Add KorUnSmile task

lm_eval/tasks/__init__.py

@@ -56,6 +56,7 @@ from . import nsmc
 from . import klue
 from . import ko_translation
 from . import korquad
+from . import korunsmile
 
 ########################################
 # Translation tasks
@@ -319,7 +320,8 @@ TASK_REGISTRY = {
     "kobest_hellaswag": kobest.HellaSwag,
     "kobest_sentineg": kobest.SentiNeg,
     "ko_en_translation": ko_translation.KoEnTranslation,
-    "en_ko_translation": ko_translation.EnKoTranslation
+    "en_ko_translation": ko_translation.EnKoTranslation,
+    "korunsmile": korunsmile.KorUnSmile
 }
 
 

lm_eval/tasks/korunsmile.py

+"""
+ Korean UnSmile Dataset
+ 
+ Github: https://github.com/smilegate-ai/korean_unsmile_dataset
+"""
+
+import numpy as np
+from lm_eval.base import MultipleChoiceTask
+from lm_eval.metrics import macro_f1_score
+
+_CITATION = """
+@misc{SmilegateAI2022KoreanUnSmileDataset,
+  title         = {Korean UnSmile dataset: Human-annotated Multi-label Korean Hate Speech Dataset},
+  author        = {Seonghyun Kim},
+  year          = {2022},
+  howpublished  = {https://github.com/smilegate-ai/korean_unsmile_dataset},
+}
+"""
+
+class KorUnSmile(MultipleChoiceTask):
+    VERSION = 0
+    DATASET_PATH = "smilegate-ai/kor_unsmile"
+    DATASET_NAME = None
+
+    def has_training_docs(self):
+        return True
+
+    def has_validation_docs(self):
+        return True
+
+    def has_test_docs(self):
+        return False
+
+    def training_docs(self):
+        if self._training_docs is None:
+            self._training_docs = list(map(self._process_doc,self.dataset["train"]))
+        return self._training_docs
+
+    def validation_docs(self):
+        return map(self._process_doc,self.dataset["valid"])
+
+    def _process_doc(self, doc):
+        out_doc = {
+            "title": doc["문장"],
+            "choices": ["여성/가족", "남성", "성소수자", "인종/국적", "연령", "지역", "종교", "기타 혐오", "악플/욕설", "clean"],
+            "gold": np.argmax(doc["labels"])
+        }
+        return out_doc
+
+    def doc_to_text(self, doc):
+        return "{}".format(doc["title"])
+
+    def doc_to_target(self, doc):
+        return " {}".format({0: "여성/가족", 1: "남성", 2: "성소수자", 3: "인종/국적", 4: "연령", 5: "지역", 6: "종교", 7: "기타 혐오", 8: "악플/욕설", 9: "clean"}[doc["gold"]])
+
+    def process_results(self, doc, results):
+        pred = np.argmax(results)
+        gold = doc["gold"]
+        return {
+            "f1": (gold, pred)
+        }
+
+    def higher_is_better(self):
+        return {
+            "f1": True
+        }
+
+    def aggregation(self):
+        return {
+            "f1": macro_f1_score
+        }