migrate to flat schema

README.md

@@ -4,7 +4,7 @@ language:
 license: apache-2.0
 tags:
   - text
-pretty_name: MSMARCO
+pretty_name: MS MARCO
 size_categories:
   - "100K<n<1M"
 source_datasets:
@@ -16,18 +16,12 @@ dataset_info:
   features:
   - name: query
     dtype: string
-  - name: pos
-    list:
-    - name: doc
-      dtype: string
-    - name: score
-      dtype: float
-  - name: neg
-    list:
-    - name: doc
-      dtype: string
-    - name: score
-      dtype: float
+  - name: positive
+    sequence:
+    - dtype: string
+  - name: negative
+    sequence:
+    - dtype: string
   splits:
     - name: train
       num_bytes: 89609915
@@ -55,23 +49,21 @@ configs:
     path: "data/dev/*"
 ---
 
-# MSMARCO dataset
+# MS MARCO dataset
 
 A dataset in a [nixietune](https://github.com/nixiesearch/nixietune) compatible format:
 
 ```json
 {
   "query": ")what was the immediate impact of the success of the manhattan project?",
-  "pos": [
-    {
-      "doc": "The presence of communication amid scientific minds was equally important to the success of the Manhattan Project as scientific intellect was. The only cloud hanging over the impressive achievement of the atomic researchers and engineers is what their success truly meant; hundreds of thousands of innocent lives obliterated.",
-      "score": 1
-    }
-  ]
+  "positive": [
+      "The presence of communication amid scientific minds was equally important to the success of the Manhattan Project as scientific intellect was. The only cloud hanging over the impressive achievement of the atomic researchers and engineers is what their success truly meant; hundreds of thousands of innocent lives obliterated."
+  ],
+  "negative": []
 }
 ```
 
-This is the original converted dataset with the following splits:
+This is the original [BeIR/msmarco](https://huggingface.co/datasets/BeIR/msmarco) converted dataset with the following splits:
 * train: 502939 queries, only positives.
 * test: 43 queries, positives and negatives.
 * dev: 6980 queries, only positives.
@@ -81,7 +73,8 @@ This is the original converted dataset with the following splits:
 ```python
 from datasets import load_dataset
 
-data = load_dataset('nixiesearch/msmarco', split="train")
+data = load_dataset('nixiesearch/ms_marco')
+print(data["train"].features)
 ```
 
 ## License

convert.py

@@ -1,10 +1,11 @@
-from datasets import load_dataset
+from datasets import load_dataset, Features, Value, Sequence
 from dataclasses import dataclass, field
 import logging
 from transformers import HfArgumentParser
 from tqdm import tqdm
 from typing import Dict, List
 import json
+import numpy as np
 
 logger = logging.getLogger()
 logger.setLevel(logging.INFO)
@@ -17,7 +18,6 @@ logger.handlers = [console_handler]
 
 @dataclass
 class ConversionAgruments:
-    path: str = field(metadata={"help": "Path to the MAMARCO dataset"})
     out: str = field(metadata={"help": "Output path"})
 
 
@@ -27,22 +27,20 @@ class QRel:
     score: int
 
 
-def load_json(path: str, split: str = "train") -> List[str]:
-    dataset = load_dataset("json", data_files=path, split=split)
-    cache: List[str] = []
-    for row in tqdm(dataset, desc=f"loading {path}"):
+def load_msmarco(path: str, split) -> Dict[int, str]:
+    dataset = load_dataset(path, split, split=split)
+    cache: Dict[int, str] = {}
+    for row in tqdm(dataset, desc=f"loading {path} split={split}"):
         index = int(row["_id"])
-        if index >= len(cache):
-            cache.extend([""] * (1 + 2 * max(index, len(cache))))
         cache[index] = row["text"]
     return cache
 
 
-def load_qrel(path: str) -> Dict[int, List[QRel]]:
-    dataset = load_dataset("csv", data_files=path, split="train", delimiter="\t")
+def load_qrel(path: str, split: str) -> Dict[int, List[QRel]]:
+    dataset = load_dataset(path, split=split)
     print(dataset.features)
     cache: Dict[int, List[QRel]] = {}
-    for row in tqdm(dataset, desc=f"loading {path}"):
+    for row in tqdm(dataset, desc=f"loading {path} split={split}"):
         qid = int(row["query-id"])
         qrel = QRel(int(row["corpus-id"]), int(row["score"]))
         if qid in cache:
@@ -52,26 +50,14 @@ def load_qrel(path: str) -> Dict[int, List[QRel]]:
     return cache
 
