Can't predict on Tabular or Collab Learners

[HenrySellers]

I am using fastai v2.0.15 and no matter what I pass into predict I get

/opt/conda/lib/python3.6/site-packages/fastai/learner.py in predict(self, item, rm_type_tfms, with_input)
    245 
    246     def predict(self, item, rm_type_tfms=None, with_input=False):
--> 247         dl = self.dls.test_dl([item], rm_type_tfms=rm_type_tfms, num_workers=0)
    248         inp,preds,_,dec_preds = self.get_preds(dl=dl, with_input=True, with_decoded=True)
    249         i = getattr(self.dls, 'n_inp', -1) ...

... TypeError: list indices must be integers or slices, not list

I’ve tried everything from the forum & other sources & also tried even passing in None types or anything at all. Same result. I’m wondering is it something to do with my environment or something else?

Any help appreciated.

[florianl (Florian)]

Can you post your Code please?

[HenrySellers]

Hi Florian,

Sure:

It doesn’t seem to matter what I pass into the predict function as the error is always the same and points to the line in my OP

[dinodelao]

Hi, I am having the same problem as well. Did you ever find a solution? I’m trying to use my collab_learner model to make predictions on a test set.

[JT_0024]

Same issue. I load my exported collab learner:

And I create a single DataLoader object from the same DataLoaders object that was used to train the model, even trying to replicate the source code for Learner.get_preds():

But I still get an error:

Please help, I need to be able to make predictions with my collab model.

[JT_0024]

Alright, I figured it out, at least for my use case. The new DataLoader needs to be created by passing the new DataFrame into learn.dls.test_dl(). Then you can call learn.get_preds() with the new DataLoader:

[maddy2u (Madhavan Sriram)]

This worked for me. Thanks. The documentation needs to be updated for this one.