فيكس

lib/ml/evaluation.py

@@ -148,7 +148,7 @@ def evaluate(self, exp: str, snapshot: str, tvx_ehr: TVxEHR) -> Dict[str, float]
         model = model.load_params_from_archive(os.path.join(self.experiment_dir[exp], 'params.zip'), snapshot)
         subjects_list = tvx_ehr.subject_ids # list of subject_ids
         # randomly select at most 1000 subjects.
-        subjects_list = random.sample(subjects_list, min(750, len(subjects_list)))
+        subjects_list = random.sample(subjects_list, min(600, len(subjects_list)))
         predictions = model.batch_predict(tvx_ehr.device_batch(subjects_list))
         return metrics(predictions).as_df(snapshot).iloc[0].to_dict()