Count distinct on an aggregated dimension

I have the following model:

class Historique(models.Model): type_of_object_choices = ( (1, 'Cadeau') , (2, 'Offre') , (3, 'Commentaire') ) event_type_choices = ( (1, 'modification') , (2, 'creation') , (4, 'suppression') ) type_of_object = models.IntegerField(choices=type_of_object_choices, null=True) event_type = models.IntegerField(choices=event_type_choices) object_pk = models.IntegerField() 

Each object can be subject to three actions, each action occurring only once per object. These actions are encoded as an integer field called event_type_choices, which can take the values 1, 2, 4. By summing all event_type_choices for a given object, we compute a value in the range [1, 2, 4, 3, 5, 6, 7] representing the cumulative actions that have occurred on this object so far. Based on this value, a derived field called type_of_action is created to represent a simplified view of the actions performed on the object, whish is shown to the user.

I want to count the number of objects there are for each combinaison of type_of_object and type_of_action. The following SQL query achieves this, but I would like to express it in a Django's query syntax.

 SELECT type_object , CASE WHEN sum_event_type in (1) THEN 'modification' WHEN sum_event_type in (2, 3) THEN 'creation' WHEN sum_event_type in (4, 6) THEN 'supression' END as type_of_action , COUNT(DISTINCT object_pk) as nb_objects FROM ( SELECT type_object , object_pk , sum(event_type) as sum_event_type FROM historique GROUP BY type_object, object_pk ) as temp GROUP BY type_object , CASE WHEN sum_event_type in (1) THEN 'modification' WHEN sum_event_type in (2, 3) THEN 'creation' WHEN sum_event_type in (4, 6) THEN 'supression' END 

I have been trying different solutions based on annotate and aggregate but all faied.