Edward J. Schwartz commited on
Commit
f57ae1a
1 Parent(s): 221e108

Add first loading script that has configs for splits

Browse files
Files changed (1) hide show
  1. oo-method-test-split.py +112 -0
oo-method-test-split.py ADDED
@@ -0,0 +1,112 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/python
2
+
3
+ import datasets
4
+
5
+ import pyarrow as pa
6
+ import pyarrow.parquet as pq
7
+
8
+ BASE_DATASET = "ejschwartz/oo-method-test"
9
+
10
+ class OOMethodTestDataset(datasets.ArrowBasedBuilder):
11
+
12
+ BUILDER_CONFIGS = [
13
+ datasets.BuilderConfig(
14
+ name="combined",
15
+ version=datasets.Version("1.0.0"),
16
+ description="All data files combined",
17
+ ),
18
+ datasets.BuilderConfig(
19
+ name="byrow",
20
+ version=datasets.Version("1.0.0"),
21
+ description="Split by example (dumb)",
22
+ ),
23
+ datasets.BuilderConfig(
24
+ name="byfuncname",
25
+ version=datasets.Version("1.0.0"),
26
+ description="Split by function name",
27
+ )
28
+
29
+ ]
30
+
31
+ def __init__(self, *args, **kwargs):
32
+ super().__init__(*args, **kwargs)
33
+
34
+ def _info(self):
35
+ return datasets.DatasetInfo()
36
+
37
+ def _split_generators(self, dl_manager):
38
+ ds = datasets.load_dataset(BASE_DATASET)
39
+
40
+ #print(files)
41
+ #print(downloaded_files)
42
+
43
+ if self.config.name == "combined":
44
+
45
+ return [
46
+ datasets.SplitGenerator(
47
+ name="combined",
48
+ gen_kwargs={
49
+ "ds": ds['combined'],
50
+ },
51
+ ),
52
+ ]
53
+
54
+ elif self.config.name == "byrow":
55
+
56
+ ds = ds['combined'].train_test_split(test_size=0.1, seed=42)
57
+ #print(ds)
58
+
59
+ return [
60
+ datasets.SplitGenerator(
61
+ name="train",
62
+ gen_kwargs={
63
+ "ds": ds['train'],
64
+ },
65
+ ),
66
+ datasets.SplitGenerator(
67
+ name="test",
68
+ gen_kwargs={
69
+ "ds": ds['test'],
70
+ },
71
+ ),
72
+
73
+ ]
74
+
75
+ elif self.config.name == "byfuncname":
76
+
77
+ ds = ds['combined']
78
+
79
+ unique_names = ds.unique('Name')
80
+ nameds = datasets.Dataset.from_dict({'Name': unique_names})
81
+
82
+ name_split = nameds.train_test_split(test_size=0.1, seed=42)
83
+ #print(name_split)
84
+
85
+ train_name = name_split['train']['Name']
86
+ test_name = name_split['test']['Name']
87
+
88
+ return [
89
+ datasets.SplitGenerator(
90
+ name="train",
91
+ gen_kwargs={
92
+ "ds": ds.filter(lambda r: r['Name'] in train_name),
93
+ },
94
+ ),
95
+ datasets.SplitGenerator(
96
+ name="test",
97
+ gen_kwargs={
98
+ "ds": ds.filter(lambda r: r['Name'] in test_name),
99
+ },
100
+ ),
101
+
102
+ ]
103
+
104
+ else:
105
+ assert False
106
+
107
+ def _generate_tables(self, ds):
108
+
109
+ # Converting to pandas is silly, but the old version of datasets doesn't
110
+ # seem to have a way to convert to Arrow?
111
+ for i, batch in enumerate(ds.to_pandas(batched=True)):
112
+ yield i, pa.Table.from_pandas(batch)