Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,207 +1,83 @@
|
|
|
|
1 |
import streamlit as st
|
2 |
-
|
3 |
-
import warnings
|
4 |
-
import gensim
|
5 |
-
from gensim.models.word2vec import Word2Vec
|
6 |
-
import gensim.downloader
|
7 |
-
import pandas as pd
|
8 |
-
from random import choice, shuffle
|
9 |
-
from mnemonic import Mnemonic
|
10 |
-
import bip32utils
|
11 |
-
import numpy as np
|
12 |
-
import sys, re
|
13 |
-
from time import sleep
|
14 |
-
from urllib.request import urlopen
|
15 |
-
import time
|
16 |
-
from joblib import Parallel, delayed
|
17 |
-
from tqdm import tqdm
|
18 |
-
from stqdm import stqdm
|
19 |
|
20 |
-
|
21 |
-
|
22 |
-
st.markdown("Welcome!")
|
23 |
-
BOUND = 10 ** 6
|
24 |
-
good_prefixes = {'19pcB', '1Bbf', '1NSme', '1wLR1', '1KBtw', '1Hu5J', '1F3n8', '172x'}
|
25 |
-
good_prefixes_small = {i.lower() for i in good_prefixes}
|
26 |
|
27 |
-
|
28 |
-
|
29 |
-
wv = gensim.downloader.load('glove-wiki-gigaword-50')
|
30 |
-
bip39_dict = [row.strip() for row in open("bip39dict.txt").readlines()]
|
31 |
-
forbiddenSet = set()
|
32 |
-
for row in open("forbidden.txt", "r").readlines():
|
33 |
-
forbiddenSet |= set(row.split())
|
34 |
-
return wv, bip39_dict, forbiddenSet
|
35 |
|
36 |
-
|
37 |
-
def get_2048():
|
38 |
-
table = {}
|
39 |
-
for word in bip39_dict:
|
40 |
-
res = []
|
41 |
-
for elem in bip39_dict:
|
42 |
-
cur_sim = wv.similarity(word, elem)
|
43 |
-
res.append((elem, cur_sim))
|
44 |
-
res.sort(key = lambda x: -x[1])
|
45 |
-
table[word] = res
|
46 |
-
return table
|
47 |
|
48 |
-
|
49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
res = []
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
return res[:topn]
|
59 |
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
blockchain_info_array.append (
|
71 |
-
float(re.search( r'%s":(\d+),' % tag, htmltext ).group(1)))
|
72 |
-
res = {}
|
73 |
-
out = "Bitcoin Address " + check_address + "\n"
|
74 |
-
for i, btc_tokens in enumerate(blockchain_info_array):
|
75 |
-
num = max(0, btc_tokens) / SATOSHIS_PER_BTC
|
76 |
-
res[blockchain_tags_json[i]] = num
|
77 |
-
out += "%s \t " % blockchain_tags_json[i] + "%.8f Bitcoin" % num + "\n"
|
78 |
-
st.text(out)
|
79 |
return res
|
80 |
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
92 |
|
93 |
-
def check(sentence, passphrase=None):
|
94 |
-
if not mnemon.check(' '.join(sentence)):
|
95 |
-
return
|
96 |
-
if passphrase is None:
|
97 |
-
for passphrase in passwords:
|
98 |
-
check(sentence, passphrase)
|
99 |
-
return
|
100 |
-
b44 = get_address(' '.join(sentence), passphrase)
|
101 |
-
flag1 = b44[:5] in good_prefixes or b44[:4] in good_prefixes
|
102 |
-
if flag1:
|
103 |
-
st.markdown(' '.join(sentence))
|
104 |
-
candidates.append((sentence, b44, passphrase))
|
105 |
-
res = check_balance(b44)
|
106 |
-
if res['final_balance'] > 0:
|
107 |
-
founded.append((sentence, b44, passphrase))
|
108 |
-
out = 'FOUNDED!\n' + '\n'.join(sentence) + '\n' + b44 + '\n' + passphrase + '\n'
|
109 |
-
st.markdown(out)
|
110 |
-
open('FOUNDED.txt', 'w').write(out)
|
111 |
-
exit(0)
|
112 |
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
return
|
117 |
-
if pos == 12:
|
118 |
-
res.append(cur)
|
119 |
-
return
|
120 |
-
for word in monthVars[pos]:
|
121 |
-
gen(pos + 1, cur + [word])
|
122 |
-
res = []
|
123 |
-
gen(0, [])
|
124 |
-
return res
|
125 |
|
126 |
-
|
127 |
-
def check_(cur):
|
128 |
-
bucket.append(cur)
|
129 |
-
return
|
130 |
-
assert len(MonthVariants) == 12
|
131 |
-
cntSearched = 0
|
