zhangir-azerbayev
commited on
Commit
•
4e223b3
1
Parent(s):
0a04f15
Revert "Fix Dataset Viewer (#2)"
Browse filesThis reverts commit 0a04f150620ad256bc01da0e83a85e4df0701186.
- README.md +0 -25
- proof-pile-2.py +157 -0
README.md
CHANGED
@@ -7,31 +7,6 @@ tags:
|
|
7 |
- math
|
8 |
size_categories:
|
9 |
- 10B<n<100B
|
10 |
-
configs:
|
11 |
-
- config_name: arxiv
|
12 |
-
data_files:
|
13 |
-
- split: train
|
14 |
-
path: arxiv/train/*.jsonl.zst
|
15 |
-
- split: validation
|
16 |
-
path: arxiv/validation/*.jsonl.zst
|
17 |
-
- split: test
|
18 |
-
path: arxiv/test/*.jsonl.zst
|
19 |
-
- config_name: open-web-math
|
20 |
-
data_files:
|
21 |
-
- split: train
|
22 |
-
path: open-web-math/train/*.jsonl.zst
|
23 |
-
- split: validation
|
24 |
-
path: open-web-math/validation/*.jsonl.zst
|
25 |
-
- split: test
|
26 |
-
path: open-web-math/test/*.jsonl.zst
|
27 |
-
- config_name: algebraic-stack
|
28 |
-
data_files:
|
29 |
-
- split: train
|
30 |
-
path: algebraic-stack/train/*.jsonl.zst
|
31 |
-
- split: validation
|
32 |
-
path: algebraic-stack/validation/*.jsonl.zst
|
33 |
-
- split: test
|
34 |
-
path: algebraic-stack/test/*.jsonl.zst
|
35 |
---
|
36 |
<img src="proofpile_logo.jpg" width="500">
|
37 |
|
|
|
7 |
- math
|
8 |
size_categories:
|
9 |
- 10B<n<100B
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
---
|
11 |
<img src="proofpile_logo.jpg" width="500">
|
12 |
|
proof-pile-2.py
ADDED
@@ -0,0 +1,157 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import os
|
3 |
+
import zstandard as zstd
|
4 |
+
|
5 |
+
import datasets
|
6 |
+
|
7 |
+
_CITATION="""\
|
8 |
+
@article{azerbayev2023llemma,
|
9 |
+
title={Llemma: an open language model for mathematics},
|
10 |
+
author={Zhangir Azerbayev and Hailey Schoelkopf and Keiran Paster and Marco Dos Santos and Stephen McAleer and Albert Q. Jiang and Jia Deng and Stella Biderman and Sean Welleck},
|
11 |
+
eprint={xyz.xyz},
|
12 |
+
archivePrefix={arXiv}
|
13 |
+
year={2023}
|
14 |
+
}
|
15 |
+
"""
|
16 |
+
|
17 |
+
_DESCRIPTION = """\
|
18 |
+
A dataset of high quality mathematical text. """
|
19 |
+
_HOMEPAGE = "https://github.com/EleutherAI/math-lm"
|
20 |
+
|
21 |
+
|
22 |
+
SPLITS = ["train", "validation", "test"]
|
23 |
+
|
24 |
+
_DATA_PATHS = {
|
25 |
+
"arxiv": {
|
26 |
+
split: [f'arxiv/{split}/arXiv_{str(i).zfill(3)}.jsonl.zst' for i in range(100)]
|
27 |
+
for split in SPLITS
|
28 |
+
},
|
29 |
+
"open-web-math": {
|
30 |
+
split: [
|
31 |
+
os.path.join(f"open-web-math/{split}", filename)
|
32 |
+
for filename in os.listdir(f"open-web-math/{split}")
|
33 |
+
]
|
34 |
+
for split in SPLITS
|
35 |
+
},
|
36 |
+
"algebraic-stack": {
|
37 |
+
split: [
|
38 |
+
os.path.join(f"algebraic-stack/{split}", filename)
|
39 |
+
for filename in os.listdir(f"algebraic-stack/{split}")
|
40 |
+
]
|
41 |
+
for split in SPLITS
|
42 |
+
}
|
43 |
+
}
|
44 |
+
|
45 |
+
class ProofPile2Config(datasets.BuilderConfig):
|
46 |
+
"""BuilderConfig for RedPajama sample."""
|
47 |
+
|
48 |
+
def __init__(self, *args, subsets, **kwargs):
|
49 |
+
"""BuilderConfig for ProofPile2.
|
50 |
+
Args:
|
51 |
+
**kwargs: keyword arguments forwarded to super.
|
52 |
+
"""
|
53 |
+
super(ProofPile2Config, self).__init__(**kwargs)
|
54 |
+
self.subsets = subsets
|
55 |
+
|
56 |
+
|
57 |
+
class ProofPile2(datasets.GeneratorBasedBuilder):
|
58 |
+
"""A large dataset of mathematical text."""
