chagned data typed of choices

README.md

@@ -18,7 +18,7 @@ dataset_info:
   - name: prompt
     dtype: string
   - name: choices
-    sequence: string
+    dtype: string
   splits:
   - name: test
     num_bytes: 119949703

smokedataset_QA.py

@@ -42,12 +42,12 @@ class smokedataset_QA(datasets.GeneratorBasedBuilder):
                     "image": datasets.Image(),
                     "label": datasets.ClassLabel(names=_NAMES),
                     "prompt": datasets.Value(dtype='string'),
-                    "choices": datasets.Sequence(datasets.Value("string"))
+                    "choices": datasets.Value(dtype='string')
                 }
             ),
             # If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
             # specify them. They'll be used if as_supervised=True in builder.as_dataset.
-            #supervised_keys=None,
+            supervised_keys=None,
             # Homepage of the dataset for documentation
             homepage=_HOMEPAGE,
             # License for the dataset if available