Convert dataset to Parquet (part 00004-of-00005)

#6
by LoMicha - opened
train_real_simple_io.tar.gz → ExeBench/test_real-00000-of-00001.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5717bb6c663ab3fe039208511fb8e52e3db7b45b7d357d3e4ca8d027b715b19e
3
- size 35982092
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:151dcd2783232480dc8c202d91c95cf99dbba525039ece5ca7c4194b5fd5835b
3
+ size 18974256
test_real.tar.gz → ExeBench/test_synth-00000-of-00001.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:39f63f236a869eb853e2704d6be7c062d35d1e47df7d9b8c777fb30c63c70632
3
- size 9241028
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:962e9ccb2a705c1728a8c3e211108962e2cb3614f401009699e58b0e655d56cf
3
+ size 33733873
test_synth.tar.gz → ExeBench/train_synth_rich_io-00008-of-00009.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:027a3a1671154ec76f78bc7f76147c22055efacbbf6c443a9644b7668e3ed8df
3
- size 16016669
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67fb35baf4d54e645212a93ae4c9c2a8dc4daf60d18896363b4f0292a815b8f1
3
+ size 77576785
valid_real.tar.gz → ExeBench/valid_real-00000-of-00001.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c06049a4e17193958cb9b67129cfbcf028fb12b1b24133333b174bcbc1cad3a7
3
- size 13382934
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc63d33157d76fed3e9a62ada37d0d251bd80951fc88352603acf5a6ea74038d
3
+ size 28423343
ExeBench/valid_synth-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb13f5313bd38b62b18d9c734673e78c57679be5e7a863146f1e26fa9c1b3f48
3
+ size 35869868
README.md CHANGED
@@ -1,3 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  # ExeBench: an ML-scale dataset of executable C functions
2
 
3
  ExeBench is a dataset of millions of C functions paired with dependencies and metadatada such that at least a subset of it can be executed with IO pairs. It is mainly inteded for machine learning applications but it is application-agnostic enough to have other usages.
 
