Spaces:
Sleeping
Sleeping
File size: 6,692 Bytes
3fef185 |
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 |
from torch.utils.data import Dataset
import os
import json
import random
import hashlib
def stable_long_hash(input_string):
hash_object = hashlib.sha256(input_string.encode())
hex_digest = hash_object.hexdigest()
int_hash = int(hex_digest, 16)
long_long_hash = (int_hash & ((1 << 63) - 1))
return long_long_hash
model_map_authscan = {
"gpt-4o-mini-text": 1,
"gemini-2.0-text": 2,
"deepseek-text": 3,
"llama-text": 4
}
model_map_llmdetectaive = {
"gemma-text": 1,
"mixtral-text": 2,
"llama3-text": 3
}
model_map_hart = {
"claude-text": 1,
"gemini-text": 2,
"gpt-text": 3
}
def load_dataset(dataset_name,path=None):
dataset = {
"train": [],
"valid": [],
"test": []
}
if dataset_name == "falconset":
model_map = model_map_authscan
elif dataset_name == "llmdetectaive":
model_map = model_map_llmdetectaive
elif dataset_name == "hart":
model_map = model_map_hart
folder = os.listdir(path)
# print(folder)
for sub in folder:
sub_path = os.path.join(path, sub)
files = os.listdir(sub_path)
for file in files:
if not file.endswith('.jsonl'):
continue
file_path = os.path.join(sub_path, file)
key_name = file.split('.')[0]
assert key_name in dataset.keys(), f'{key_name} is not in dataset.keys()'
with open(file_path, 'r') as f:
data = [json.loads(line) for line in f]
for i in range(len(data)):
dct = {}
dct['text'] = data[i]['text']
if sub == "human-text":
dct['label'] = "human"
dct['label_detailed'] = "human"
dct['index'] = (1,0,0)
elif sub.startswith("human---"):
dct['label'] = "human+AI"
model = sub.split("---")[1]
dct['label_detailed'] = model
dct['index'] = (1, 1, model_map[model])
else:
dct['label'] = "AI"
dct['label_detailed'] = sub
dct['index'] = (0, 10^3, model_map[sub])
dataset[key_name].append(dct)
return dataset
def load_outdomain_dataset(path):
dataset = {
"valid": [],
"test": []
}
folder = os.listdir(path)
for sub in folder:
sub_path = os.path.join(path, sub)
files = os.listdir(sub_path)
for file in files:
if not file.endswith('.jsonl'):
continue
file_path = os.path.join(sub_path, file)
key_name = file.split('.')[0]
assert key_name in dataset.keys(), f'{key_name} is not in dataset.keys()'
with open(file_path, 'r', encoding='utf-8') as f:
data = [json.loads(line) for line in f]
for i in range(len(data)):
dct = {}
dct['text'] = data[i]['text']
if sub == "human-text":
dct['label'] = "human"
dct['label_detailed'] = "human"
dct['index'] = (1,0)
elif sub.startswith("human---"):
dct['label'] = "human+AI"
model = sub.split("---")[1]
dct['label_detailed'] = model
dct['index'] = (1, 1)
else:
dct['label'] = "AI"
dct['label_detailed'] = sub
dct['index'] = (0, 10^3)
dataset[key_name].append(dct)
return dataset
def load_dataset_conditional_lang(path=None, language='vi', seed=42):
dataset = {
"train": [],
"val": [],
"test": []
}
combined_data = []
random.seed(seed) # for reproducibility
folder = os.listdir(path)
print("Subfolders:", folder)
for sub in folder:
sub_path = os.path.join(path, sub)
if not os.path.isdir(sub_path):
continue
files = os.listdir(sub_path)
for file in files:
if not file.endswith('.jsonl') or language not in file:
continue
file_path = os.path.join(sub_path, file)
with open(file_path, 'r', encoding='utf-8') as f:
data = [json.loads(line) for line in f]
for entry in data:
if 'content' not in entry:
print("Key does not exist!")
continue
dct = {}
dct['text'] = entry['content']
if sub == "human":
dct['label'] = "human"
dct['label_detailed'] = "human"
dct['index'] = (1, 0, 0)
elif sub == "human+AI":
model = entry['label_detailed'].split("+")[1]
dct['label'] = "human+AI"
dct['label_detailed'] = model
dct['index'] = (1, 1, model_map[model])
else:
dct['label'] = "AI"
dct['label_detailed'] = entry['label_detailed']
dct['index'] = (0, 10**3, model_map[entry['label_detailed']])
combined_data.append(dct)
random.shuffle(combined_data)
total = len(combined_data)
train_end = int(total * 0.9)
val_end = train_end + int(total * 0.05)
dataset['train'] = combined_data[:train_end]
dataset['val'] = combined_data[train_end:val_end]
dataset['test'] = combined_data[val_end:]
print(f"Total: {total} | Train: {len(dataset['train'])} | Val: {len(dataset['val'])} | Test: {len(dataset['test'])}")
return dataset
class TextDataset(Dataset):
def __init__(self, dataset,need_ids=True,out_domain=0):
self.dataset = dataset
self.need_ids=need_ids
self.out_domain = out_domain
def get_class(self):
return self.classes
def __len__(self):
return len(self.dataset)
def __getitem__(self, idx):
text, label, label_detailed, index = self.dataset[idx].values()
id = stable_long_hash(text)
if self.out_domain:
label, is_mixed = index
if self.need_ids:
return int(id), text, int(label), int(is_mixed)
return text, int(label), int(is_mixed)
else:
label, is_mixed, write_model = index
if self.need_ids:
return int(id), text, int(label), int(is_mixed), int(write_model)
return text, int(label), int(is_mixed), int(write_model)
|