Datasets:
Upload latex-formulas.py
Browse files- latex-formulas.py +86 -0
latex-formulas.py
ADDED
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import datasets
|
2 |
+
import json
|
3 |
+
|
4 |
+
|
5 |
+
RAW_METADATA_URL = r'https://huggingface.co/datasets/OleehyO/latex-formulas/resolve/main/raw_formulas.jsonl'
|
6 |
+
|
7 |
+
FILTERED_METADATA_URL = r'https://huggingface.co/datasets/OleehyO/latex-formulas/resolve/main/formulas_finally.jsonl'
|
8 |
+
FILTERED_IMG_URL = r'https://huggingface.co/datasets/OleehyO/latex-formulas/resolve/main/formulas_pic.tar.gz'
|
9 |
+
|
10 |
+
|
11 |
+
class LatexFormulasConfig(datasets.BuilderConfig):
|
12 |
+
def __init__(self, img_archive_url, metadata_url, **kwargs):
|
13 |
+
super().__init__(**kwargs)
|
14 |
+
self.img_archive_url = img_archive_url
|
15 |
+
self.metadata_url = metadata_url
|
16 |
+
|
17 |
+
|
18 |
+
class LatexFormulas(datasets.GeneratorBasedBuilder):
|
19 |
+
BUILDER_CONFIGS = [
|
20 |
+
LatexFormulasConfig(
|
21 |
+
name="raw_formulas",
|
22 |
+
img_archive_url=None,
|
23 |
+
metadata_url=RAW_METADATA_URL
|
24 |
+
),
|
25 |
+
LatexFormulasConfig(
|
26 |
+
name="filtered_formulas",
|
27 |
+
img_archive_url=FILTERED_IMG_URL,
|
28 |
+
metadata_url=FILTERED_METADATA_URL
|
29 |
+
)
|
30 |
+
]
|
31 |
+
|
32 |
+
def _info(self):
|
33 |
+
if self.config.name == "raw_formulas":
|
34 |
+
return datasets.DatasetInfo(
|
35 |
+
features=datasets.Features({
|
36 |
+
"latex_formula": datasets.Value("string")
|
37 |
+
})
|
38 |
+
)
|
39 |
+
if self.config.name == "filtered_formulas":
|
40 |
+
return datasets.DatasetInfo(
|
41 |
+
features=datasets.Features({
|
42 |
+
"image": datasets.Image(),
|
43 |
+
"latex_formula": datasets.Value("string")
|
44 |
+
})
|
45 |
+
)
|
46 |
+
|
47 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager):
|
48 |
+
metadata_path = dl_manager.download(self.config.metadata_url)
|
49 |
+
|
50 |
+
if self.config.name == "filtered_formulas":
|
51 |
+
image_archive_path = self.config.img_archive_url
|
52 |
+
images = dl_manager.iter_archive(image_archive_path)
|
53 |
+
elif self.config.name == "raw_formulas":
|
54 |
+
images = None
|
55 |
+
|
56 |
+
return [
|
57 |
+
datasets.SplitGenerator(
|
58 |
+
name=datasets.Split.TRAIN,
|
59 |
+
gen_kwargs={
|
60 |
+
"images": images,
|
61 |
+
"metadata_path": metadata_path
|
62 |
+
}
|
63 |
+
)
|
64 |
+
]
|
65 |
+
|
66 |
+
def _generate_examples(self, images, metadata_path):
|
67 |
+
if images is not None:
|
68 |
+
img_formula_pair = {}
|
69 |
+
with open(metadata_path, 'r', encoding="utf-8") as f:
|
70 |
+
for line in f:
|
71 |
+
single_json = json.loads(line)
|
72 |
+
img_formula_pair[single_json["id"]] = single_json["formula"]
|
73 |
+
|
74 |
+
for img_path, img_obj in images:
|
75 |
+
img_name = img_path.split('/')[-1]
|
76 |
+
if img_name in img_formula_pair:
|
77 |
+
yield img_path, {
|
78 |
+
"image": {"path": img_path, "bytes": img_obj.read()},
|
79 |
+
"latex_formula": img_formula_pair[img_name]
|
80 |
+
}
|
81 |
+
else:
|
82 |
+
with open(metadata_path, 'r', encoding="utf-8") as f:
|
83 |
+
for idx, line in enumerate(f):
|
84 |
+
yield idx, {
|
85 |
+
"latex_formula": json.loads(line)["formula"]
|
86 |
+
}
|