Upload split_utils.py with huggingface_hub
Browse files- split_utils.py +34 -32
split_utils.py
CHANGED
@@ -1,4 +1,5 @@
|
|
1 |
import itertools
|
|
|
2 |
import re
|
3 |
from typing import Dict
|
4 |
|
@@ -8,8 +9,7 @@ from .stream import Stream
|
|
8 |
|
9 |
|
10 |
def parse_random_mix_string(input_str):
|
11 |
-
"""
|
12 |
-
Parses a string of format "source1[percentage1%]+source2[value2]+..." and returns a dictionary.
|
13 |
|
14 |
Args:
|
15 |
input_str (str): A string containing source names and their respective proportions. The format is
|
@@ -28,23 +28,27 @@ def parse_random_mix_string(input_str):
|
|
28 |
>>> parse_random_mix_string("dale[90%]+oren[0.7]+mike")
|
29 |
{'dale': 0.9, 'oren': 0.7, 'mike': 1.0}
|
30 |
"""
|
31 |
-
|
32 |
-
|
|
|
|
|
33 |
raise ValueError(f"Invalid input format for split '{input_str}'")
|
34 |
|
35 |
pattern = re.compile(r"([a-zA-Z]+)(\[\d*\.?\d*%?\])?")
|
36 |
matches = pattern.findall(input_str)
|
37 |
|
38 |
return {
|
39 |
-
name: float(value.strip("[]%")) / 100
|
|
|
|
|
40 |
for name, value in matches
|
41 |
}
|
42 |
|
43 |
|
44 |
def parse_slices_string(input_str):
|
45 |
-
"""
|
46 |
-
|
47 |
-
{"source1": [(value1,value2)], "source2": [(value2, None)], "source3": [(None,None)]...}
|
48 |
|
49 |
If a source appears multiple times with different indices, all index pairs are included in the list.
|
50 |
|
@@ -65,7 +69,6 @@ def parse_slices_string(input_str):
|
|
65 |
>>> parse_slices_string("oren[:50]+jake[24:]+test+oren[5:10]")
|
66 |
{'oren': [(None, 50), (5, 10)], 'jake': [(24, None)], 'test': [(None, None)]}
|
67 |
"""
|
68 |
-
|
69 |
result_dict = {}
|
70 |
|
71 |
# Split the input string into a list of sources
|
@@ -82,7 +85,9 @@ def parse_slices_string(input_str):
|
|
82 |
name = source
|
83 |
start = end = None
|
84 |
else:
|
85 |
-
raise ValueError(
|
|
|
|
|
86 |
|
87 |
if name not in result_dict:
|
88 |
result_dict[name] = [(start, end)]
|
@@ -100,14 +105,12 @@ def slice_stream(stream, start, end):
|
|
100 |
if end is not None:
|
101 |
stream = itertools.islice(stream, end)
|
102 |
|
103 |
-
|
104 |
-
yield item
|
105 |
# return stream
|
106 |
|
107 |
|
108 |
def slice_streams(input_streams, mapping):
|
109 |
-
"""
|
110 |
-
Slices multiple input streams according to a mapping and chains the results together.
|
111 |
|
112 |
Args:
|
113 |
input_streams (dict): A dictionary where the keys are the names of the input streams
|
@@ -129,12 +132,13 @@ def slice_streams(input_streams, mapping):
|
|
129 |
>>> slice_streams(old_streams, mapping)
|
130 |
{"new_train": [1, 2, 3, 4, 5, 8, 9], "new_test": [12, 13, 14]}
|
131 |
"""
|
132 |
-
|
133 |
new_streams = {}
|
134 |
for new_stream, sources in mapping.items():
|
135 |
|
136 |
def generator(new_stream, sources):
|
137 |
for old_stream, slices in sources.items():
|
|
|
|
|
138 |
old_stream_content = input_streams[old_stream]
|
139 |
for start, end in slices:
|
140 |
yield from slice_stream(old_stream_content, start, end)
|
@@ -147,8 +151,7 @@ def slice_streams(input_streams, mapping):
|
|
147 |
|
148 |
|
149 |
def build_stream_routing(mapping):
|
150 |
-
"""
|
151 |
-
Builds the stream mapping dictionary based on the provided mapping.
|
152 |
|
153 |
The stream mapping dictionary represents the mapping of old streams to new streams
|
154 |
and their respective probabilities. It ensures that the probabilities for each old stream
|
@@ -176,16 +179,15 @@ def build_stream_routing(mapping):
|
|
176 |
}
|
177 |
}
|
178 |
stream_mapping = build_stream_mapping(mapping)
|
179 |
-
|
180 |
# Output: {'my_old_stream1': (['my_new_stream', 'my_new_stream2'], [0.6, 0.4]),
|
181 |
# 'my_old_stream2': (['my_new_stream', 'my_new_stream2'], [0.2, 0.8])}
|
182 |
"""
|
183 |
-
|
184 |
stream_mapping = {}
|
185 |
|
186 |
# Calculate total weight for each old stream
|
187 |
total_weights = {}
|
188 |
-
for
|
189 |
for old_stream, weight in old_streams.items():
|
190 |
if old_stream not in total_weights:
|
191 |
total_weights[old_stream] = weight
|
@@ -203,13 +205,12 @@ def build_stream_routing(mapping):
|
|
203 |
if total_weights[old_stream] < 1:
|
204 |
stream_mapping[old_stream][None] = 1 - total_weights[old_stream]
|
205 |
|
206 |
-
|
207 |
-
return stream_mapping
|
208 |
|
209 |
|
210 |
def rename_split(input_streams: Dict[str, Stream], mapping: Dict[str, str]):
|
211 |
-
"""
|
212 |
-
|
213 |
Args:
|
214 |
input_streams (dict): A dictionary containing the input streams, where each key is
|
215 |
the name of the stream and the value is an iterable or generator
|
@@ -219,11 +220,14 @@ def rename_split(input_streams: Dict[str, Stream], mapping: Dict[str, str]):
|
|
219 |
|
220 |
Returns:
|
221 |
dict: A dictionary containing the generated new streams, where each key is the name
|
222 |
-
|
|
|
223 |
return {mapping.get(key, key): val for key, val in input_streams.items()}
|
224 |
|
225 |
|
226 |
-
def random_mix_generator(
|
|
|
|
|
227 |
for old_stream_name in new_stream_sources:
|
228 |
optinal_streams, weights = stream_routing[old_stream_name]
|
229 |
with nested_seed(old_stream_name) as rand:
|
@@ -237,8 +241,7 @@ def random_mix_generator(new_stream_name, new_stream_sources, stream_routing, in
|
|
237 |
|
238 |
|
239 |
def random_mix_streams(input_streams, mapping):
|
240 |
-
"""
|
241 |
-
Creates new streams based on the provided input streams and mapping.
|
242 |
|
243 |
The create_streams function generates new streams by selectively including items from
|
244 |
the old streams based on the specified mapping. Each item will be included in at most
|
@@ -273,11 +276,10 @@ def random_mix_streams(input_streams, mapping):
|
|
273 |
}
|
274 |
new_streams = create_streams(input_streams, mapping)
|
275 |
for new_stream_name, new_stream in new_streams.items():
|
276 |
-
|
277 |
for _, item in zip(range(10), new_stream):
|
278 |
-
|
279 |
"""
|
280 |
-
|
281 |
new_streams = {}
|
282 |
|
283 |
# Build stream routing
|
@@ -300,4 +302,4 @@ def random_mix_streams(input_streams, mapping):
|
|
300 |
|
301 |
|
302 |
if __name__ == "__main__":
|
303 |
-
|
|
|
1 |
import itertools
|
2 |
+
import logging
|
3 |
import re
|
4 |
from typing import Dict
|
5 |
|
|
|
9 |
|
10 |
|
11 |
def parse_random_mix_string(input_str):
|
12 |
+
"""Parses a string of format "source1[percentage1%]+source2[value2]+..." and returns a dictionary.
|
|
|
13 |
|
14 |
Args:
|
15 |
input_str (str): A string containing source names and their respective proportions. The format is
|
|
|
28 |
>>> parse_random_mix_string("dale[90%]+oren[0.7]+mike")
|
29 |
{'dale': 0.9, 'oren': 0.7, 'mike': 1.0}
|
30 |
"""
|
31 |
+
if not re.fullmatch(
|
32 |
+
r"(([a-zA-Z]+\[\d*\.?\d*%?\]|[a-zA-Z]+)\+)*([a-zA-Z]+\[\d*\.?\d*%?\]|[a-zA-Z]+)",
|
33 |
+
input_str,
|
34 |
+
):
|
35 |
raise ValueError(f"Invalid input format for split '{input_str}'")
|
36 |
|
37 |
pattern = re.compile(r"([a-zA-Z]+)(\[\d*\.?\d*%?\])?")
|
38 |
matches = pattern.findall(input_str)
|
39 |
|
40 |
return {
|
41 |
+
name: float(value.strip("[]%")) / 100
|
42 |
+
if "%" in value
|
43 |
+
else (float(value.strip("[]")) if value else 1.0)
|
44 |
for name, value in matches
|
45 |
}
|
46 |
|
47 |
|
48 |
def parse_slices_string(input_str):
|
49 |
+
"""Parses a string of format "source1[value1:value2] + source2[value2:] + source3 + ..." and returns a dictionary.
|
50 |
+
|
51 |
+
{"source1": [(value1,value2)], "source2": [(value2, None)], "source3": [(None,None)]...}.
|
52 |
|
53 |
If a source appears multiple times with different indices, all index pairs are included in the list.
|
54 |
|
|
|
69 |
>>> parse_slices_string("oren[:50]+jake[24:]+test+oren[5:10]")
|
70 |
{'oren': [(None, 50), (5, 10)], 'jake': [(24, None)], 'test': [(None, None)]}
|
71 |
"""
|
|
|
72 |
result_dict = {}
|
73 |
|
74 |
# Split the input string into a list of sources
|
|
|
85 |
name = source
|
86 |
start = end = None
|
87 |
else:
|
88 |
+
raise ValueError(
|
89 |
+
f'The input string "{input_str}" is not in the correct format.'
