def test_recs(RecommenderClass: Type[BaseRecommender]) -> None:
"""Test the learning of recommenders exit normally, and they are picklable.
Args:
RecommenderClass (Type[BaseRecommender]): The recommender class to be tested.
"""
rec = RecommenderClass(X_train)
rec.learn()
scores = rec.get_score(np.arange(X_train.shape[0]))
eval = Evaluator(X_test, 0, 20)
with pytest.raises(ValueError):
eval.get_score(rec)
metrics = eval.get_scores(rec, cutoffs=[X_train.shape[1]])
assert np.all(np.isfinite(scores))
assert np.all(~np.isnan(scores))
for value in metrics.values():
assert ~np.isnan(value)
assert np.isfinite(value)
with open("temp.pkl", "wb") as ofs:
pickle.dump(rec, ofs)
with open("temp.pkl", "rb") as ifs:
rec_dumped: BaseRecommender = pickle.load(ifs)
score_from_dumped = rec_dumped.get_score(np.arange(X_train.shape[0]))
np.testing.assert_allclose(scores, score_from_dumped)
test_configs: List[Tuple[Type[BaseOptimizer], int]] = [
(TopPopOptimizer, 1),
(CosineKNNOptimizer, 40),
(AsymmetricCosineKNNOptimizer, 40),
(TverskyIndexKNNOptimizer, 40),
(DenseSLIMOptimizer, 20),
(P3alphaOptimizer, 40),
(RP3betaOptimizer, 40),
(IALSOptimizer, 40),
# (BPRFMOptimizer, 40),
# (MultVAEOptimizer, 5),
# (SLIMOptimizer, 40),
]
for optimizer_class, n_trials in test_configs:
name = optimizer_class.__name__
optimizer: BaseOptimizer = optimizer_class(X_train_all,
valid_evaluator)
(best_param,
validation_results) = optimizer.optimize(timeout=14400,
n_trials=n_trials)
validation_results.to_csv(f"{name}_validation_scores.csv")
test_recommender = optimizer.recommender_class(X_train_val_all,
**best_param)
test_recommender.learn()
test_scores = test_evaluator.get_scores(test_recommender, [5, 10, 20])
test_results.append(
dict(name=name, best_param=best_param, **test_scores))
with open("test_results.json", "w") as ofs:
json.dump(test_results, ofs, indent=2)