Repository navigation
Expand file tree
/
Copy pathevaluate_leakage.py
More file actions
133 lines (116 loc) · 4.3 KB
/
Copy pathevaluate_leakage.py
File metadata and controls
133 lines (116 loc) · 4.3 KB
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
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.
import json
import os
from analysis.leak_calculator import (
gen_classification_metrics,
print_classification_metrics,
save_classification_metrics,
)
import transformers
from src.pii_leakage.arguments.attack_args import AttackArgs
from src.pii_leakage.arguments.config_args import ConfigArgs
from src.pii_leakage.arguments.dataset_args import DatasetArgs
from src.pii_leakage.arguments.env_args import EnvArgs
from src.pii_leakage.arguments.evaluation_args import EvaluationArgs
from src.pii_leakage.arguments.model_args import ModelArgs
from src.pii_leakage.arguments.ner_args import NERArgs
from src.pii_leakage.attacks.attack_factory import AttackFactory
from src.pii_leakage.attacks.extraction.naive_extraction import (
NaiveExtractionAttack,
)
from src.pii_leakage.dataset.dataset_factory import DatasetFactory
from src.pii_leakage.models.language_model import LanguageModel
from src.pii_leakage.models.model_factory import ModelFactory
from src.pii_leakage.ner.pii_results import ListPII
from src.pii_leakage.utils.output import print_dict_highlighted
def parse_args():
parser = transformers.HfArgumentParser(
(
ModelArgs,
NERArgs,
DatasetArgs,
AttackArgs,
EvaluationArgs,
EnvArgs,
ConfigArgs,
)
)
return parser.parse_args_into_dataclasses()
def evaluate(
model_args: ModelArgs,
ner_args: NERArgs,
dataset_args: DatasetArgs,
attack_args: AttackArgs,
eval_args: EvaluationArgs,
env_args: EnvArgs,
config_args: ConfigArgs,
):
print("Evaluating baseline leakage - no patching...")
if config_args.exists():
model_args = config_args.get_model_args()
dataset_args = config_args.get_dataset_args()
attack_args = config_args.get_attack_args()
ner_args = config_args.get_ner_args()
env_args = config_args.get_env_args()
print_dict_highlighted(vars(attack_args))
# Load the target model (trained on private data)
lm: LanguageModel = ModelFactory.from_model_args(
model_args, env_args=env_args
).load(verbose=True)
train_dataset = DatasetFactory.from_dataset_args(
dataset_args=dataset_args.set_split("train"), ner_args=ner_args
)
real_pii: ListPII = train_dataset.load_pii().flatten(attack_args.pii_class)
print(
f"Sample 20 real PII out of {len(real_pii.unique().mentions())}: {real_pii.unique().mentions()[:20]}"
)
if "gpt2" in model_args.architecture:
model_arch = "gpt2"
elif "Qwen" in model_args.architecture:
model_arch = "qwen3-06"
else:
model_arch = "EleutherAI-pythia-160m-deduped"
# get the public PII
if "gpt2" in model_args.architecture:
pii_public_baseline = "gpt2"
else:
pii_public_baseline = "EleutherAI-pythia-160m-deduped"
baseline_pii = json.load(
open(f"pii/baseline_pii_{pii_public_baseline}_{attack_args.pii_class}.list")
)
print(f"Baseline PII count: {len(baseline_pii)}")
print(f"Baseline PII sample: {baseline_pii[:20]}")
"""Evaluate a model and attack pair."""
attack: NaiveExtractionAttack = AttackFactory.from_attack_args(
attack_args, ner_args=ner_args, env_args=env_args
)
for i in range(0, 3):
print(f"\nEvaluation run {i}")
generated_pii = set(attack.attack(lm).keys())
real_pii_set = set(real_pii.unique().mentions())
leaked_pii = generated_pii.difference(baseline_pii)
print(f"Generated: {len(generated_pii)}")
# Overall metrics
precision, recall, f1_score = gen_classification_metrics(
real_pii_set, leaked_pii
)
model_name = model_args.model_ckpt.split("/")[-1]
save_classification_metrics(
model_name,
attack_args.pii_class,
real_pii_set,
leaked_pii,
precision,
recall,
f1_score,
i,
is_circuits=False,
threshold=0,
ablation=None,
patch=None,
)
# ----------------------------------------------------------------------------
if __name__ == "__main__":
evaluate(*parse_args())
# ----------------------------------------------------------------------------