close to done for bboxes

DUDE_loader.py

@@ -103,17 +103,15 @@ class DUDE(datasets.GeneratorBasedBuilder):
                 "questionId": datasets.Value("string"),
                 "question": datasets.Value("string"),
                 "answers": datasets.Sequence(datasets.Value("string")),
-                # ''' OPTIONAL?
-                # "answers_page_bounding_boxes": datasets.Sequence(
-                #     {
-                #         "left": datasets.Value("int32"),
-                #         "top": datasets.Value("int32"),
-                #         "width": datasets.Value("int32"),
-                #         "height": datasets.Value("int32"),
-                #         "page": datasets.Value("int32"),
-                #     }
-                # ),
-                # '''
+                "answers_page_bounding_boxes": datasets.Sequence(
+                    {
+                        "left": datasets.Value("int32"),
+                        "top": datasets.Value("int32"),
+                        "width": datasets.Value("int32"),
+                        "height": datasets.Value("int32"),
+                        "page": datasets.Value("int32"),
+                    }
+                ),
                 "answers_variants": datasets.Sequence(datasets.Value("string")),
                 "answer_type": datasets.Value("string"),
                 "data_split": datasets.Value("string"),
@@ -186,6 +184,12 @@ class DUDE(datasets.GeneratorBasedBuilder):
             a["data_split"] = split
             a["document"] = retrieve_doc(pdfs_archive, a["docId"])
             a["OCR"] = retrieve_OCR(OCR_archive, a["docId"])
-            a.pop("answers_page_bounding_boxes")  # fix later
 
+            # FIXES for faulty generation
+            #a.pop("answers_page_bounding_boxes")  # fix later            
+            if a["answers_page_bounding_boxes"] in [ [], [[]] ]:
+                a["answers_page_bounding_boxes"] = None
+            else:
+                if isinstance(a['answers_page_bounding_boxes'][0], list):
+                    a["answers_page_bounding_boxes"] = a['answers_page_bounding_boxes'][0]                    
             yield i, a