the big change

mapped_multitask.py

@@ -30,18 +30,6 @@ def to_dict_element(el, cols):
     return final_dict
 
 
-
-def mega_hash(func,dataset_name,dataset_config,dataset_obj,split):
-    hasher = Hasher()
-    hasher.update(repr(dataset_obj))
-    hasher.update(pickle.dumps(func))
-    hasher.update(split)
-    hasher.update(dataset_config)
-    hasher.update(dataset_name)
-    
-
-    return hasher.hexdigest()
-
 logger = datasets.logging.get_logger(__name__)
 
 
@@ -81,57 +69,6 @@ class MappedMultitaskConfig(datasets.BuilderConfig):
         self.feature_format = feature_format
 
 
-
-
-def to_source_target(example):
-    # print(type())
-    source = []
-    target = []
-    meta_list = []
-    for ctx, ans_for, title, question, qids, cids, answer_ids, answer_list in zip(
-        example["positive_ctxs.text"],
-        example["positive_ctxs.pid"],
-        example["positive_ctxs.title"],
-        example["question_text"],
-        example["qid"],
-        example["positive_ctxs.chunk_id"],
-        example["answer_list.aid"],
-        example["answer_list.answer_text"],
-    ):
-        ans_mapping = {idx.split("__")[-1]: ans_str for idx, ans_str in zip(answer_ids, answer_list)}
-        for c, a, t, _, _, cid in zip(ctx, ans_for, title, question, qids, cids):
-            source.append(f"Title: {t}\nText: {c}\nQuestion: {question}\n")
-            target.append(f"Answer: {ans_mapping[a.split('__')[-2]]}")
-            meta_list.append({"id": cid, "qid": qids, "question": question, "title": t, "text": c})
-    for ctx, title, question, qids, cids in zip(
-        example["hard_negative_ctxs.text"],
-        example["hard_negative_ctxs.title"],
-        example["question_text"],
-        example["qid"],
-        example["positive_ctxs.chunk_id"],
-    ):
-        for c, t, _, _, cid in zip(ctx, title, question, qids, cids):
-            source.append(f"Title: {t}\nText: {c}\nQuestion: {question}\n")
-            target.append("Not relevant")
-            meta_list.append({"id": cid, "qid": qids, "question": question, "title": t, "text": c})
-    return {"target": target, "source": source, "meta": meta_list}
-
-
-def transform_dpr(dataset,dataset_name,dataset_config):
-    
-    for split in dataset.column_names:
-        _split_ds = dataset[split].flatten()
-        fingerprint = mega_hash(to_source_target,dataset_name,
-                            dataset_config,_split_ds,split)
-        dataset[split] = _split_ds.map(
-            to_source_target,
-            batched=True,
-            remove_columns=_split_ds.column_names,
-            new_fingerprint = fingerprint
-        )
-    return dataset
-
-
 class MappedMultitask(datasets.GeneratorBasedBuilder):
 
     BUILDER_CONFIGS = [
@@ -189,21 +126,16 @@ class MappedMultitask(datasets.GeneratorBasedBuilder):
 
     def _generate_examples(self, split):
         """This function returns the examples in the raw (text) form."""
-        qampari_path = "/home/ohadr/ssd/dalle-mini/qampari/mapped_qampari.py"
-        nq_path = "/home/ohadr/ssd/dalle-mini/qampari/mapped_nq.py"
         dataset_list = []
-        qampari = load_dataset(qampari_path, self.info.config_name)
-        if split in get_config_splits(qampari_path)[self.info.config_name] and split in qampari:
+        qo_path = "/home/joberant/home/ohadr/qampari/mapped_qampari.py"
+        qampari = load_dataset(qo_path, self.info.config_name)
+        if split in get_config_splits(qo_path)[self.info.config_name] and split in qampari:
             dataset_list.append(qampari[split].flatten())
-        
+        nq_path = "/home/joberant/home/ohadr/qampari/mapped_nq.py"
         nq = load_dataset(nq_path, self.info.config_name)
         if split in get_config_splits(nq_path)[self.info.config_name] and split in nq:
-            dataset_list.append(nq[split].flatten())
-        
-        
-        
+            dataset_list.append(nq[split].flatten())        
         flattened_dataset = interleave_datasets(datasets=dataset_list).flatten()
-        # print(flattened_dataset)
         