1
+ ---
2
+ dataset_info:
3
+ config_name: ExeBench
4
+ features:
5
+ - name: path
6
+ dtype: string
7
+ - name: func_def
8
+ dtype: string
9
+ - name: func_head
10
+ dtype: string
11
+ - name: func_head_types
12
+ dtype: string
13
+ - name: fname
14
+ dtype: string
15
+ - name: signature
16
+ sequence: string
17
+ - name: asm
18
+ sequence:
19
+ - name: target
20
+ dtype: string
21
+ - name: code
22
+ dtype: string
23
+ - name: synth_deps
24
+ dtype: string
25
+ - name: real_deps
26
+ dtype: string
27
+ - name: synth_io_pairs
28
+ sequence:
29
+ - name: input
30
+ sequence:
31
+ - name: var
32
+ dtype: string
33
+ - name: value
34
+ dtype: string
35
+ - name: output
36
+ sequence:
37
+ - name: var
38
+ dtype: string
39
+ - name: value
40
+ dtype: string
41
+ - name: dummy_funcs
42
+ dtype: string
43
+ - name: dummy_funcs_seed
44
+ dtype: int64
45
+ - name: real_io_pairs
46
+ sequence:
47
+ - name: input
48
+ sequence:
49
+ - name: var
50
+ dtype: string
51
+ - name: value
52
+ dtype: string
53
+ - name: output
54
+ sequence:
55
+ - name: var
56
+ dtype: string
57
+ - name: value
58
+ dtype: string
59
+ - name: dummy_funcs
60
+ dtype: string
61
+ - name: dummy_funcs_seed
62
+ dtype: int64
63
+ - name: synth_exe_wrapper
64
+ dtype: string
65
+ - name: real_exe_wrapper
66
+ dtype: string
67
+ - name: ref
68
+ dtype: string
69
+ - name: synth_iospec
70
+ dtype: string
71
+ - name: real_iospec
72
+ dtype: string
73
+ splits:
74
+ - name: train_not_compilable
75
+ num_bytes: 2059245891
76
+ num_examples: 715700
77
+ - name: train_synth_compilable
78
+ num_bytes: 60287736644
79
+ num_examples: 4159373
80
+ - name: train_real_compilable
81
+ num_bytes: 11988407074
82
+ num_examples: 885074
83
+ - name: train_synth_simple_io
84
+ num_bytes: 20089650210
85
+ num_examples: 640116
86
+ - name: train_real_simple_io
87
+ num_bytes: 266865540
88
+ num_examples: 42769
89
+ - name: train_synth_rich_io
90
+ num_bytes: 4358613407
91
+ num_examples: 107250
92
+ - name: valid_synth
93
+ num_bytes: 207011272
94
+ num_examples: 5000
95
+ - name: valid_real
96
+ num_bytes: 55126948
97
+ num_examples: 2133
98
+ - name: test_synth
99
+ num_bytes: 192119154
100
+ num_examples: 5000
101
+ - name: test_real
102
+ num_bytes: 41191987
103
+ num_examples: 2134
104
+ download_size: 18777417895
105
+ dataset_size: 99545968127
106
+ configs:
107
+ - config_name: ExeBench
108
+ data_files:
109
+ - split: train_not_compilable
110
+ path: ExeBench/train_not_compilable-*
111
+ - split: train_synth_compilable
112
+ path: ExeBench/train_synth_compilable-*
113
+ - split: train_real_compilable
114
+ path: ExeBench/train_real_compilable-*
115
+ - split: train_synth_simple_io
116
+ path: ExeBench/train_synth_simple_io-*
117
+ - split: train_real_simple_io
118
+ path: ExeBench/train_real_simple_io-*
119
+ - split: train_synth_rich_io
120
+ path: ExeBench/train_synth_rich_io-*
121
+ - split: valid_synth
122
+ path: ExeBench/valid_synth-*
123
+ - split: valid_real
124
+ path: ExeBench/valid_real-*
125
+ - split: test_synth
126
+ path: ExeBench/test_synth-*
127
+ - split: test_real
128
+ path: ExeBench/test_real-*
129
+ default: true
130
+ ---
131
  # ExeBench: an ML-scale dataset of executable C functions
132
 
