Update docci.py

docci.py

@@ -107,7 +107,7 @@ class DOCCI(datasets.GeneratorBasedBuilder):
         with open(data["descriptions"], "r") as f:
             examples = [json.loads(l.strip()) for l in f]
 
-            for ex in annotations["images"]:
+            for ex in examples:
                 if split == "train":
                     if not (ex["split"] == "train" and ex['example_id'].startswith("train")):
                         continue
@@ -134,7 +134,7 @@ class DOCCI(datasets.GeneratorBasedBuilder):
                 yield _ex["example_id"], _ex
 
     def _generate_examples_docci_aar(self, data, split):
-        image_files = glob.glob(data[split])
+        image_files = glob.glob(data["images"])
         
         for image_path in image_files: