def warden_metadata():
from utils import pickle_it
meta = {}
meta['full_df'] = specter_df()
meta['wallet_list'] = current_app.specter.wallet_alias_list()
# Load pickle with previous checkpoint df
df_pkl = 'txs_pf.pkl'
data = pickle_it(action='load', filename=df_pkl)
if not isinstance(data, pd.DataFrame):
if data == 'file not found':
meta['df_old'] = None
else:
meta['df_old'] = data
# load difference / changes in addresses from file
ack_file = 'txs_diff.pkl'
data = pickle_it(action='load', filename=ack_file)
if data == 'file not found':
meta['ack_file'] = None
else:
meta['ack_file'] = data
meta['old_new_df_old'] = data['deleted']
meta['old_new_df_new'] = data['added']
return (meta)
def get_local_ip():
from utils import pickle_it
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
s.connect(('8.8.8.8', 1)) # connect() for UDP doesn't send packets
local_ip_address = s.getsockname()[0]
pickle_it('save', 'local_ip_address.pkl', local_ip_address)
return (local_ip_address)
def run_experiment(directory, dataset_path, params, nproc=None):
global A, y
# ray.init(address="auto")
if not os.path.exists(directory):
os.makedirs(directory)
pickle_it(params, 'params', directory)
print('load dataset')
with open(dataset_path, 'rb') as f:
A, y = joblib.load(f)
# A_obj = put(A)
# y_obj = put(y)
# for i in range(0, n_repeat):
# params[i].A_obj = A_obj
# params[i].y_obj = y_obj
# print("A.shape= ", A.shape, params[i].A_obj, params[i].y_obj)
print('start experiment')
# pool = Pool(ray_address='auto')
# results = pool.map(run_logistic, params)
results = run_logistic(params[0])
print(results)
pickle_it(results, 'results', directory)
print('results saved in "{}"'.format(directory))
def refresh_txs(self, load=True):
# Returns a dictionary with keys
# dict_keys(['pageCount', 'txlist', 'last_update'])
try:
if load:
data = pickle_it(action='load', filename='specter_txs.pkl')
if data != 'file not found':
return (data)
session = self.init_session()
response = session.post(self.tx_url, data=self.tx_payload)
specter_data = response.json()
# Include last update_time
specter_data['last_update'] = datetime.now().strftime(
'%m/%d/%Y, %H:%M:%S')
specter_data['txlist'] = json.loads(specter_data['txlist'])
# Save to pickle file
pickle_it(action='save',
filename='specter_txs.pkl',
data=specter_data)
return (specter_data)
except requests.exceptions.Timeout as e:
return ('[Specter Error] [refresh] Could not login to ' +
f'{self.base_url} <> Check address <> Error: {e}')
except Exception as e:
return ('[Specter Error] [refresh] {0}'.format(e))
def test_RealTimeBTC():
from pricing_engine.engine import realtime_price
ticker = 'BTC'
results = realtime_price(ticker)
run_time = datetime.now()
if results is None:
health = False
price = None
error_message = 'Realtime Price returned None'
try:
price = results['price']
health = True
error_message = None
except Exception as e:
health = False
price = None
error_message = f"Realtime Price returned an error: {e}"
data = {
'health': health,
'price': price,
'error_message': error_message,
'run_time': run_time
}
filename = 'status_realtime_btc.pkl'
pickle_it(action='save', filename=filename, data=data)
return (data)
def refresh_txs(self, load=True):
# Returns a dictionary with keys
# dict_keys(['pageCount', 'txlist', 'last_update'])
try:
if load:
data = pickle_it(action='load', filename='specter_txs.pkl')
if data != 'file not found':
return (data)
session = self.init_session()
response = session.post(self.tx_url,
data=self.tx_payload,
verify=False)
specter_data = response.json()
session.close()
# Include last update_time
specter_data['last_update'] = datetime.now().strftime(
'%m/%d/%Y, %H:%M:%S')
specter_data['txlist'] = json.loads(specter_data['txlist'])
'''
Handling for multiple outputs, CoinJoin transactions
