update script and readme

README.md

@@ -1,8 +1,9 @@
 ---
 annotations_creators:
-  - crowdsourced
+  - found
 language_creators:
-  - crowdsourced
+  - found
+  - expert-generated
 language:
   - en
   - es
@@ -21,7 +22,7 @@ multilinguality:
 size_categories:
   - 1K<n<10K
 source_datasets:
-  - original
+  - extended|gsm8k
 task_categories:
   - text2text-generation
 task_ids: []

mgsm.py

@@ -13,9 +13,14 @@
 # limitations under the License.
 """Multilingual Grade School Math Benchmark (MGSM)"""
 
-import datasets
 import csv
-from mgsm import exemplars
+from .exemplars import (
+    EXEMPLAR_NUMBER_ANSWERS,
+    EXEMPLAR_EQUATION_SOLUTIONS,
+    MGSM_EXEMPLARS,
+)
+import datasets
+
 
 _CITATION = """\
 @article{cobbe2021gsm8k,
@@ -112,9 +117,9 @@ class MGSM(datasets.GeneratorBasedBuilder):
     def _generate_examples(self, filepath):
         if filepath.endswith(".py"):
             name = self.config.name
-            examples = exemplars.MGSM_EXEMPLARS[name]
-            number_answers = exemplars.EXEMPLAR_NUMBER_ANSWERS
-            equation_solutions = exemplars.EXEMPLAR_EQUATION_SOLUTIONS
+            examples = MGSM_EXEMPLARS[name]
+            number_answers = EXEMPLAR_NUMBER_ANSWERS
+            equation_solutions = EXEMPLAR_EQUATION_SOLUTIONS
             for key, data in examples.items():
                 yield key, {
                     "question": data["q"],
@@ -128,8 +133,6 @@ class MGSM(datasets.GeneratorBasedBuilder):
                     csv_file,
                     quotechar='"',
                     delimiter="\t",
-                    quoting=csv.QUOTE_ALL,
-                    skipinitialspace=True,
                 )
                 for key, row in enumerate(csv_reader):
                     yield key, {