udpated configuration. Test ok
Browse files- Readme.md +78 -3
- SynStOp.py +76 -54
- usage.py +0 -16
Readme.md
CHANGED
@@ -29,18 +29,93 @@ dataset_info:
|
|
29 |
- name: input
|
30 |
dtype: string
|
31 |
- name: output
|
32 |
-
dtype: string
|
33 |
- name: code
|
34 |
-
dtype: string
|
35 |
- name: res_var
|
36 |
dtype: string
|
37 |
- name: operation
|
38 |
dtype: string
|
|
|
|
|
39 |
splits:
|
|
|
|
|
|
|
40 |
- name: test
|
|
|
41 |
num_examples: 14661
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
- name: train
|
43 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
44 |
---
|
45 |
|
46 |
# Dataset Card for Small String Operations Dataset
|
|
|
29 |
- name: input
|
30 |
dtype: string
|
31 |
- name: output
|
32 |
+
dtype: string
|
33 |
- name: code
|
34 |
+
dtype: string
|
35 |
- name: res_var
|
36 |
dtype: string
|
37 |
- name: operation
|
38 |
dtype: string
|
39 |
+
- name: id
|
40 |
+
dtype: int32
|
41 |
splits:
|
42 |
+
- name: train
|
43 |
+
num_bytes: 3222948
|
44 |
+
num_examples: 33939
|
45 |
- name: test
|
46 |
+
num_bytes: 1392252
|
47 |
num_examples: 14661
|
48 |
+
download_size: 1178254
|
49 |
+
dataset_size: 4615200
|
50 |
+
- config_name: small10
|
51 |
+
features:
|
52 |
+
- name: input
|
53 |
+
dtype: string
|
54 |
+
- name: output
|
55 |
+
dtype: string
|
56 |
+
- name: code
|
57 |
+
dtype: string
|
58 |
+
- name: res_var
|
59 |
+
dtype: string
|
60 |
+
- name: operation
|
61 |
+
dtype: string
|
62 |
+
- name: id
|
63 |
+
dtype: int32
|
64 |
+
splits:
|
65 |
- name: train
|
66 |
+
num_bytes: 956996
|
67 |
+
num_examples: 11313
|
68 |
+
- name: test
|
69 |
+
num_bytes: 413404
|
70 |
+
num_examples: 4887
|
71 |
+
download_size: 312419
|
72 |
+
dataset_size: 1370400
|
73 |
+
- config_name: small15
|
74 |
+
features:
|
75 |
+
- name: input
|
76 |
+
dtype: string
|
77 |
+
- name: output
|
78 |
+
dtype: string
|
79 |
+
- name: code
|
80 |
+
dtype: string
|
81 |
+
- name: res_var
|
82 |
+
dtype: string
|
83 |
+
- name: operation
|
84 |
+
dtype: string
|
85 |
+
- name: id
|
86 |
+
dtype: int32
|
87 |
+
splits:
|
88 |
+
- name: train
|
89 |
+
num_bytes: 1074316
|
90 |
+
num_examples: 11313
|
91 |
+
- name: test
|
92 |
+
num_bytes: 464084
|
93 |
+
num_examples: 4887
|
94 |
+
download_size: 393420
|
95 |
+
dataset_size: 1538400
|
96 |
+
- config_name: small20
|
97 |
+
features:
|
98 |
+
- name: input
|
99 |
+
dtype: string
|
100 |
+
- name: output
|
101 |
+
dtype: string
|
102 |
+
- name: code
|
103 |
+
dtype: string
|
104 |
+
- name: res_var
|
105 |
+
dtype: string
|
106 |
+
- name: operation
|
107 |
+
dtype: string
|
108 |
+
- name: id
|
109 |
+
dtype: int32
|
110 |
+
splits:
|
111 |
+
- name: train
|
112 |
+
num_bytes: 1191636
|
113 |
+
num_examples: 11313
|
114 |
+
- name: test
|
115 |
+
num_bytes: 514764
|
116 |
+
num_examples: 4887
|
117 |
+
download_size: 472415
|
118 |
+
dataset_size: 1706400
|
119 |
---
|
120 |
|
121 |
# Dataset Card for Small String Operations Dataset
|
SynStOp.py
CHANGED
@@ -35,11 +35,42 @@ year={2023}
|
|
35 |
# TODO: Add description of the dataset here
|
36 |
# You can copy an official description
|
37 |
_DESCRIPTION = """\
|
38 |
-
Minimal dataset for intended for LM development and testing using python string operations.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
"""
|
40 |
|
41 |
# TODO: Add a link to an official homepage for the dataset here
|
42 |
-
_HOMEPAGE = ""
|
43 |
|
44 |
# TODO: Add the licence for the dataset here if you can find it
|
45 |
_LICENSE = "Apache 2.0 License"
|
@@ -48,19 +79,44 @@ _LICENSE = "Apache 2.0 License"
|
|
48 |
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
|
49 |
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
50 |
_URLS = {
|
51 |
-
"small":
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
}
|
60 |
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
# TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case
|
63 |
-
class
|
64 |
"""TODO: Short description of my dataset."""
