What’s going on is that fastai is treating your dependent variable as different classes, and you probably have 15 unique values of sales price so you are getting a single class for each of the values. For predictions, fastai is giving you probabilities for each of the 15 classes, hence why the size is 1459x15 and the values are small.
To fix this, pass inlabel_cls=FloatList to label_from_df to tell fastai to treat the dependent variable as a list of floats rather than a list of class labels.