'''
for tx in specter_data["txlist"]:
# sum outputs when more than one exists
if isinstance(tx["address"], list):
tx["amount"] = np.sum(np.array(tx["amount"], dtype=float))
# fees are only relevant for spends
if tx["category"] == "send":
# save a copy, as the original is still needed if it gets updated
try:
fee = tx["fee"]
# accounting for CoinJoins is unique because there's no realized fee for re-mixes
if tx["amount"] == fee:
tx["fee"] = 0
tx["amount"] -= fee
# ensure positivity for WARden (this needs to happen last)
tx["fee"] = abs(tx["fee"])
except Exception:
tx['fee'] = 0
# Save to pickle file
pickle_it(action='save',
filename='specter_txs.pkl',
data=specter_data)
return (specter_data)
except requests.exceptions.Timeout as e:
return ('[Specter Error] [refresh] Could not login to ' +
f'{self.base_url} <> Check address <> Error: {e}')
except Exception as e:
return ('[Specter Error] [refresh] {0}'.format(e))
def host_list():
services = pickle_it('load', 'services_found.pkl')
hosts = pickle_it('load', 'hosts_found.pkl')
if request.method == "GET":
delete = request.args.get("delete")
if delete is not None:
try:
del services[delete]
except Exception:
pass
try:
host = delete.strip("http://")
host = host.strip("https://")
host = host.strip("/")
if ':' in host:
host = host.split(":")[0]
for key, item in hosts.items():
if item['host'] == host:
del hosts[key]
break
except Exception:
pass
pickle_it('save', 'services_found.pkl', services)
pickle_it('save', 'hosts_found.pkl', hosts)
return redirect(url_for("warden.running_services"))
return (json.dumps(hosts))
if request.method == "POST":
url = request.form.get("new_url")
url = url_parser(url)
hosts[url] = {'ip': url, 'host': url, 'last_time': None}
pickle_it('save', 'hosts_found.pkl', hosts)
try:
if not '.onion' in url:
host_ip = socket.gethostbyname(url)
else:
host_ip = url
services[host_ip] = {
'url': url,
'status': 'Loading...',
'port': None,
'service': 'Checking Status'
}
pickle_it('save', 'services_found.pkl', services)
except Exception:
pass
return redirect(url_for("warden.running_services"))
def wallet_info(self, wallet_alias, load=True):
if wallet_alias is None:
return None
if load:
data = pickle_it(action='load',
filename=f'wallet_info_{wallet_alias}.pkl')
if data != 'file not found':
return (data)
url = self.base_url + f'wallets/wallet/{wallet_alias}/settings/'
metadata = {}
session = self.init_session()
page = session.get(url, verify=False)
soup = BeautifulSoup(page.text, 'html.parser')
session.close()
metadata['url'] = url
# Get device list
div_id = 'wallet_info_settings_tab'
data = soup.find("div", {"id": div_id})
if data is None:
metadata = {'error': True}
return metadata
metadata['title'] = data.find('h2').get_text()
if metadata['title'] == 'Devices':
metadata['subtitle'] = data.find_all('p')[0].get_text()
else:
metadata['subtitle'] = 'Single key'
# Get list of devices for this wallet
data = data.find_all('a', href=True)
metadata['devices'] = {}
for element in data:
device_info = {}
link = element['href']
alias = list(filter(None, link.split('/')))[-1]
device_info['url'] = self.base_url[:-1] + link
device_info['image'] = self.base_url[:-1] + element.find(
'img').get('src')
tmp = element.get_text().split('\n')
tmp = list(filter(None, tmp))
device_info['name'] = tmp[0].lstrip()
metadata['devices'][alias] = device_info
metadata['error'] = False
metadata['rescan'] = self.rescan_progress(wallet_alias=wallet_alias,
load=load,
session=session)
# Save to pickle file
pickle_it(action='save',
filename=f'wallet_info_{wallet_alias}.pkl',
data=metadata)
return metadata
def estimate_parameters(self, docs, tokenized_docs, feature_names):
"""
Estimates probability of each class, each feature in each class and saves the trained parameters.