         for i,element in enumerate(flattened_dataset):
             new_element = dict(source=element['source'],target=element['target'])
@@ -213,31 +145,9 @@ class MappedMultitask(datasets.GeneratorBasedBuilder):
                                         title=element['meta.title'],
                                         text=element['meta.text'],
                                         )
-            
-            # element = to_dict_element(element,cols=flattened_dataset.column_names)
-            # print(element)
             yield i, new_element
             
 
-        # if self.feature_format=="reranking":
-        #     fingerprint = mega_hash(to_source_target,"multitask",
-        #                         self.info.config_name,flattened_dataset,split)
-        #     transformed_dataset = flattened_dataset.map(
-        #                 to_source_target,
-        #                 batched=True,
-        #                 remove_columns=flattened_dataset.column_names,
-        #                 new_fingerprint = fingerprint
-        #             )
-        #     for i,element in enumerate(transformed_dataset):
-        #         yield i,element
-        # elif self.feature_format=="inference":
-        
-
-        # else:
-        #     assert False
-
-
-
 
 
 

mapped_nq.py

@@ -77,7 +77,7 @@ inference_mapped_features = Features(base_features)
 
 
 class MappedNQConfig(datasets.BuilderConfig):
-    """BuilderConfig for MappedQampariDPR."""
+    """BuilderConfig for MappedNQ."""
 
     def __init__(self, features=None, retriever=None, feature_format=None, **kwargs):
         super(MappedNQConfig, self).__init__(**kwargs)
@@ -159,6 +159,14 @@ class MappedNQ(datasets.GeneratorBasedBuilder):
             retriever="dprnq",
             feature_format="inference",
         ),
+        MappedNQConfig(
+            name="inference_bm25",
+            version=datasets.Version("1.0.1", ""),
+            description="MappedNQ dataset in DPR format with the bm25 retrieval results",
+            features=inference_mapped_features,
+            retriever="bm25",
+            feature_format="inference",
+        ),
     ]
 
     def _info(self):
@@ -175,20 +183,24 @@ class MappedNQ(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        return [
-            datasets.SplitGenerator(
+        split_list = dict(
+        train=datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={"split": "train"},
             ),
-            datasets.SplitGenerator(
+        
+        validation=datasets.SplitGenerator(
                 name=datasets.Split.VALIDATION,
                 gen_kwargs={"split": "validation"},
             ),
-            datasets.SplitGenerator(
+        test=datasets.SplitGenerator(
                 name=datasets.Split.TEST,
                 gen_kwargs={"split": "test"},
-            ),
-        ]
+            )
+            )
+        if self.info.config_name=="inference_bm25":
+            split_list.pop("train")
+        return list(split_list.values())
 
     def _prepare_split(self, split_generator, **kwargs):
         self.info.features = self.config.features
@@ -196,13 +208,17 @@ class MappedNQ(datasets.GeneratorBasedBuilder):
 
     def _generate_examples(self, split):
         """This function returns the examples in the raw (text) form."""
-        
-        path = "/home/ohadr/ssd/dalle-mini/qampari/nq.py"
+        # datasets.inspect_dataset("iohadrubin/nq","tmp_nq")
+        path = "/home/joberant/home/ohadr/qampari/nq.py"
+        flattened_dataset = load_dataset(path,
+                                                    self.info.config_name,
+                                                    split=split).flatten()
+
         flattened_dataset = load_dataset(path, self.info.config_name).flatten()
         if split not in get_config_splits(path)[self.info.config_name] or split not in flattened_dataset:
             return
         flattened_dataset = flattened_dataset[split]
-        print(self.info)
+        # print(self.info)
         if self.feature_format=="reranking":
             
