make hotpotqa less weird

mrqa.py

@@ -123,38 +123,39 @@ class MRQA(datasets.GeneratorBasedBuilder):
                 {
                     "subset": datasets.Value("string"),
                     "context": datasets.Value("string"),
-                    "context_tokens": datasets.Sequence(
-                        {
-                            "tokens": datasets.Value("string"),
-                            "offsets": datasets.Value("int32"),
-                        }
-                    ),
+                    # "context_tokens": datasets.Sequence(
+                    #     {
+                    #         "tokens": datasets.Value("string"),
+                    #         "offsets": datasets.Value("int32"),
+                    #     }
+                    # ),
                     "qid": datasets.Value("string"),
                     "question": datasets.Value("string"),
-                    "question_tokens": datasets.Sequence(
-                        {
-                            "tokens": datasets.Value("string"),
-                            "offsets": datasets.Value("int32"),
-                        }
-                    ),
-                    "detected_answers": datasets.Sequence(
-                        {
-                            "text": datasets.Value("string"),
-                            "char_spans": datasets.Sequence(
-                                {
-                                    "start": datasets.Value("int32"),
-                                    "end": datasets.Value("int32"),
-                                }
-                            ),
-                            "token_spans": datasets.Sequence(
-                                {
-                                    "start": datasets.Value("int32"),
-                                    "end": datasets.Value("int32"),
-                                }
-                            ),
-                        }
-                    ),
+                    # "question_tokens": datasets.Sequence(
+                    #     {
+                    #         "tokens": datasets.Value("string"),
+                    #         "offsets": datasets.Value("int32"),
+                    #     }
+                    # ),
+                    # "detected_answers": datasets.Sequence(
+                    #     {
+                    #         "text": datasets.Value("string"),
+                    #         "char_spans": datasets.Sequence(
+                    #             {
+                    #                 "start": datasets.Value("int32"),
+                    #                 "end": datasets.Value("int32"),
+                    #             }
+                    #         ),
+                    #         "token_spans": datasets.Sequence(
+                    #             {
+                    #                 "start": datasets.Value("int32"),
+                    #                 "end": datasets.Value("int32"),
+                    #             }
+                    #         ),
+                    #     }
+                    # ),
                     "answers": datasets.Sequence(datasets.Value("string")),
+                    "answer": datasets.Value("string"),
                 }
             ),
             supervised_keys=None,
@@ -203,29 +204,33 @@ class MRQA(datasets.GeneratorBasedBuilder):
                 for row in f:
                     paragraph = json.loads(row)
                     context = paragraph["context"].strip()
-                    context_tokens = [{"tokens": t[0], "offsets": t[1]} for t in paragraph["context_tokens"]]
+                    if subset == "HotpotQA":
+                        context = context.replace("[PAR] ", "\n\n")
+                        context = context.replace("[TLE]", "Title:")
+                        context = context.replace("[SEP]", "\nPassage:").strip()
+                    # context_tokens = [{"tokens": t[0], "offsets": t[1]} for t in paragraph["context_tokens"]]
                     for qa in paragraph["qas"]:
                         qid = qa["qid"]
                         question = qa["question"].strip()
-                        question_tokens = [{"tokens": t[0], "offsets": t[1]} for t in qa["question_tokens"]]
-                        detected_answers = []
-                        for detect_ans in qa["detected_answers"]:
-                            detected_answers.append(
-                                {
-                                    "text": detect_ans["text"].strip(),
-                                    "char_spans": [{"start": t[0], "end": t[1]} for t in detect_ans["char_spans"]],
-                                    "token_spans": [{"start": t[0], "end": t[1]} for t in detect_ans["token_spans"]],
-                                }
-                            )
+                        # question_tokens = [{"tokens": t[0], "offsets": t[1]} for t in qa["question_tokens"]]
+                        # detected_answers = []
+                        # for detect_ans in qa["detected_answers"]:
+                        #     detected_answers.append(
+                        #         {
+                        #             "text": detect_ans["text"].strip(),
+                        #             "char_spans": [{"start": t[0], "end": t[1]} for t in detect_ans["char_spans"]],
+                        #             "token_spans": [{"start": t[0], "end": t[1]} for t in detect_ans["token_spans"]],
+                        #         }
+                        #     )
                         answers = qa["answers"]
                         final_row = {
                             "subset": subset,
                             "context": context,
-                            "context_tokens": context_tokens,
+                            # "context_tokens": context_tokens,
                             "qid": qid,
                             "question": question,
-                            "question_tokens": question_tokens,
-                            "detected_answers": detected_answers,
+                            # "question_tokens": question_tokens,
+                            # "detected_answers": detected_answers,
                             "answers": answers,
                             "answer": answers[0]
                         }
@@ -234,5 +239,5 @@ class MRQA(datasets.GeneratorBasedBuilder):
 
 if __name__ == '__main__':
     from datasets import load_dataset
-    ssfd_debug = load_dataset("/Users/yuvalkirstain/repos/mrqa", name="hotpotqa")
+    ssfd_debug = load_dataset("/Users/yuvalkirstain/repos/mrqa", name="newsqa")
     x = 5