|
65 |
|
66 |
VERSION = datasets.Version("0.0.1")
|
@@ -72,31 +128,24 @@ class StopDataset(datasets.GeneratorBasedBuilder):
|
|
72 |
# You will be able to load one or the other configurations in the following list with
|
73 |
# data = datasets.load_dataset('my_dataset', 'small')
|
74 |
# data = datasets.load_dataset('my_dataset', 'second_domain')
|
75 |
-
BUILDER_CONFIGS = [
|
76 |
-
|
77 |
-
datasets.BuilderConfig(name="small[filter]", version=VERSION, description="Small string operations dataset with string slices only. [] allows to specify a comma separated list of filters on the length (i.e. l=X) and operations (i.e. o=y)"),
|
78 |
-
|
79 |
-
]
|
80 |
|
81 |
DEFAULT_CONFIG_NAME = "small" # It's not mandatory to have a default configuration. Just use one if it make sense.
|
82 |
|
83 |
def _info(self):
|
84 |
# TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
|
85 |
-
|
86 |
-
features = datasets.Features(
|
87 |
{
|
88 |
"input": datasets.Value("string"),
|
89 |
"output": datasets.Value("string"),
|
90 |
"code": datasets.Value("string"),
|
91 |
"res_var": datasets.Value("string"),
|
92 |
-
"operation": datasets.Value("string")
|
|
|
93 |
# These are the features of your dataset like images, labels ...
|
94 |
}
|
95 |
)
|
96 |
-
self._init_filters(self.config.name[len("small"):].strip("[]").split(","))
|
97 |
-
self.config.name= self.config.name[:len("small")]
|
98 |
-
else:
|
99 |
-
raise NotImplementedError()
|
100 |
return datasets.DatasetInfo(
|
101 |
# This is the description that will appear on the datasets page.
|
102 |
description=_DESCRIPTION,
|
@@ -112,18 +161,6 @@ class StopDataset(datasets.GeneratorBasedBuilder):
|
|
112 |
# Citation for the dataset
|
113 |
citation=_CITATION,
|
114 |
)
|
115 |
-
|
116 |
-
def _init_filters(self, filters):
|
117 |
-
self.filter_operations = []
|
118 |
-
self.filter_len = []
|
119 |
-
for filter in filters:
|
120 |
-
if filter =="": continue
|
121 |
-
k, v = filter.split("=")
|
122 |
-
if k=="l":
|
123 |
-
self.filter_len.append(int(v))
|
124 |
-
elif k=="o":
|
125 |
-
self.filter_operations.append(re.compile(v))
|
126 |
-
|
127 |
def _split_generators(self, dl_manager):
|
128 |
# TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration
|
129 |
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
|
@@ -131,16 +168,14 @@ class StopDataset(datasets.GeneratorBasedBuilder):
|
|
131 |
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
|
132 |
# It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
|
133 |
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
134 |
-
urls =
|
135 |
-
if len(self.filter_len)>0:
|
136 |
-
urls = [url for url in urls if any([f"stop_{str(len)}_t" in url for len in self.filter_len])]
|
137 |
data_dir = dl_manager.download_and_extract(urls)
|
138 |
return [
|
139 |
datasets.SplitGenerator(
|
140 |
name=datasets.Split.TRAIN,
|
141 |
# These kwargs will be passed to _generate_examples
|
142 |
gen_kwargs={
|
143 |
-
"filepath":
|
144 |
"split": "train",
|
145 |
},
|
146 |
),
|
@@ -149,22 +184,12 @@ class StopDataset(datasets.GeneratorBasedBuilder):
|
|
149 |
name=datasets.Split.TEST,
|
150 |
# These kwargs will be passed to _generate_examples
|
151 |
gen_kwargs={
|
152 |
-
"filepath":
|
153 |
"split": "test",
|
154 |
},
|
155 |
),
|
156 |
]
|
157 |
|
158 |
-
def _match_operations_filter(self, operation):
|
159 |
-
if self.filter_operations is not None:
|
160 |
-
matches = False
|
161 |
-
for filter in self.filter_operations:
|
162 |
-
if filter.matches(operation):
|
163 |
-
matches = True
|
164 |
-
break
|
165 |
-
return matches
|
166 |
-
else: return True
|
167 |
-
|
168 |
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
169 |
def _generate_examples(self, filepath, split):
|
170 |
# TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
|
@@ -176,10 +201,6 @@ class StopDataset(datasets.GeneratorBasedBuilder):
|
|
176 |
for ix, data in enumerate(dataset):
|
177 |
|
178 |
if self.config.name.startswith("small"):
|
179 |
-
|
180 |
-
if self._match_operations_filter(data["operation"]):
|
181 |
-
continue
|
182 |
-
|
183 |
# Yields examples as (key, example) tuples
|
184 |
id = data["id"] if "id" in data else count
|
185 |
count = count + 1
|
@@ -188,6 +209,7 @@ class StopDataset(datasets.GeneratorBasedBuilder):
|
|
188 |
"output": data["output"],
|
189 |
"code": data["code"],
|
190 |
"res_var": data["res_var"],
|
|
|
191 |
"operation": data["operation"]
|
192 |
}
|
193 |
else:
|
|
|
35 |
# TODO: Add description of the dataset here
|
36 |
# You can copy an official description
|
37 |
_DESCRIPTION = """\
|
38 |
+
Minimal dataset for intended for LM development and testing using python string operations.