:param list docs: Documents with defined classes
:param tokenized_docs: Tokenized documents
:param feature_names: Feature vector which is computed with scikit-learn TfidfVectorizer module
:return:
"""
self.class_probability(docs)
self.feature_probability(feature_names, tokenized_docs)
pickle_it(self, "naive_bayes_model")
print("Parameter estimation complete. Predicting trained documents...")
def trim_graph(graph, reduce_sample=True, pickle=True, from_scratch=True):
if not graph and not from_scratch:
graph = reload_json(USER_GRAPH_FNAME, transform=nx.node_link_graph)
return graph
rng_state = reload_object(RNG_FNAME, random.getstate)
random.setstate(rng_state)
print("Trimming graph...")
significant_id_set = set()
for direct in (Direct.IN, Direct.OUT):
sample = []
ids = []
for user_id in graph:
ids.append(user_id)
num_neighb = direct.deg_view(graph)[user_id]
sample.append(num_neighb)
sample_mean = mean(sample)
pop_stdev = stdev(sample)
for i, degree in enumerate(sample):
if abs(degree - sample_mean) > STDEV_MOD * pop_stdev:
user_id = ids[i]
significant_id_set.add((user_id, degree))
by_asc_degree = sorted(list(significant_id_set), key=lambda x: x[1])
significant_ids = [i[0] for i in by_asc_degree]
to_subgraph = set()
for user_id in significant_ids:
try:
others = set(graph.neighbors(user_id))
except KeyError:
continue
if reduce_sample and len(others) != 0:
others = random.sample(others, int(len(others) * OTHERS_MOD))
if len(others) == 0:
continue
to_subgraph.add(user_id)
for other in others:
to_subgraph.add(other)
pickle_it(rng_state, RNG_FNAME)
user_graph = graph.subgraph(to_subgraph)
if pickle:
json_it(user_graph, USER_GRAPH_FNAME, nx.node_link_data)
return user_graph
def onion_string():
from utils import pickle_it
if app.settings['SERVER'].getboolean('onion_server'):
try:
pickle_it('save', 'onion_address.pkl',
app.tor_service_id + '.onion')
return (f"""
{emoji.emojize(':onion:')} Tor Onion server running at:
{yellow(app.tor_service_id + '.onion')}
""")
except Exception:
return (yellow("[!] Tor Onion Server Not Running"))
else:
return ('')
def start_hidden_service(app):
if app.controller is None:
return
app.controller.reconnect()
key_path = os.path.abspath(os.path.join(home_path(), ".tor_service_key"))
app.tor_service_id = None
if not os.path.exists(key_path):
service = app.controller.create_ephemeral_hidden_service(
{app.tor_port: app.port}, await_publication=True)
app.tor_service_id = service.service_id
print("* Started a new hidden service with the address of %s.onion" %
app.tor_service_id)
with open(key_path, "w") as key_file:
key_file.write("%s:%s" %
(service.private_key_type, service.private_key))
else:
with open(key_path) as key_file:
key_type, key_content = key_file.read().split(":", 1)
try:
service = app.controller.create_ephemeral_hidden_service(
{app.tor_port: app.port},
key_type=key_type,
key_content=key_content,
await_publication=True,
)
app.tor_service_id = service.service_id
print("")
print(success("✅ Resumed %s.onion" % app.tor_service_id))
from utils import pickle_it
pickle_it('save', 'onion_address.pkl',
app.tor_service_id + '.onion')
print("")
except Exception:
pass
try:
# save address to file
if app.save_tor_address_to is not None:
with open(app.save_tor_address_to, "w") as f:
f.write("%s.onion" % app.tor_service_id)
app.tor_service_id = app.tor_service_id
from utils import pickle_it
pickle_it('save', 'onion_address.pkl', app.tor_service_id + '.onion')
app.tor_enabled = True
except Exception:
pass
def internet_connected(host="8.8.8.8", port=53, timeout=3):
"""
Host: 8.8.8.8 (google-public-dns-a.google.com)
OpenPort: 53/tcp
Service: domain (DNS/TCP)
"""
connected = False
try:
socket.setdefaulttimeout(timeout)
socket.socket(socket.AF_INET, socket.SOCK_STREAM).connect((host, port))
connected = True
except socket.error as ex:
connected = False
pickle_it('save', 'diags_internet_connected.pkl', connected)
return (connected)
def background_specter_health():
health = app.specter.is_healthy()
if health is False:
message = Message(
category='Specter Server',