|
59 |
+
VERSION = datasets.Version("1.1.0")
|
60 |
+
|
61 |
+
# This is an example of a dataset with multiple configurations.
|
62 |
+
# If you don't want/need to define several sub-sets in your dataset,
|
63 |
+
# just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.
|
64 |
+
|
65 |
+
# If you need to make complex sub-parts in the datasets with configurable options
|
66 |
+
# You can create your own builder configuration class to store attribute, inheriting from ProofPile2Config
|
67 |
+
# BUILDER_CONFIG_CLASS = MyBuilderConfig
|
68 |
+
|
69 |
+
# You will be able to load one or the other configurations in the following list with
|
70 |
+
# data = datasets.load_dataset('my_dataset', 'first_domain')
|
71 |
+
# data = datasets.load_dataset('my_dataset', 'second_domain')
|
72 |
+
BUILDER_CONFIGS = [
|
73 |
+
ProofPile2Config(
|
74 |
+
name='default',
|
75 |
+
subsets=list(_DATA_PATHS.keys()),
|
76 |
+
version=VERSION,
|
77 |
+
description="All subsets"
|
78 |
+
),
|
79 |
+
ProofPile2Config(
|
80 |
+
name='arxiv',
|
81 |
+
subsets=["arxiv"],
|
82 |
+
version=VERSION,
|
83 |
+
description="ArXiv subset"
|
84 |
+
),
|
85 |
+
ProofPile2Config(
|
86 |
+
name='open-web-math',
|
87 |
+
subsets=['open-web-math'],
|
88 |
+
version=VERSION,
|
89 |
+
description="OpenWebMath"
|
90 |
+
),
|
91 |
+
ProofPile2Config(
|
92 |
+
name='algebraic-stack',
|
93 |
+
subsets=['algebraic-stack'],
|
94 |
+
version=VERSION,
|
95 |
+
description="Code subset"
|
96 |
+
),
|
97 |
+
]
|
98 |
+
|
99 |
+
|
100 |
+
def _info(self):
|
101 |
+
features = datasets.Features(
|
102 |
+
{
|
103 |
+
"text": datasets.Value("string"),
|
104 |
+
"meta": datasets.Value("string")
|
105 |
+
}
|
106 |
+
)
|
107 |
+
return datasets.DatasetInfo(
|
108 |
+
description=_DESCRIPTION,
|
109 |
+
features=features,
|
110 |
+
homepage=_HOMEPAGE,
|
111 |
+
citation=_CITATION,
|
112 |
+
)
|
113 |
+
|
114 |
+
def _split_generators(self, dl_manager):
|
115 |
+
return [
|
116 |
+
datasets.SplitGenerator(
|
117 |
+
name=datasets.Split.TRAIN,
|
118 |
+
# These kwargs will be passed to _generate_examples
|
119 |
+
gen_kwargs={
|
120 |
+
"data_files": list(map(
|
121 |
+
dl_manager.download,
|
122 |
+
[x for subset in self.config.subsets for x in _DATA_PATHS[subset]["train"]]
|
123 |
+
)),
|
124 |
+
},
|
125 |
+
),
|
126 |
+
datasets.SplitGenerator(
|
127 |
+
name=datasets.Split.VALIDATION,
|
128 |
+
# These kwargs will be passed to _generate_examples
|
129 |
+
gen_kwargs={
|
130 |
+
"data_files": list(map(
|
131 |
+
dl_manager.download,
|
132 |
+
[x for subset in self.config.subsets for x in _DATA_PATHS[subset]["validation"]]
|
133 |
+
)),
|
134 |
+
},
|
135 |
+
),
|
136 |
+
datasets.SplitGenerator(
|
137 |
+
name=datasets.Split.TEST,
|
138 |
+
gen_kwargs={
|
139 |
+
"data_files": list(map(
|
140 |
+
dl_manager.download,
|
141 |
+
[x for subset in self.config.subsets for x in _DATA_PATHS[subset]["test"]]
|
142 |
+
)),
|
143 |
+
},
|
144 |
+
),
|
145 |
+
]
|
146 |
+
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
147 |
+
def _generate_examples(self, data_files):
|
148 |
+
key = 0
|
149 |
+
for name in data_files:
|
150 |
+
with zstd.open(open(name, "rb"), "rt", encoding="utf-8") as f:
|
151 |
+
for x in f.readlines():
|
152 |
+
instance = json.loads(x)
|
153 |
+
if instance:
|
154 |
+
if "meta" not in instance:
|
155 |
+
instance["meta"] = dict()
|
156 |
+
yield key, {"text": instance["text"], "meta": json.dumps(instance["meta"])}
|
157 |
+
key += 1
|