I have already put the testing images into dls.test_dl, I would want to try to predict using the images in the test_dl (but it failed because test_dl is type of TfmdDL).

You should be able to get the predictions from the test_dl by doing learn.get_preds(dl=test_dl)

1 Like