thomwolf HF Staff commited on
Commit
dd27573
·
1 Parent(s): e2885e9
Files changed (2) hide show
  1. data/test.jsonl +0 -0
  2. wikitext_103.py +50 -0
data/test.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
wikitext_103.py ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import datasets
2
+ import os
3
+ import json
4
+
5
+
6
+ _CITATION = ""
7
+ _DESCRIPTION = """
8
+ Wikitext-103 dataset from this paper:
9
+ https://arxiv.org/pdf/1609.07843.pdf
10
+
11
+ Gopher's authors concatenate all the articles, set context length to n/2 (n = max_seq_len),
12
+ and use the "closed vocabulary" variant of the dataset for evaluation.
13
+
14
+ In contrast, we evaluate the model on each article independently, use single token contexts
15
+ (except for the last sequence in each document), and use the raw dataset.
16
+ """
17
+
18
+ class Wikitext103(datasets.GeneratorBasedBuilder):
19
+ VERSION = datasets.Version("1.0.0")
20
+
21
+ def _info(self):
22
+ features = datasets.Features(
23
+ {
24
+ "text": datasets.Value("string"),
25
+
26
+ }
27
+ )
28
+ return datasets.DatasetInfo(
29
+ description=_DESCRIPTION,
30
+ features=features,
31
+ homepage="",
32
+ license="",
33
+ citation=_CITATION,
34
+ )
35
+
36
+ def _split_generators(self, dl_manager):
37
+ test_json = dl_manager.download(os.path.join("data", "test.jsonl"))
38
+
39
+ return [
40
+ datasets.SplitGenerator(
41
+ name=datasets.Split.TEST,
42
+ gen_kwargs={"path": test_json},
43
+ )
44
+ ]
45
+
46
+ # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
47
+ def _generate_examples(self, path):
48
+ with open(path, encoding="utf-8") as f:
49
+ for key, row in enumerate(f):
50
+ yield key, json.loads(row)