Stack the features of multiple models (Ensemble)

Hello everyone,
I am a newbie of Deep learning and fastai too.
I have been working with Keras before and now I found the fastai is awesome.

I have been known a method called as model stacking that is one of ensembled technical.
Please correct me if I am wrong.
By model stacking, we train multiple models and get their “features”. The features in here, we defined as an output of convolution block before pass through fully connected layers.
Let say (n_samples, n_features) is the output of convolution blocks
Where:
n_samples is the size of the training set
n_features is the number of features.

The stacked features of multiple models should be:
stacked_features = (n_samples, n_features_1 + n_features_2 +…+ n_features_m)
where m is the number of models you want to stack together.

Then, the stacked_features will be an input of a fully connected model we would like to continue training.

Above is the step I have been done with Keras. So, I am wondering if there is a similar way to do with fastai.

As my investigation,
act, val_act, test_act = learn.activations

  1. Would act, val_act and test_act be the features that I mentioned above?
  2. If (1) is correct, they are carray struct. How can I stack the features together?

Thank,