133
  ExeBench is a dataset of millions of C functions paired with dependencies and metadatada such that at least a subset of it can be executed with IO pairs. It is mainly inteded for machine learning applications but it is application-agnostic enough to have other usages.
exebench.py DELETED
@@ -1,223 +0,0 @@
1
- # coding=utf-8
2
- # Copyright 2022 ExeBench authors
3
- # The code required to produce and load this dataset is licensed under MIT License.
4
- # The code samples included in this dataset keep their own licenses, which can be retrieved via their metadata.
5
- # Unless required by applicable law or agreed to in writing, software
6
- # distributed under the License is distributed on an "AS IS" BASIS,
7
- # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
8
- # See the License for the specific language governing permissions and
9
- # limitations under the License.
10
-
11
- # Please note that the dataset release is still work in progress.
12
-
13
- """The ExeBench dataset."""
14
-
15
- import json
16
-
17
- import datasets
18
-
19
- from pathlib import Path
20
-
21
-
22
- _CITATION = """\
23
- @inproceedings{10.1145/3520312.3534867,
24
- author = {Armengol-Estap\'{e}, Jordi and Woodruff, Jackson and Brauckmann, Alexander and Magalh\~{a}es, Jos\'{e} Wesley de Souza and O'Boyle, Michael F. P.},
25
- title = {ExeBench: An ML-Scale Dataset of Executable C Functions},
26
- year = {2022},
27
- isbn = {9781450392730},
28
- publisher = {Association for Computing Machinery},
29
- address = {New York, NY, USA},
30
- url = {https://doi.org/10.1145/3520312.3534867},
31
- doi = {10.1145/3520312.3534867},
32
- abstract = {Machine-learning promises to transform compilation and software engineering, yet is frequently limited by the scope of available datasets. In particular, there is a lack of runnable, real-world datasets required for a range of tasks ranging from neural program synthesis to machine learning-guided program optimization. We introduce a new dataset, ExeBench, which attempts to address this. It tackles two key issues with real-world code: references to external types and functions and scalable generation of IO examples. ExeBench is the first publicly available dataset that pairs real-world C code taken from GitHub with IO examples that allow these programs to be run. We develop a toolchain that scrapes GitHub, analyzes the code, and generates runnable snippets of code. We analyze our benchmark suite using several metrics, and show it is representative of real-world code. ExeBench contains 4.5M compilable and 700k executable C functions. This scale of executable, real functions will enable the next generation of machine learning-based programming tasks.},
33
- booktitle = {Proceedings of the 6th ACM SIGPLAN International Symposium on Machine Programming},
34
- pages = {50–59},
35
- numpages = {10},
36
- keywords = {Code Dataset, Program Synthesis, Mining Software Repositories, C, Machine Learning for Code, Compilers},
37
- location = {San Diego, CA, USA},
38
- series = {MAPS 2022}
39
- }
40
- """
41
-
42
- _DESCRIPTION = """\
43
- An ML-scale dataset of executable C functions
44
- """ # TODO: expand
45
-
46
- _HOMEPAGE = "https://github.com/jordiae/exebench"
47
-
48
- _LICENSE = "Multiple: see each function license (fields 'ref' and 'path')"
49
-
50
- _URL = "" # "https://huggingface.co/datasets/jordiae/exebench-test/resolve/main/"
51
-
52
- _REMOVED_FEATURES = ["doc", "angha_error", "real_error", "angha_io_error", "real_io_error",
53
- "angha_io_pairs_are_trivial", "real_io_pairs_are_trivial"]
54
-
55
- _RENAMED_FEATURES = {"angha_deps": "synth_deps", "angha_io_pairs": "synth_io_pairs",
56
- "angha_exe_wrapper": "synth_exe_wrapper", "angha_iospec": "synth_iospec"}
57
-
58
- _FEATURES = datasets.Features(
59
- {
60
- "path": datasets.Value("string"),
61
- "func_def": datasets.Value("string"),
62
- "func_head": datasets.Value("string"),
63
- "func_head_types": datasets.Value("string"),
64