|
39 |
+
The dataset is created by running different one line python string operations on random strings
|
40 |
+
The idea is, that transformer implementation can learn the string operations and that this task is a good
|
41 |
+
proxy tasks for other transformer operations on real languages and real tasks. Consequently, the
|
42 |
+
data set is small and can be used in the development process without large scale infrastructures.
|
43 |
+
|
44 |
+
There are different configurations for the data set.
|
45 |
+
|
46 |
+
- `small`: contains below 50k instances of various string length and only contains slicing operations, i.e. all python operations expressable with `s[i:j:s]` (which also includes string reversal).
|
47 |
+
- you can further choose different subsets according to either length or the kind of operation
|
48 |
+
- `small10`: like small, but only strings to length 10
|
49 |
+
- `small15`: like small, but only strings to length 15
|
50 |
+
- `small20`: like small, but only strings to length 20
|
51 |
+
|
52 |
+
The fields have the following meaning:
|
53 |
+
|
54 |
+
- `input`: input string, i.e. the string and the string operation
|
55 |
+
- `output`: output of the string operation
|
56 |
+
- `code`: code for running the string operation in python,
|
57 |
+
- `res_var`: name of the result variable
|
58 |
+
- `operation`: kind of operation:
|
59 |
+
- `step_x` for `s[::x]`
|
60 |
+
- `char_at_x` for `s[x]`
|
61 |
+
- `slice_x:y` for `s[x:y]`
|
62 |
+
- `slice_step_x:y:z` for `s[x:y:z]`
|
63 |
+
- `slice_reverse_i:j:k` for `s[i:i+j][::k]`
|
64 |
+
|
65 |
+
Siblings of `data` contain additional metadata information about the dataset.
|
66 |
+
|
67 |
+
- `prompt` describes possible prompts based on that data splitted into input prompts / output prompts
|
68 |
+
|
69 |
+
|
70 |
"""
|
71 |
|
72 |
# TODO: Add a link to an official homepage for the dataset here
|
73 |
+
_HOMEPAGE = "https://huggingface.co/PaDaS-Lab/SynStOp"
|
74 |
|
75 |
# TODO: Add the licence for the dataset here if you can find it
|
76 |
_LICENSE = "Apache 2.0 License"
|
|
|
79 |
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
|
80 |
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
81 |
_URLS = {
|
82 |
+
"small": {
|
83 |
+
"train": ["./small/stop_10_train.json.gz", "./small/stop_20_train.json.gz", "./small/stop_15_train.json.gz",],
|
84 |
+
"test": ["./small/stop_10_test.json.gz", "./small/stop_20_test.json.gz", "./small/stop_15_test.json.gz",]
|
85 |
+
},
|
86 |
+
"small15": {
|
87 |
+
"train": [ "./small/stop_15_train.json.gz",],
|
88 |
+
"test": [ "./small/stop_15_test.json.gz",]
|
89 |
+
},
|
90 |
+
"small10": {
|
91 |
+
"train": ["./small/stop_10_train.json.gz"],
|
92 |
+
"test": ["./small/stop_10_test.json.gz"]
|
93 |
+
},
|
94 |
+
"small20": {
|
95 |
+
"train": [ "./small/stop_20_train.json.gz"],
|
96 |
+
"test": [ "./small/stop_20_test.json.gz"]
|
97 |
+
}
|
98 |
}
|
99 |
|
100 |
|
101 |
+
class SynStOpDatasetConfig(datasets.BuilderConfig):
|
102 |
+
|
103 |
+
def __init__(self, subset="small", length=(10,15,20), **kwargs):
|
104 |
+
"""BuilderConfig for SynStOpDatasetConfig.
