-
Notifications
You must be signed in to change notification settings - Fork 0
/
compute_data_overlap_metrics.py
266 lines (227 loc) · 11.3 KB
/
compute_data_overlap_metrics.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
import json
import os
import glob
from typing import List, Tuple, Set, DefaultDict
from nltk import ngrams
from typing import Dict
from tqdm import tqdm
from collections import defaultdict
from light_scenario import LightInstance, LightScenario, LightScenarioKey
from data_overlap_spec import (
DataOverlapStats,
DataOverlapStatsKey,
OverlapProtocolSpec,
EntryDataOverlapKey,
EntryOverlapNgrams,
)
from light_tokenizer import LightTokenizer
from load_documents import get_document_iterator
from common.hierarchical_logger import hlog, htrack_block
from common.general import asdict_without_nones
from common.arguments import get_data_overlap_args
from common.util import get_tokenizer
from scenarios.scenario import ScenarioSpec
PART_INPUT: str = "input"
PART_REF: str = "references"
# type alias for overlap-related data structures
Ngram = Tuple[str, ...]
NgramIndex = Dict[int, Dict[Ngram, Set[EntryDataOverlapKey]]]
NgramCounter = Dict[EntryDataOverlapKey, Dict[Ngram, int]]
def load_light_scenarios_from_jsonl(path: str) -> List[LightScenario]:
"""
Create a list of light scenarios from a jsonl file, where each json represents a LightScenario object.
Input file format:
Instance JSON 1
Instance JSON 2
Instance JSON 3
...
"""
def create_light_instance_from_dict(instance_dict: dict) -> LightInstance:
return LightInstance(
input=instance_dict[PART_INPUT], references=instance_dict[PART_REF], id=instance_dict["id"]
)
light_scenarios: List[LightScenario] = []
light_scenario_jsons = open(path, "r").readlines()
for light_scenario_json in light_scenario_jsons:
light_scenario_dict: dict = json.loads(light_scenario_json)
light_scenario_key_dict: dict = light_scenario_dict["scenario_key"]
# if the light_scenarios are exported from helm, they will have a scenario_spec field
scenario_spec = ScenarioSpec(**light_scenario_key_dict["scenario_spec"])
light_scenario_key = LightScenarioKey(scenario_spec=scenario_spec, split=light_scenario_key_dict["split"])
light_instances: List[LightInstance] = [
create_light_instance_from_dict(instance_dict) for instance_dict in light_scenario_dict["instances"]
]
light_scenarios.append(LightScenario(scenario_key=light_scenario_key, instances=light_instances))
return light_scenarios
def create_ngram_index(
light_scenarios: List[LightScenario],
n_values: List[int],
tokenizer: LightTokenizer,
stats_key_counts: Dict[DataOverlapStatsKey, int],
) -> NgramIndex:
"""
Given a list of scenarios and n values, initialize ngram_index.
stats_key_counts is passed in and updated, counting the number of times a stats_key occurs
"""
ngram_index: NgramIndex = {n: {} for n in n_values}
for scenario in light_scenarios:
hlog(f"Building ngram indexes for {scenario.scenario_key}")
for n in n_values:
stats_key = DataOverlapStatsKey(
light_scenario_key=scenario.scenario_key, overlap_protocol_spec=OverlapProtocolSpec(n=n)
)
stats_key_counts[stats_key] = len(scenario.instances)
for i, instance in enumerate(scenario.instances):
id = instance.id
assert id
input_tokens = tokenizer.tokenize(instance.input)
for input_ngram in ngrams(input_tokens, n):
if input_ngram not in ngram_index[n]:
ngram_index[n][input_ngram] = set()
ngram_index[n][input_ngram].add(
EntryDataOverlapKey(stats_key=stats_key, instance_id=id, part=PART_INPUT)
)
# compute reference ngrams
for reference in instance.references:
reference_unigrams = tokenizer.tokenize(reference)
for reference_ngram in ngrams(reference_unigrams, n):
if reference_ngram not in ngram_index[n]:
ngram_index[n][reference_ngram] = set()
ngram_index[n][reference_ngram].add(
EntryDataOverlapKey(stats_key=stats_key, instance_id=id, part=PART_REF)
)
return ngram_index
def compute_all_data_overlap(
training_file_path: str,
file_format: str,
ngram_index: NgramIndex,
tokenizer: LightTokenizer,
stats_key_to_input_ids: DefaultDict[DataOverlapStatsKey, Set[str]],
stats_key_to_reference_ids: DefaultDict[DataOverlapStatsKey, Set[str]],
entry_overlap_key_to_ngram_counts: DefaultDict[EntryDataOverlapKey, DefaultDict[str, int]],
output_ngrams: bool,
) -> None:
"""
Given an input file, compute a overlap stats for each n and each scenario by calling
`compute_document_data_overlap()` for each document in the file. The function writes
to the stats_key_to_input_ids and stats_key_to_reference_ids directly and does not return anything.