message_txt=
f"<span class='text-danger'>Health Check Returned False at {app.specter.base_url}</span>"
)
app.message_handler.add_message(message)
health_text = pickle_it('load', 'specter_health.pkl')
message = Message(
category='Specter Server',
message_txt=
f"<span class='text-danger'>Health Check Result: {health_text}</span>"
)
app.message_handler.add_message(message)
# Check for other URLs where Specter could be located
seeker = app.specter.seek_specter()
if seeker is not None:
# Open Config and update
config_file = Config.config_file
config_settings = configparser.ConfigParser()
config_settings.read(config_file)
config_settings['SPECTER']['url'] = seeker['url']
with open(config_file, 'w') as file:
config_settings.write(file)
message = Message(
category='Specter Server',
message_txt=
f"<span class='text-success'>Found an alternative server at {app.specter.base_url}.<br>Updated to use this server.</span>"
)
app.message_handler.add_message(message)
return
else:
message = Message(
category='Specter Server',
message_txt=
f"<span class='text-danger'>Could not find any alternative servers running</span>"
)
app.message_handler.add_message(message)
health_text = pickle_it('load', 'specter_health.pkl')
message = Message(
category='Specter Server',
message_txt=
f"<span class='text-info'>Health Check Result: {health_text}</span>")
app.message_handler.add_message(message)
def get_pickle():
filename = request.args.get("filename")
if not filename:
return None
filename += ".pkl"
data_loader = pickle_it(action='load', filename=filename)
return (json.dumps(data_loader, default=lambda o: '<not serializable>'))
def run_experiment(directory, dataset_pickle, params, nproc=None):
global X, y
if not os.path.exists(directory):
os.makedirs(directory)
pickle_it(params, 'params', directory)
print('load dataset')
with open(dataset_pickle, 'rb') as f:
X, y = pickle.load(f)
print('start experiment')
with mp.Pool(nproc) as pool:
results = pool.map(run_logistic, params)
pickle_it(results, 'results', directory)
print('results saved in "{}"'.format(directory))
def word_to_vec(self, data, is_test=True):
"""
Converts documents to tf-idf vectors.
:param list data: Text documents to be converted
:param bool is_test: If true then uses trained tf-idf model. Else, trains and saved tf-idf model
:return:
"""
if is_test:
vectorizer = unpickle_it("tfidf_model")
self.tfidf_vectors = vectorizer.transform(data)
else:
vectorizer = TfidfVectorizer()
vectorizer.fit(data)
self.tfidf_vectors = vectorizer.transform(data)
pickle_it(vectorizer, "tfidf_model")
self.feature_names = vectorizer.get_feature_names()
def run_experiment(directory, dataset_path, params, nproc=None):
global A, y
if not os.path.exists(directory):
os.makedirs(directory)
pickle_it(params, 'params', directory)
print('load dataset')
with open(dataset_path, 'rb') as f:
A, y = pickle.load(f)
print("A.shape= ", A.shape)
print('start experiment')
# with mp.Pool() as pool:
# results = pool.map(run_logistic, params)
results = run_logistic(params[0])
print(results)
pickle_it(results, 'results', directory)
print('results saved in "{}"'.format(directory))
def rescan_progress(self, wallet_alias, load=True, session=None):
if load:
data = pickle_it(action='load',
filename=f'specter_rescan_{wallet_alias}.pkl')
if data != 'file not found':
return (data)
try:
url = self.base_url + f"wallets/wallet/{wallet_alias}/rescan_progress"
if not session:
session = self.init_session()
response = session.get(url)
data = response.json()
# Save to pickle file
pickle_it(action='save',
filename=f'specter_rescan_{wallet_alias}.pkl',
data=data)
return (data)
except Exception as e:
return ('[Specter Error] [rescan] {0}'.format(e))
def build_graph(pickle=False, from_scratch=True):
user_graph = nx.DiGraph()
if not from_scratch:
print("Loading graph data.")
user_graph = reload_object(USER_GRAPH_FNAME, nx.DiGraph)
if user_graph:
return user_graph
print("Building graph...")
if VALID_USER_FRAME.empty:
sys.stderr.write("ERROR: A user or tweet dictionary is empty.")