- "fname": datasets.Value("string"),
65
- "signature": datasets.Sequence(datasets.Value("string")),
66
- # "doc": datasets.Value("string"),
67
- # "angha_error": datasets.Value("string"),
68
- # "real_error": datasets.Value("string"),
69
- "asm": datasets.Sequence({'target': datasets.Value("string"), 'code': datasets.Value("string")}), # unflat dict#Optional[Dict[str, Optional[FuncAsm]]] = None
70
- "synth_deps": datasets.Value("string"),
71
- "real_deps": datasets.Value("string"),
72
- "synth_io_pairs": datasets.Sequence({
73
- "input": datasets.Sequence({'var': datasets.Value("string"), 'value': datasets.Value("string")}),
74
- "output": datasets.Sequence({'var': datasets.Value("string"), 'value': datasets.Value("string")}),
75
- "dummy_funcs": datasets.Value("string"),
76
- "dummy_funcs_seed": datasets.Value("int64")
77
- }),
78
- "real_io_pairs": datasets.Sequence({
79
- "input": datasets.Sequence({'var': datasets.Value("string"), 'value': datasets.Value("string")}),
80
- "output": datasets.Sequence({'var': datasets.Value("string"), 'value': datasets.Value("string")}),
81
- "dummy_funcs": datasets.Value("string"),
82
- "dummy_funcs_seed": datasets.Value("int64")
83
- }),
84
- # "angha_io_error": datasets.Value("string"),
85
- # "real_io_error": datasets.Value("string"),
86
- "synth_exe_wrapper": datasets.Value("string"),
87
- "real_exe_wrapper": datasets.Value("string"),
88
- # "angha_io_pairs_are_trivial": datasets.Value("bool"),
89
- # "real_io_pairs_are_trivial": datasets.Value("bool"),
90
- "ref": datasets.Value("string"),
91
- "synth_iospec": datasets.Value("string"), # serialized, TODO: improve
92
- "real_iospec": datasets.Value("string")
93
- }
94
- )
95
-
96
-
97
- class ExeBenchConfig(datasets.BuilderConfig):
98
- """BuilderConfig for ExeBench."""
99
-
100
- def __init__(self, *args, **kwargs):
101
- """BuilderConfig for The Pile.
102
- Args:
103
- **kwargs: keyword arguments forwarded to super.
104
- """
105
- super().__init__(
106
- *args,
107
- **kwargs,
108
- )
109
-
110
-
111
- class ExeBench(datasets.GeneratorBasedBuilder):
112
- """ExeBench dataset"""
113
-
114
- BUILDER_CONFIGS = [
115
- ExeBenchConfig(
116
- name="ExeBench",
117
- version=datasets.Version("1.0.4"),
118
- description="Executable C dataset"
119
- ),
120
- ]
121
-
122
- def _info(self):
123
- """Give information and typings for the dataset."""
124
- return datasets.DatasetInfo(
125
- # This is the description that will appear on the datasets page.
126
- description=_DESCRIPTION,
127
- # This defines the different columns of the dataset and their types
128
- features=_FEATURES,
129
- # If there's a common (input, target) tuple from the features,
130
- # specify them here. They'll be used if as_supervised=True in
131
- # builder.as_dataset.
132
- supervised_keys=None,
133
- # Homepage of the dataset for documentation
134
- homepage=_HOMEPAGE,
135
- # License for the dataset if available
136
- license=_LICENSE,
137
- # Citation for the dataset
138
- citation=_CITATION,
139
- )
140
-
141
- def _split_generators(self, dl_manager):
142
- """Returns SplitGenerators."""
143
- urls_to_download = {
144
- "train_not_compilable": f"{_URL}train_no_compilable.tar.gz",
145
- "train_synth_compilable": f"{_URL}train_synth_compilable.tar.gz",
146
- "train_real_compilable": f"{_URL}train_real_compilable.tar.gz",
147
- "train_synth_simple_io": f"{_URL}train_synth_simple_io.tar.gz",
148
- "train_real_simple_io": f"{_URL}train_real_simple_io.tar.gz",
149
- "train_synth_rich_io": f"{_URL}train_synth_rich_io.tar.gz",
150
- "valid_synth": f"{_URL}valid_synth.tar.gz",
151
- "valid_real": f"{_URL}valid_real.tar.gz",
152
- "test_synth": f"{_URL}test_synth.tar.gz",
153
- "test_real": f"{_URL}test_real.tar.gz",
154
- }
155
- downloaded_files = dl_manager.download_and_extract(urls_to_download)
156
-
157
- return [
158
- datasets.SplitGenerator(name='train_not_compilable',
159
- gen_kwargs={"files": downloaded_files["train_not_compilable"]}),