|
105 |
+
Args:
|
106 |
+
**kwargs: keyword arguments forwarded to super.
|
107 |
+
"""
|
108 |
+
super(SynStOpDatasetConfig, self).__init__(**kwargs)
|
109 |
+
self.subset = subset
|
110 |
+
self.length = length
|
111 |
+
self.files = {
|
112 |
+
"train": ["./{subset}".format(subset=subset) + "/stop_{length}_train.json.gz".format(length=length) for length in length],
|
113 |
+
"test": ["./{subset}".format(subset=subset) + "/stop_{length}_test.json.gz".format(length=length) for length in length],
|
114 |
+
}
|
115 |
+
|
116 |
+
|
117 |
+
|
118 |
# TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case
|
119 |
+
class SynStOpDataset(datasets.GeneratorBasedBuilder):
|
120 |
"""TODO: Short description of my dataset."""
|
121 |
|
122 |
VERSION = datasets.Version("0.0.1")
|
|
|
128 |
# You will be able to load one or the other configurations in the following list with
|
129 |
# data = datasets.load_dataset('my_dataset', 'small')
|
130 |
# data = datasets.load_dataset('my_dataset', 'second_domain')
|
131 |
+
BUILDER_CONFIGS = [SynStOpDatasetConfig(name="small", length=(10,15,20),version=VERSION, description="Small set of string operations with string slices only")] +\
|
132 |
+
[SynStOpDatasetConfig(name=f"small{l1}", length=(l1,), version=datasets.Version("0.0.1"), description="Small set of string operations with string slices only") for l1 in [10,15, 20]]
|
|
|
|
|
|
|
133 |
|
134 |
DEFAULT_CONFIG_NAME = "small" # It's not mandatory to have a default configuration. Just use one if it make sense.
|
135 |
|
136 |
def _info(self):
|
137 |
# TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
|
138 |
+
features = datasets.Features(
|
|
|
139 |
{
|
140 |
"input": datasets.Value("string"),
|
141 |
"output": datasets.Value("string"),
|
142 |
"code": datasets.Value("string"),
|
143 |
"res_var": datasets.Value("string"),
|
144 |
+
"operation": datasets.Value("string"),
|
145 |
+
"id": datasets.Value("int32"),
|
146 |
# These are the features of your dataset like images, labels ...
|
147 |
}
|
148 |
)
|
|
|
|
|
|
|
|
|
149 |
return datasets.DatasetInfo(
|
150 |
# This is the description that will appear on the datasets page.
|
151 |
description=_DESCRIPTION,
|
|
|
161 |
# Citation for the dataset
|
162 |
citation=_CITATION,
|
163 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
164 |
def _split_generators(self, dl_manager):
|
165 |
# TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration
|
166 |
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
|
|
|
168 |
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
|
169 |
# It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
|
170 |
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
171 |
+
urls = self.config.files
|
|
|
|
|
172 |
data_dir = dl_manager.download_and_extract(urls)
|
173 |
return [
|
174 |
datasets.SplitGenerator(
|
175 |
name=datasets.Split.TRAIN,
|
176 |
# These kwargs will be passed to _generate_examples
|
177 |
gen_kwargs={
|
178 |
+
"filepath": data_dir["train"],
|
179 |
"split": "train",
|
180 |
},
|
181 |
),
|
|
|
184 |
name=datasets.Split.TEST,
|
185 |
# These kwargs will be passed to _generate_examples
|
186 |
gen_kwargs={
|
187 |
+
"filepath": data_dir["test"],
|
188 |
"split": "test",
|
189 |
},
|
190 |
),
|
191 |
]
|
192 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
193 |
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
194 |
def _generate_examples(self, filepath, split):
|
195 |
# TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
|
|
|
201 |
for ix, data in enumerate(dataset):
|
202 |
|
203 |
if self.config.name.startswith("small"):
|
|
|
|
|
|
|
|
|
204 |
# Yields examples as (key, example) tuples
|
205 |
id = data["id"] if "id" in data else count
|
206 |
count = count + 1
|
|
|
209 |
"output": data["output"],
|
210 |
"code": data["code"],
|
211 |
"res_var": data["res_var"],
|
212 |
+
"id": id,
|
213 |
"operation": data["operation"]
|
214 |
}
|
215 |
else:
|
usage.py
DELETED
@@ -1,16 +0,0 @@
|
|
1 |
-
import datasets
|
2 |
-
|
3 |
-
if __name__=="__main__":
|
4 |
-
# load locally from this repo
|
5 |
-
ds = datasets.load_dataset("./stop.py", "small")
|
6 |
-
|
7 |
-
ds.push_to_hub("PaDaS-Lab/stop-small")
|
8 |
-
|
9 |
-
from datasets import load_dataset
|
10 |
-
|
11 |
-
dataset = load_dataset("PaDaS-Lab/stop-small")
|
12 |
-
print(dataset)
|
13 |
-
|
14 |
-
# load locally from this repo
|
15 |
-
# load locally from this repo
|
16 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|