Proposing to add functionality to save the list of not ranked recommendations (i.e., predicted candidate news and corresponding scores) for each user in JSON format.
Pull request adressing:
5ccc9f83e07529e5ce43295b026d3f7c0991506e - Streamlining how user IDs (from original dataset) and user indices (e.g., for indexing user embedding matrix) are handled across datasets.
bbf494bbd06bc98acda9124ccc58526bc760b75c - Adds functions to retrieve and save recommendations for given users and predictions in AbstractRecomender.
e0a1baab85930f3aa067544de51d54d7c05d0ae2 - Add explicit news IDs to batch to be passed to the model.
eca9f54f6c7c9dee72868ba7c84b0bdc64464ddc - Update models with option to save the recommendation lists for the users in the test set on on_test_epoch_end.
1b33d384ea5b3e77d1c8c78e206e9a6129bf768e - Update model configurations to reflect changes in the model hyperparameters for saving the recommendations lists.
Proposing to add functionality to save the list of not ranked recommendations (i.e., predicted candidate news and corresponding scores) for each user in JSON format.
Pull request adressing:
AbstractRecomender.on_test_epoch_end.