training_file_path: file path of the training data
file_format: format of the training file(s)
ngram_index: The ngram index that maps from ngrams to overlap stats
tokenizer: The tokenizer used to break documents in the file into tokens
stats_key_to_input_ids: a dict mapping the stats key to the overlapping input ids
stats_key_to_reference_ids: a dict mapping the stats key to the overlapping reference ids
entry_overlap_key_to_ngram_counts: a dict mapping the key to the overlapping ngrams
output_ngrams: whether we should output ngrams
"""
document_iterator = get_document_iterator(file_path=training_file_path, file_format=file_format)
for document in document_iterator:
compute_document_data_overlap(
document=document,
ngram_index=ngram_index,
tokenizer=tokenizer,
stats_key_to_input_ids=stats_key_to_input_ids,
stats_key_to_reference_ids=stats_key_to_reference_ids,
entry_overlap_key_to_ngram_counts=entry_overlap_key_to_ngram_counts,
output_ngrams=output_ngrams,
)
def compute_document_data_overlap(
document: str,
ngram_index: NgramIndex,
tokenizer: LightTokenizer,
stats_key_to_input_ids: DefaultDict[DataOverlapStatsKey, Set[str]],
stats_key_to_reference_ids: DefaultDict[DataOverlapStatsKey, Set[str]],
entry_overlap_key_to_ngram_counts: DefaultDict[EntryDataOverlapKey, DefaultDict[str, int]],
output_ngrams: bool,
) -> None:
"""
Given a document, compute a overlap stats for each n and each scenario. The function
writes to the overlap stats directly and does not return anything.
ngram_index: The ngram index that maps from ngrams to overlap stats
tokenizer: The tokenizer used to break the document into tokens
stats_key_to_input_ids: Dict to keep track of input_ids that are overlapping
stats_key_to_reference_ids: Dict to keep track of reference_ids that are overlapping
entry_overlap_key_to_ngram_counts: a dict mapping the key to the overlapping ngrams
output_ngrams: whether we should output ngrams
"""
document_tokens = tokenizer.tokenize(document)
for n in ngram_index.keys():
for document_ngram in ngrams(document_tokens, n):
if document_ngram in ngram_index[n]:
for entry_overlap_key in ngram_index[n][document_ngram]:
id = entry_overlap_key.instance_id
part = entry_overlap_key.part
if part == PART_INPUT:
stats_key_to_input_ids[entry_overlap_key.stats_key].add(id)
elif part == PART_REF:
stats_key_to_reference_ids[entry_overlap_key.stats_key].add(id)
if output_ngrams:
entry_overlap_key_to_ngram_counts[entry_overlap_key][document_ngram] += 1
if __name__ == "__main__":
args = get_data_overlap_args()
tokenizer: LightTokenizer = get_tokenizer(args.normalization)
input_file_paths: List[str]
if os.path.isdir(args.input_data):
input_file_paths = []
for file_path in glob.iglob(os.path.join(args.input_data, "**/*"), recursive=True):
if os.path.isfile(file_path):
input_file_paths.append(file_path)
else:
input_file_paths = [args.input_data]
hlog(f"The input data will be loaded from {input_file_paths}")
hlog(f"Loading scenario data from {args.scenario_data}")
light_scenarios = load_light_scenarios_from_jsonl(args.scenario_data)
stats_key_counts: DefaultDict[DataOverlapStatsKey, int] = defaultdict(int)
with htrack_block("Initializing the stats, ngram_index, and ngram_counter"):
ngram_index: NgramIndex
ngram_index = create_ngram_index(
light_scenarios=light_scenarios, n_values=args.N, tokenizer=tokenizer, stats_key_counts=stats_key_counts
)
# DataOverlapStatsKey -> Set[str] for ids
stats_key_to_input_ids: DefaultDict[DataOverlapStatsKey, Set] = defaultdict(set)
stats_key_to_reference_ids: DefaultDict[DataOverlapStatsKey, Set] = defaultdict(set)
entry_overlap_key_to_ngram_counts: DefaultDict[EntryDataOverlapKey, DefaultDict[str, int]] = defaultdict(
lambda: defaultdict(int)
)
# commpute the stats
with htrack_block("Computing overlap stats"):
for input_file_index in tqdm(
range(len(input_file_paths)), desc="Computing overlap stats for input files", disable=None
):
input_file_path: str = input_file_paths[input_file_index]
compute_all_data_overlap(
training_file_path=input_file_path,
file_format=args.input_format,
ngram_index=ngram_index,
tokenizer=tokenizer,
stats_key_to_input_ids=stats_key_to_input_ids,
stats_key_to_reference_ids=stats_key_to_reference_ids,
entry_overlap_key_to_ngram_counts=entry_overlap_key_to_ngram_counts,
output_ngrams=not args.no_output_ngrams,
)
if not args.no_output_ngrams:
all_entry_overlap_ngrams = []
with open(f"{args.output_stats}_ngrams", "w") as f:
for entry_overlap_key in entry_overlap_key_to_ngram_counts:
ngram_counts = [
ngram_count for ngram_count in entry_overlap_key_to_ngram_counts[entry_overlap_key].items()
]
entry_overlap_ngrams = EntryOverlapNgrams(
entry_data_overlap_key=entry_overlap_key, overlapping_ngram_counts=ngram_counts
)
all_entry_overlap_ngrams.append(entry_overlap_ngrams)
f.write(f"{json.dumps(asdict_without_nones(entry_overlap_ngrams))}\n")
all_data_overlap_stats = []
for stats_key, count in stats_key_counts.items():
data_overlap_stats = DataOverlapStats(
data_overlap_stats_key=stats_key,
instance_ids_with_overlapping_input=sorted(stats_key_to_input_ids[stats_key]),
instance_ids_with_overlapping_reference=sorted(stats_key_to_reference_ids[stats_key]),
num_instances=count,
)
all_data_overlap_stats.append(data_overlap_stats)
with open(args.output_stats, "w") as f:
f.writelines(
f"{json.dumps(asdict_without_nones(data_overlap_stats))}\n" for data_overlap_stats in all_data_overlap_stats
)
hlog(f"Written {len(all_data_overlap_stats)} results to {args.output_stats}")