             # flattened_dataset = flattened_dataset[split]
@@ -223,7 +239,6 @@ class MappedNQ(datasets.GeneratorBasedBuilder):
                 element = to_dict_element(element,cols=flattened_dataset.column_names)
                 for j,ctx in enumerate(element['ctxs']):
                     qid,ctx,question = element['qid'],ctx,element["question"]
-                    # ctx.pop("score",None)
                     ctx.pop("score",None)
                     assert "id" in ctx
                     source_element = {"source": f"Title: {ctx['title']}\nText: {ctx['text']}\nQuestion: {question}\n",
@@ -234,4 +249,4 @@ class MappedNQ(datasets.GeneratorBasedBuilder):
                     yield f"{qid}__{ctx['id']}", source_element
 
         else:
-            assert False
+            assert False

mapped_qampari.py

@@ -200,7 +200,9 @@ class MappedQampari(datasets.GeneratorBasedBuilder):
 
     def _generate_examples(self, split):
         """This function returns the examples in the raw (text) form."""
-        flattened_dataset = load_dataset("/home/ohadr/ssd/dalle-mini/qampari/qampari.py",
+        # datasets.inspect_dataset("iohadrubin/qampari","tmp_qampari")
+
+        flattened_dataset = load_dataset("/home/joberant/home/ohadr/qampari/qampari.py",
                                                     self.info.config_name,
                                                     split=split).flatten()
         if self.feature_format=="reranking":

nq.py

@@ -115,7 +115,8 @@ RETDPR_INF_URLS = {
 }
 
 RETBM25_INF_URLS = {
-    # split:f"https://dl.fbaipublicfiles.com/dpr/data/retriever_results/single/nq-{split}.json.gz" for split in ["train", "dev","test"]
+    split:f"https://www.cs.tau.ac.il/~ohadr/nq-{split}.json.gz" for split in ["dev","test"]
+    
 }
 RETBM25_RERANKING_features = Features(
     {
@@ -196,7 +197,9 @@ RETDPR_INF_features = Features(
 )
 URL_DICT = {"reranking_dprnq":RETDPR_RERANKING_URLS, 
             "reranking_bm25":RETBM25_RERANKING_URLS,
-            "inference_dprnq":RETDPR_INF_URLS}
+            "inference_dprnq":RETDPR_INF_URLS,
+            "inference_bm25":RETBM25_INF_URLS,
+            }
 
 class NatQuestions(datasets.GeneratorBasedBuilder):
 
@@ -228,15 +231,16 @@ class NatQuestions(datasets.GeneratorBasedBuilder):
             feature_format="inference",
             url=URL_DICT,
         ),
-        # NatQuestionsConfig(
-        #     name="inference_bm25",
-        #     version=datasets.Version("1.0.1", ""),
-        #     description="NatQuestions dataset in a format accepted by the inference model, performing reranking on the bm25 retrieval results",
-        #     features=inference_features,
-        #     retriever="bm25",
-        #     feature_format="inference",
-        #     url=RETBM25_INF_URLS
-        # ),
+        NatQuestionsConfig(
+            name="inference_bm25",
+            version=datasets.Version("1.0.1", ""),
+            description="NatQuestions dataset in a format accepted by the inference model, performing reranking on the bm25 retrieval results",
+            features=RETDPR_INF_features,
+            retriever="bm25",
+            feature_format="inference",
+            url=URL_DICT,
+        ),
+
     ]
 
     def _info(self):
@@ -263,6 +267,7 @@ class NatQuestions(datasets.GeneratorBasedBuilder):
         # filepath = "/home/joberant/home/ohadr/testbed/notebooks/NatQuestions_retrievers"
         
         result = []
+
         if "train" in filepath[self.info.config_name]:
             result.append(
                 datasets.SplitGenerator(