save

bangalore_score.py

@@ -229,7 +229,15 @@ class Bangalore_Score(evaluate.Metric):
         # print(f'\n\naggregate_score: {aggregate_score}')
         return {'amsr': aggregate_score}
 
-    def _compute(self, predictions, references, num_proc=None):
+    def _compute(
+            self,
+            predictions,
+            references,
+            metric,
+            metric_key,
+            max_score=True,
+            num_proc=None
+    ):
         """Returns the scores"""
         import json
         import evaluate
@@ -246,13 +254,12 @@ class Bangalore_Score(evaluate.Metric):
         proc_ds = Dataset.from_dict({'pred': predictions, 'ref': references})
 
         ### compute amsr ###
-        metric, metric_key = evaluate.load('evaluate-metric/meteor'), 'meteor'
         proc_ds = proc_ds.map(
-            lambda x: self.amsr_fn(x, metric=metric, metric_key=metric_key),
+            lambda x: self.amsr_fn(x, metric=metric, metric_key=metric_key, max_score=max_score),
             num_proc=num_proc, desc='computing amsr'
         )
 
         amsr_mean = sum(proc_ds['amsr']) / len(proc_ds['amsr'])
-        return {'amsr': amsr_mean}
+        return {f'amsr_{metric_key}': amsr_mean}