132 |
-
variants = generateAll(MonthVariants)
|
133 |
-
shuffle(variants)
|
134 |
-
st.markdown(f"Количество вариантов для перебора: {len(variants)}")
|
135 |
-
bucket = []
|
136 |
-
for curSentence in stqdm(variants):
|
137 |
-
cntSearched += 1
|
138 |
-
for i in range(12):
|
139 |
-
cur = curSentence[i:] + curSentence[:i]
|
140 |
-
check_(cur)
|
141 |
-
cur = cur[::-1]
|
142 |
-
check_(cur)
|
143 |
-
if len(bucket) > 1000:
|
144 |
-
Parallel(n_jobs=n_jobs)(delayed(check)(words) for words in bucket)
|
145 |
-
bucket = []
|
146 |
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
similar_input = st.text_input("Введите слово")
|
152 |
-
with col2:
|
153 |
-
topn = st.slider("Сколько топ похожих вывести?", 10, 100, value=10)
|
154 |
-
submit_button = st.form_submit_button(label='Submit')
|
155 |
-
if submit_button:
|
156 |
-
show_forb = st.checkbox(label='Показывать выброшенные слова')
|
157 |
-
wv, bip39_dict, forbiddenSet = get()
|
158 |
-
res = top_from_bip39(similar_input.lower(), topn)
|
159 |
-
if not show_forb:
|
160 |
-
res2 = []
|
161 |
-
for word, num in res:
|
162 |
-
if word not in forbiddenSet:
|
163 |
-
res2.append((word, num))
|
164 |
-
res = res2.copy()
|
165 |
-
st.markdown("Топ похожих слов")
|
166 |
-
def highlight(s):
|
167 |
-
if s['Слово'] in forbiddenSet:
|
168 |
-
return ['background-color: red'] * len(s)
|
169 |
-
else:
|
170 |
-
return [None] * len(s)
|
171 |
-
df = pd.DataFrame(res, columns=['Слово', 'Похожесть'])
|
172 |
-
st.table(df.style.apply(highlight, axis=1))
|
173 |
|
174 |
-
with st.form(key='check_address'):
|
175 |
-
st.write("Проверить баланс на адресе")
|
176 |
-
col1, col2 = st.columns([2, 1])
|
177 |
-
with col1:
|
178 |
-
sentence_input = st.text_input(label="Введите предложение из 12 слов")
|
179 |
-
with col2:
|
180 |
-
password_input = st.text_input(label="Введите пароль")
|
181 |
-
submit_button = st.form_submit_button(label='Submit')
|
182 |
-
if submit_button:
|
183 |
-
b44 = get_address(sentence_input, password_input)
|
184 |
-
res = check_balance(b44)
|
185 |
-
if res['final_balance'] > 0:
|
186 |
-
out = 'FOUNDED!\n' + '\n'.join(sentence) + '\n' + b44 + '\n' + passphrase + '\n'
|
187 |
-
st.markdown(out)
|
188 |
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
cand_input = st.text_area("Введите список кандидатов", height=320, value=open("monthvars.txt").read())
|
194 |
-
with col2:
|
195 |
-
pass_input = st.text_area("Введите список паролей", height=320, value=open("passwords.txt").read())
|
196 |
-
n_jobs = 1
|
197 |
-
# n_jobs = st.slider("Количество потоков для параллеивания", 1, 10, value=1)
|
198 |
-
# debug = st.radio("print intermediate steps", [True, False])
|
199 |
-
submit_button = st.form_submit_button(label='Запустить перебор')
|
200 |
-
if submit_button:
|
201 |
-
candidates = []
|
202 |
-
founded = []
|
203 |
-
monthVars = [row.split() for row in cand_input.split('\n')]
|
204 |
-
passwords = [i.strip() for i in pass_input.split('\n')]
|
205 |
-
start_time = time.time()
|
206 |
-
searchSeed(monthVars[:12])
|
207 |
-
|
1 |
+
import torch
|
2 |
import streamlit as st
|
3 |
+
import transformers
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
+
from transformers import BertTokenizer, BertForMaskedLM
|
6 |
+
from transformers import BertForSequenceClassification, DataCollatorWithPadding
|
|
|
|
|
|
|
|
|
7 |
|
8 |
+
st.set_page_config(page_title="style transfer", layout="centered")
|
9 |
+
st.markdown("Welcome to text style transfer. Wait a few seconds for the model to load...")
