Create bluemo.py
Browse files
bluemo.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# bluemo.py
|
2 |
+
import json, datasets
|
3 |
+
from datasets import Features, Value, Sequence, Split, SplitGenerator
|
4 |
+
|
5 |
+
_CITATION = ""
|
6 |
+
_DESCRIPTION = "BlueMO dataset."
|
7 |
+
|
8 |
+
class BlueMO(datasets.GeneratorBasedBuilder):
|
9 |
+
BUILDER_CONFIGS = [
|
10 |
+
datasets.BuilderConfig(name="default"),
|
11 |
+
datasets.BuilderConfig(name="proof"),
|
12 |
+
datasets.BuilderConfig(name="calculation"),
|
13 |
+
datasets.BuilderConfig(name="mathtext"),
|
14 |
+
]
|
15 |
+
|
16 |
+
def _info(self):
|
17 |
+
if self.config.name in ("proof", "calculation", "default"):
|
18 |
+
feats = Features({
|
19 |
+
"source_file": Value("string"),
|
20 |
+
"problem_type": Value("string"),
|
21 |
+
"problem": Value("string"),
|
22 |
+
"solution": Value("string"),
|
23 |
+
"remark": Value("string"),
|
24 |
+
"figures": Sequence(Value("string")),
|
25 |
+
})
|
26 |
+
else: # mathtext
|
27 |
+
feats = Features({
|
28 |
+
"source_file": Value("string"),
|
29 |
+
"text": Value("string"),
|
30 |
+
"figures": Sequence(Value("string")),
|
31 |
+
})
|
32 |
+
return datasets.DatasetInfo(description=_DESCRIPTION, citation=_CITATION, features=feats)
|
33 |
+
|
34 |
+
def _split_generators(self, dl_manager):
|
35 |
+
# match your README configs
|
36 |
+
if self.config.name == "proof":
|
37 |
+
paths = dl_manager.iter_files(["processed_dataset/proof"])
|
38 |
+
elif self.config.name == "calculation":
|
39 |
+
paths = dl_manager.iter_files(["processed_dataset/calculation"])
|
40 |
+
elif self.config.name == "mathtext":
|
41 |
+
paths = dl_manager.iter_files(["processed_dataset/text"])
|
42 |
+
else: # default = proof + calculation
|
43 |
+
paths = dl_manager.iter_files(["processed_dataset/proof", "processed_dataset/calculation"])
|
44 |
+
return [SplitGenerator(name=Split.TRAIN, gen_kwargs={"paths": list(paths)})]
|
45 |
+
|
46 |
+
def _generate_examples(self, paths):
|
47 |
+
for i, p in enumerate(sorted(paths)):
|
48 |
+
with open(p, "r", encoding="utf-8") as f:
|
49 |
+
obj = json.load(f)
|
50 |
+
# normalize keys & types
|
51 |
+
if "figures" in obj and obj["figures"] is None:
|
52 |
+
obj["figures"] = []
|
53 |
+
if "figures" in obj:
|
54 |
+
obj["figures"] = [str(x) for x in (obj["figures"] or [])]
|
55 |
+
yield str(i), obj
|