Has anyone had experience where the Tabular Learner outputs the same values for all predictions

Deep Learning
1

For regression tasks, I’m finding that my learner is giving the same output for all predictions. I suspected that it was an issue with the data, but putting it through a random forest gives me different predictions. I’ve tried changing the model architecture, but nothing so far has worked

Has anyone experienced this issue?

2

Hard to tell without example code and data.

3

Yes. And I am still stuck.
learner.get_preds(ds_type=DatasetType.Test) returns all zeros in the prediction column.

4

Found the fix.

Just do this :

predictions, *_ = learner.get_preds(DatasetType.Test)
labels = np.argmax(predictions, 1)

I hope I’m not too late, just found it out myself.

1 Like