|
90 |
+
)
|
91 |
|
92 |
if name not in result_dict:
|
93 |
result_dict[name] = [(start, end)]
|
|
|
105 |
if end is not None:
|
106 |
stream = itertools.islice(stream, end)
|
107 |
|
108 |
+
yield from stream
|
|
|
109 |
# return stream
|
110 |
|
111 |
|
112 |
def slice_streams(input_streams, mapping):
|
113 |
+
"""Slices multiple input streams according to a mapping and chains the results together.
|
|
|
114 |
|
115 |
Args:
|
116 |
input_streams (dict): A dictionary where the keys are the names of the input streams
|
|
|
132 |
>>> slice_streams(old_streams, mapping)
|
133 |
{"new_train": [1, 2, 3, 4, 5, 8, 9], "new_test": [12, 13, 14]}
|
134 |
"""
|
|
|
135 |
new_streams = {}
|
136 |
for new_stream, sources in mapping.items():
|
137 |
|
138 |
def generator(new_stream, sources):
|
139 |
for old_stream, slices in sources.items():
|
140 |
+
if old_stream not in input_streams:
|
141 |
+
raise ValueError(f"'{old_stream}' is not available in input stream")
|
142 |
old_stream_content = input_streams[old_stream]
|
143 |
for start, end in slices:
|
144 |
yield from slice_stream(old_stream_content, start, end)
|
|
|
151 |
|
152 |
|
153 |
def build_stream_routing(mapping):
|
154 |
+
"""Builds the stream mapping dictionary based on the provided mapping.
|
|
|
155 |
|
156 |
The stream mapping dictionary represents the mapping of old streams to new streams
|
157 |
and their respective probabilities. It ensures that the probabilities for each old stream
|
|
|
179 |
}
|
180 |
}
|
181 |
stream_mapping = build_stream_mapping(mapping)
|
182 |
+
logging.info(stream_mapping)
|
183 |
# Output: {'my_old_stream1': (['my_new_stream', 'my_new_stream2'], [0.6, 0.4]),
|
184 |
# 'my_old_stream2': (['my_new_stream', 'my_new_stream2'], [0.2, 0.8])}
|
185 |
"""
|
|
|
186 |
stream_mapping = {}
|
187 |
|
188 |
# Calculate total weight for each old stream
|
189 |
total_weights = {}
|
190 |
+
for _new_stream, old_streams in mapping.items():
|
191 |
for old_stream, weight in old_streams.items():
|
192 |
if old_stream not in total_weights:
|
193 |
total_weights[old_stream] = weight
|
|
|
205 |
if total_weights[old_stream] < 1:
|
206 |
stream_mapping[old_stream][None] = 1 - total_weights[old_stream]
|
207 |
|
208 |
+
return {k: (list(v.keys()), list(v.values())) for k, v in stream_mapping.items()}
|
|
|
209 |
|
210 |
|
211 |
def rename_split(input_streams: Dict[str, Stream], mapping: Dict[str, str]):
|
212 |
+
"""Renames the streams.
|
213 |
+
|
214 |
Args:
|
215 |
input_streams (dict): A dictionary containing the input streams, where each key is
|
216 |
the name of the stream and the value is an iterable or generator
|
|
|
220 |
|
221 |
Returns:
|
222 |
dict: A dictionary containing the generated new streams, where each key is the name
|
223 |
+
of the new stream and the value is a generator representing the stream.
|
224 |
+
"""
|
225 |
return {mapping.get(key, key): val for key, val in input_streams.items()}
|
226 |
|
227 |
|
228 |
+
def random_mix_generator(
|
229 |
+
new_stream_name, new_stream_sources, stream_routing, input_streams
|
230 |
+
):
|
231 |
for old_stream_name in new_stream_sources:
|
232 |
optinal_streams, weights = stream_routing[old_stream_name]
|
233 |
with nested_seed(old_stream_name) as rand:
|
|
|
241 |
|
242 |
|
243 |
def random_mix_streams(input_streams, mapping):
|
244 |
+
"""Creates new streams based on the provided input streams and mapping.
|
|
|
245 |
|
246 |
The create_streams function generates new streams by selectively including items from
|
247 |
the old streams based on the specified mapping. Each item will be included in at most
|
|
|
276 |
}
|
277 |
new_streams = create_streams(input_streams, mapping)
|
278 |
for new_stream_name, new_stream in new_streams.items():
|
279 |
+
logging.info(f"{new_stream_name}:")
|
280 |
for _, item in zip(range(10), new_stream):
|
281 |
+
logging.info(item)
|
282 |
"""
|
|
|
283 |
new_streams = {}
|
284 |
|
285 |
# Build stream routing
|
|
|
302 |
|
303 |
|
304 |
if __name__ == "__main__":
|
305 |
+
logging.info(parse_random_mix_string("dale[90%]+oren[0.7]+mike"))
|