fix non-seriarizable bug

__main__.py

@@ -50,7 +50,7 @@ results = metric.compute(
 )
 logger.info(f"Time taken: {time.time() - s}")
 
-print(results)
+print(json.dumps(results))
 if args.output:
     with open(args.output, "w") as f:
         json.dump(results, f)

matching_series.py

@@ -208,7 +208,7 @@ class matching_series(evaluate.Metric):
         else:
             distance = self._compute_metric(predictions[:, None], references, metric=metric, axis=1)
 
-        index_distance = distance.diagonal(axis1=0, axis2=1).mean()
+        index_distance = distance.diagonal(axis1=0, axis2=1).mean().item()
 
         # matching scores
         distance_mean = distance.mean(axis=-1)
@@ -218,12 +218,12 @@ class matching_series(evaluate.Metric):
 
         # matching distance
         # shape: (num_generation,)
-        precision_distance = distance_mean[np.arange(len(best_match)), best_match].mean()
+        precision_distance = distance_mean[np.arange(len(best_match)), best_match].mean().item()
 
         # best match for each reference time series
         # shape: (num_reference,)
         best_match_inv = np.argmin(distance_mean, axis=0)
-        recall_distance = distance_mean[best_match_inv, np.arange(len(best_match_inv))].mean()
+        recall_distance = distance_mean[best_match_inv, np.arange(len(best_match_inv))].mean().item()
 
         f1_distance = 2 / (1 / precision_distance + 1 / recall_distance)
         mean_distance = (precision_distance + recall_distance) / 2
@@ -246,11 +246,11 @@ class matching_series(evaluate.Metric):
         cuc_features = []
         for f in range(predictions.shape[-1]):
             distance_f = distance[:, :, f]
-            index_distance_f = distance_f.diagonal(axis1=0, axis2=1).mean()
+            index_distance_f = (distance_f.diagonal(axis1=0, axis2=1).mean()).item()
             best_match_f = np.argmin(distance_f, axis=-1)
-            precision_distance_f = distance_f[np.arange(len(best_match_f)), best_match_f].mean()
+            precision_distance_f = (distance_f[np.arange(len(best_match_f)), best_match_f].mean()).item()
             best_match_inv_f = np.argmin(distance_f, axis=0)
-            recall_distance_f = distance_f[best_match_inv_f, np.arange(len(best_match_inv_f))].mean()
+            recall_distance_f = (distance_f[best_match_inv_f, np.arange(len(best_match_inv_f))].mean()).item()
             f1_distance_f = 2 / (1 / precision_distance_f + 1 / recall_distance_f)
             mean_distance_f = (precision_distance_f + recall_distance_f) / 2
             precision_distance_features.append(precision_distance_f)
@@ -361,7 +361,7 @@ class matching_series(evaluate.Metric):
                 sample = np.random.choice(match, size=n_sample, replace=False)  # type: ignore
                 coverage += len(np.unique(sample)) / n_reference
             coverages.append(coverage / n_calculation)
-        cuc = np.trapz(coverages, n_samples) / len(n_samples) / max(n_samples)
+        cuc = (np.trapz(coverages, n_samples) / len(n_samples) / max(n_samples)).item()
         return coverages, cuc
 
     @staticmethod