160
- datasets.SplitGenerator(name='train_synth_compilable',
161
- gen_kwargs={"files": downloaded_files["train_synth_compilable"]}),
162
- datasets.SplitGenerator(name='train_real_compilable',
163
- gen_kwargs={"files": downloaded_files["train_real_compilable"]}),
164
- datasets.SplitGenerator(name='train_synth_simple_io',
165
- gen_kwargs={"files": downloaded_files["train_synth_simple_io"]}),
166
- datasets.SplitGenerator(name='train_real_simple_io',
167
- gen_kwargs={"files": downloaded_files["train_real_simple_io"]}),
168
- datasets.SplitGenerator(name='train_synth_rich_io',
169
- gen_kwargs={"files": downloaded_files["train_synth_rich_io"]}),
170
- datasets.SplitGenerator(name='valid_synth',
171
- gen_kwargs={"files": downloaded_files["valid_synth"]}),
172
- datasets.SplitGenerator(name='valid_real',
173
- gen_kwargs={"files": downloaded_files["valid_real"]}),
174
- datasets.SplitGenerator(name='test_synth',
175
- gen_kwargs={"files": downloaded_files["test_synth"]}),
176
- datasets.SplitGenerator(name='test_real',
177
- gen_kwargs={"files": downloaded_files["test_real"]}),
178
- ]
179
-
180
- def _generate_examples(self, files):
181
- """Yield examples as (key, example) tuples."""
182
- key = 0
183
- import zstandard as zstd
184
-
185
- for path in Path(files).rglob('*.jsonl.zst'):
186
- with zstd.open(open(path, "rb"), "rt", encoding="utf-8") as f:
187
- for row in f:
188
- data = json.loads(row)
189
- data = data['text']
190
- data = self._fixes(data)
191
- for io_pairs_kind in ('synth_io_pairs', 'real_io_pairs'):
192
- if data[io_pairs_kind]:
193
- new_io_pairs = []
194
- for e in data[io_pairs_kind]:
195
- new_e = {}
196
- new_e['input'] = [{'var': var, 'value': json.dumps(value)} for (var, value) in e['input'].items()] if e['input'] else []
197
- new_e['output'] = [{'var': var, 'value': json.dumps(value)} for (var, value) in e['output'].items()] if e['output'] else []
198
- new_e['dummy_funcs'] = e['dummy_funcs']
199
- new_e['dummy_funcs_seed'] = e['dummy_funcs_seed']
200
- new_io_pairs.append(new_e)
201
- data[io_pairs_kind] = new_io_pairs
202
- data['synth_iospec'] = json.dumps(data['synth_iospec'])
203
- data['real_iospec'] = json.dumps(data['real_iospec'])
204
- yield key, data
205
- key += 1
206
-
207
- def _fixes(self, row):
208
- if 'angha_iospec' not in row:
209
- row['angha_iospec'] = None
210
- if 'real_iospec' not in row:
211
- row['real_iospec'] = None
212
- if 'func_head_types' not in row:
213
- row['func_head_types'] = ''
214
- row['asm'] = [{'target': target, 'code': code['func_asm'] if code else None} for (target, code) in
215
- row['asm'].items()] # TODO: pre_asm etc
216
- for removed_key in _REMOVED_FEATURES:
217
- if removed_key in row:
218
- del row[removed_key]
219
- for original_key, new_key in _RENAMED_FEATURES.items():
220
- row[new_key] = row[original_key]
221
- del row[original_key]
222
- return row
223
-
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
train_no_compilable.tar.gz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:c182b4c62dd8793e42fa33e34151053c9fa6ec12ede9145661b046e29182ede5
3
- size 465629449
 
 
 
 
train_real_compilable.tar.gz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:3dee594ee4cf70fe2e35eca6ee395157458a8d78dce8ddc1747b864d3b4d8326
3
- size 1574773269
 
 
 
 
train_synth_compilable.tar.gz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:5c5396923f14b56bd7d0481f3a84be3349e8aa36c66e62174cccd557408ff8cc
3
- size 8263470418
 
 
 
 
train_synth_rich_io.tar.gz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:a3871d931615f852eb32213946f215a03d04f6741efe3b961805d8988715bc9f
3
- size 368732571
 
 
 
 
train_synth_simple_io.tar.gz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:dc1a819467a4814ff6fce0f6def277046ba4b485cf36bec079f55471b4b09118
3
- size 1296875463
 
 
 
 
valid_synth.tar.gz DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:5824f4d0eabc89fdac44a050fdb43584f426b97f62b38f24284e54c0f80bb396
3
- size 16583586