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
+
|
14 |
+
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
|
15 |
+
bert_mlm_positive = BertForMaskedLM.from_pretrained('bert-base-uncased', return_dict=True).to(device).train(False)
|
16 |
+
bert_mlm_negative = BertForMaskedLM.from_pretrained('bert-base-uncased', return_dict=True).to(device).train(False)
|
17 |
+
bert_cls = BertForSequenceClassification.from_pretrained(
|
18 |
+
'bert-base-uncased', return_dict=True, problem_type="multi_label_classification", num_labels=2
|
19 |
+
).to(device).train(False)
|
20 |
+
|
21 |
+
def get_replacements(sentence: str, num_tokens, k_best, epsilon=1e-3):
|
22 |
+
"""
|
23 |
+
- split the sentence into tokens using the INGSOC-approved BERT tokenizer
|
24 |
+
- find :num_tokens: tokens with the highest ratio (see above)
|
25 |
+
- replace them with :k_best: words according to bert_mlm_positive
|
26 |
+
:return: a list of all possible strings (up to k_best * num_tokens)
|
27 |
+
"""
|
28 |
res = []
|
29 |
+
sentence_ix = tokenizer(sentence, return_tensors='pt')
|
30 |
+
sentence_ix = {key: value.to(device) for key, value in sentence_ix.items()}
|
31 |
+
length = len(sentence_ix['input_ids'][0])
|
32 |
+
probs_positive = bert_mlm_positive(**sentence_ix).logits.softmax(dim=-1)[0]
|
33 |
+
probs_negative = bert_mlm_negative(**sentence_ix).logits.softmax(dim=-1)[0]
|
34 |
+
p_tokens_positive = probs_positive[torch.arange(length), sentence_ix['input_ids'][0]]
|
35 |
+
p_tokens_negative = probs_negative[torch.arange(length), sentence_ix['input_ids'][0]]
|
|
|
36 |
|
37 |
+
p_relative = (p_tokens_positive + epsilon) / (p_tokens_negative + epsilon)
|
38 |
+
best_pos = torch.argsort(p_relative[1:-1], dim=0)[:num_tokens] + 1
|
39 |
+
best_pos_tokens = torch.argsort(probs_positive, dim=1)[..., -k_best:]
|
40 |
+
for pos in best_pos:
|
41 |
+
for replace_token in best_pos_tokens[pos]:
|
42 |
+
new_tensor = sentence_ix['input_ids'][0].cpu().numpy()
|
43 |
+
new_tensor[pos] = replace_token
|
44 |
+
new_sentence = tokenizer.decode(new_tensor[1:-1])
|
45 |
+
res.append(new_sentence)
|
46 |
+
# print(new_sentence)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
return res
|
48 |
|
49 |
+
def beamSearch(sentence, n_rounds=5):
|
50 |
+
labels = torch.tensor([[1, 1]], dtype=torch.float).to(device)
|
51 |
+
for i in range(n_rounds):
|
52 |
+
cur_res = get_replacements(sentence, num_tokens=num_tokens, k_best=k_best)
|
53 |
+
max_prob = -1
|
54 |
+
best_sentence = None
|
55 |
+
for candidate_sentence in cur_res:
|
56 |
+
inputs = tokenizer(candidate_sentence, return_tensors="pt").to(device)
|
57 |
+
outputs = bert_cls(**inputs, labels=labels)
|
58 |
+
prob_good = outputs.logits.softmax(dim=-1)[0][1]
|
59 |
+
if prob_good > max_prob:
|
60 |
+
max_prob = prob_good
|
61 |
+
best_sentence = candidate_sentence
|
62 |
+
if debug:
|
63 |
+
st.markdown(f"cur_sentence: {best_sentence}")
|
64 |
+
sentence = best_sentence
|
65 |
+
return sentence
|
66 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
67 |
|
68 |
+
bert_mlm_positive.load_state_dict(torch.load('mlm_positive.pth', map_location=torch.device('cpu')))
|
69 |
+
bert_mlm_negative.load_state_dict(torch.load('mlm_negative.pth', map_location=torch.device('cpu')))
|
70 |
+
# bert_cls.load_state_dict(torch.load('bert_cls.pth', map_location=torch.device('cpu')))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
|
72 |
+
user_input = st.text_input("Please enter something review")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
|
74 |
+
n_rounds = st.slider("Pick a number of rounds in beamSearch", 1, 10, value=5)
|
75 |
+
k_best = st.slider("Pick k_best parameter", 1, 5, value=3)
|
76 |
+
num_tokens = st.slider("Pick num_tokens parameter", 1, 5, value=3)
|
77 |
+
debug = st.radio("print intermediate steps", [True, False])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
78 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
|
80 |
+
if len(user_input.split()) > 0:
|
81 |
+
res = beamSearch(user_input, n_rounds=n_rounds)
|
82 |
+
st.markdown("Processed review:")
|
83 |
+
st.markdown(f"{res}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|