Correcting input types

keywords_evaluate.py

@@ -47,7 +47,7 @@ Examples:
     Here is an exemple on how to use the metric:
 
     >>> metric = evaluate.load("rfr2003/keywords_evaluate")
-    >>> results = metric.compute(generations=["yes", "no"], golds=[["yes"], ["yes"]], keywords={'yes', 'no'})
+    >>> results = metric.compute(generations=["yes", "no"], golds=[["yes"], ["yes"]], keywords=['yes', 'no'])
     >>> print(results)
     {'accuracy': 0.5}
 """
@@ -77,7 +77,7 @@ class Keywords_evaluate(evaluate.Metric):
         # TODO: Download external resources if needed
         pass
 
-    def _compute(self, generations, golds, keywords={'yes', 'no'}, strict=True):
+    def _compute(self, generations, golds, keywords=['yes', 'no'], strict=True):
         '''Calculate Accuracy scores between model generations and golden answers where the task is to generate the good(s) keyword(s) among a list of them. If strict is True, we expect to find all the expected keywords generated, if not we want only one'''
         assert len(generations) == len(golds)
         assert isinstance(golds, list)
@@ -85,6 +85,7 @@ class Keywords_evaluate(evaluate.Metric):
         correct, total = 0, 0
         
         if keywords:
+            keywords = set(keywords)
             pattern = r"\b(" + "|".join(map(re.escape, keywords)) + r")\b"
             
         else:

tests.py

@@ -2,21 +2,21 @@ test_cases = [
     {
         "generations": ["yes", "no"],
         "golds": [["yes"], ["yes"]],
-        "keywords": {'yes', 'no'}
+        "keywords": ['yes', 'no']
         "strict": True,
         "result": {"accuracy": 1.0}
     },
     {
         "generations": ["[up, left]", "[right]"],
         "golds": [['up', 'left'], ['right', 'down']],
-        "keywords": {'up', 'left', 'right', 'down'}
+        "keywords": ['up', 'left', 'right', 'down']
         "strict": True,
         "result": {"accuracy": 0.5}
     },
     {
         "generations": ["[up, left]", "[right]"],
         "golds": [['up', 'left'], ['right', 'down']],
-        "keywords": {'up', 'left', 'right', 'down'}
+        "keywords": ['up', 'left', 'right', 'down']
         "strict": False,
         "result": {"accuracy": 1.0}
     }