Merge pull request #859 from baberabb/nqopen_baber_fi

[Refactor] NQopen

lm_eval/api/task.py

@@ -674,11 +674,11 @@ class ConfigurableTask(Task):
             check_choices = test_choice
         else:
             check_choices = [test_target]
-
+        if self.config.doc_to_choice is not None:
             for choice in check_choices:
-            choice_has_whitespace = True if " " in choice else False
+                choice_has_whitespace = True if choice[0].isspace() else False
                 delimiter_has_whitespace = (
-                True if " " in self.config.target_delimiter else False
+                    True if self.config.target_delimiter[-1].isspace() else False
                 )
 
                 if delimiter_has_whitespace and choice_has_whitespace:
@@ -1067,6 +1067,9 @@ class ConfigurableTask(Task):
                 # it assumes that doc_to_target returns a number.
                 choices = self.doc_to_choice(doc)
                 gold = choices[gold]
+            # we expect multiple_targets to be a list.
+            elif self.multiple_target:
+                gold = list(gold)
             else:
                 gold = str(gold)
 
@@ -1077,6 +1080,10 @@ class ConfigurableTask(Task):
                     # return true if any are true
                     # TODO: this may break for multipLe_target, non zero-or-1 metrics
                     scores = []
+                    if not isinstance(gold, list):
+                        # sometimes, a multiple_target dataset has exceptions where one doc has only one string answer
+                        # print(gold)
+                        gold = [gold]
                     for gold_option in gold:
                         try:
                             result_score = self._metric_fn_list[metric](

lm_eval/tasks/nq_open/nq_open.yaml

+task: nq_open
+dataset_path: nq_open
+output_type: greedy_until
+training_split: train
+validation_split: validation
+description: "Answer these questions:\n"
+doc_to_text: "Q: {{question}}?\nA:"
+doc_to_target: "{{answer}}" # TODO: should be multi-target
+fewshot_delimiter: "\n"
+generation_kwargs:
+  until:
+    - "\n"
+    - "."
+    - ","
+  do_sample: false
+  temperature: 0.0
+filter_list:
+  - name: remove_whitespace
+    filter:
+      - function: remove_whitespace
+      - function: take_first
+target_delimiter: " "
+metric_list:
+  - metric: exact_match
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: true
+    regexes_to_ignore:
+    - "\ban|a|the\b"