 
-def process(
-    qrels: Dict[int, List[QRel]], queries: List[str], corpus: List[str]
+def process_raw(
+    qrels: Dict[int, List[QRel]], queries: Dict[int, str], corpus: Dict[int, str]
 ) -> List[Dict]:
     result = []
     for query, rels in tqdm(qrels.items(), desc="processing split"):
-        pos = [
-            {"doc": corpus[rel.doc], "score": rel.score}
-            for rel in rels
-            if rel.doc < len(corpus) and rel.score > 0 and corpus[rel.doc] != ""
-        ]
-        neg = [
-            {"doc": corpus[rel.doc], "score": rel.score}
-            for rel in rels
-            if rel.doc < len(corpus) and rel.score == 0 and corpus[rel.doc] != ""
-        ]
-        group = {"query": queries[query], "pos": pos}
-        if len(neg) == 0:
-            group["neg"] = None
-        else:
-            group["neg"] = neg
+        pos = [corpus[rel.doc] for rel in rels if rel.doc in corpus and rel.score > 0]
+        neg = [corpus[rel.doc] for rel in rels if rel.doc in corpus and rel.score == 0]
+        group = {"query": queries[query], "positive": pos, "negative": neg}
         result.append(group)
     return result
 
@@ -81,15 +67,15 @@ def main():
     (args,) = parser.parse_args_into_dataclasses()
     print(f"Args: {args}")
     qrels = {
-        "dev": load_qrel(f"{args.path}/qrels/dev.tsv"),
-        "train": load_qrel(f"{args.path}/qrels/train.tsv"),
-        "test": load_qrel(f"{args.path}/qrels/test.tsv"),
+        "train": load_qrel("BeIR/msmarco-qrels", split="train"),
+        "test": load_qrel("BeIR/msmarco-qrels", split="test"),
+        "dev": load_qrel("BeIR/msmarco-qrels", split="validation"),
     }
-    corpus = load_json(f"{args.path}/corpus.jsonl", split="train")
-    queries = load_json(f"{args.path}/queries.jsonl")
+    queries = load_msmarco("BeIR/msmarco", split="queries")
+    corpus = load_msmarco("BeIR/msmarco", split="corpus")
     print("processing done")
     for split, data in qrels.items():
-        dataset = process(data, queries, corpus)
+        dataset = process_raw(data, queries, corpus)
         with open(f"{args.out}/{split}.jsonl", "w") as out:
             for item in dataset:
                 json.dump(item, out)

data/dev/dev.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dbd532e51439da3bd74c79105f9f9f27beafd1db73a1b2155713ec37c8fd376d
-size 1206403
+oid sha256:fdf0538cbf00e68f4904763b074ef192128ef75f668b509e36826e13316103c3
+size 1199510

data/test/test.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b454c2d3e35adf12716e64699cc6637f703d9e82fc46f036cfa5af59433edc5
-size 969945
+oid sha256:d7daf1e3f97722ae59b888591756a12846b2fc02b89ee536d189faa27e684136
+size 951166

data/train/train.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d56a52df9c3bbd14ee6f92cef2aeb61bcf6d33bc9e85012559b53a5f246157a0
-size 89609915
+oid sha256:ce3b897ce69b424181aef6c25b8e17e20d0928a3d591f72d739db92b9346c573
+size 89120825

use.py

@@ -1,20 +1,4 @@
-from datasets import load_dataset, Features, Value, Sequence
+from datasets import load_dataset
 
-
-# data = load_dataset("ms_marco", )
-
-data = load_dataset(
-    "json",
-    data_files={
-        "train": "/home/shutty/data/nixiesearch-datasets/msmarco/train.jsonl",
-        "test": "/home/shutty/data/nixiesearch-datasets/msmarco/test.jsonl",
-    },
-    features=Features(
-        {
-            "query": Value("string"),
-            "pos": [{"doc": Value("string"), "score": Value("int32")}],
-            "neg": [{"doc": Value("string"), "score": Value("int32")}],
-        }
-    ),
-)
-print(data["test"].features)
+data = load_dataset("nixiesearch/ms_marco")
+print(data)