sys.exit(NO_DATA_EXIT_CODE)
for i, user_id in enumerate(USER_LIST):
new_edges = {Direct.IN: [], Direct.OUT: []}
for direct in (Direct.IN, Direct.OUT):
num, expected = get_expected_connection_bounds(user_graph,
user_id,
direct=direct)
if num == 0:
others = (VALID_USER_FRAME.loc[user_id])[direct.twit_key()]
for ident in others:
edge = direct.make_edge(int(user_id), int(ident))
new_edges[direct].append(edge)
elif num != expected:
print(
f"followers mismatch for node {user_id}: {num}, {expected}, {direct}"
)
print("Skipping")
print()
new_in = new_edges[Direct.IN]
new_out = new_edges[Direct.OUT]
if len(new_in) != 0 or len(new_out) != 0:
user_graph.add_edges_from(new_in)
user_graph.add_edges_from(new_out)
if (i + 1) % 100 == 0:
print(f"Analyzed %d users" % (i + 1))
if pickle:
pickle_it(user_graph, FULL_GRAPH_FNAME)
return user_graph
def is_healthy(self):
reach = self.is_reachable()
if reach is False:
pickle_it("save", "specter_health.pkl", "Unreacheable")
return False
auth = self.is_auth()
if auth is False:
pickle_it("save", "specter_health.pkl", "Could not Authenticate")
return False
# Check Node Health
try:
if self.home_parser()['bitcoin_core_is_syncing'] is True:
pickle_it("save", "specter_health.pkl",
"Bitcoin Node Not Fully Synched")
return False
else:
pickle_it("save", "specter_health.pkl", "Running")
return True
except Exception as e:
pickle_it("save", "specter_health.pkl", f"Error: {e}")
return False
def is_service_running(service, expiry=None):
# Expiry in seconds since last time reached
# usage: is_service_running('WARden Server', 10)
services_found = pickle_it('load', 'services_found.pkl')
if services_found == 'file not found':
return (False, None)
for key, val in services_found.items():
if val['service'] == service:
if expiry is not None:
utc_time = datetime.utcnow()
epoch_time = (utc_time - datetime(1970, 1, 1)).total_seconds()
if epoch_time - val['last_update'] > expiry:
continue
return (True, val)
return (False, None)
def alert_activity():
alerts = False
# Don't send any alerts as activity still being downloaded
if current_app.downloading:
return alerts
ack_file = 'txs_diff.pkl'
try:
data = pickle_it(action='load', filename=ack_file)
if data == 'file not found':
raise FileNotFoundError
if data['changes_detected_on'] is not None:
return (True)
else:
return (False)
except Exception:
return (False)
def apikey(source, required=True):
# Check if a cryptocompare key is stored at home directory
if source == 'cryptocompare':
from utils import pickle_it
API_KEY = pickle_it('load', 'cryptocompare_api.pkl')
if "\n" in API_KEY:
API_KEY = API_KEY.strip("\n")
if API_KEY != 'file not found':
return API_KEY
# GET API_KEY
if load_config().has_option('API', source):
API_KEY = load_config()['API'][source]
else:
API_KEY = None
if required and API_KEY is None:
raise Exception(f'{source} requires an API KEY and none was found.')
return API_KEY
def init_session(self):
with requests.session() as session:
if "onion" in self.base_url:
config = load_config()
port = config['TOR'].get('port')
TOR = pickle_it('load', 'tor.pkl')
port = TOR['port']
if port is None or port == 'failed':
port = 9050
session.proxies = {
"http": "socks5h://0.0.0.0:" + port,
"https": "socks5h://0.0.0.0:" + port,
}
response = session.post(self.login_url,
data=self.login_payload,
timeout=60,
verify=False)
# Check if authorized
if response.status_code == 401:
raise Exception(
'Unauthorized Login to Specter. Check Username and/or Password.'
)
return (session)
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20]],
'Thursday': [[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20]],
"Friday": [[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20]],
"Saturday": [[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20]],
"Sunday": [[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20],
[20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20], [20, 20, 20, 20]]
}
if __name__ == '__main__':
from utils import pickle_it
pickle_it(DEFAULT_TIMER_SETTINGS, "timer")
def run_parallel_experiment(directory, dataset_pickle, models, cores, baseline, repeat=3):
if not os.path.exists(directory):
os.makedirs(directory)
pickle_it([m(1) for m in models], 'models', directory)
pickle_it(cores, 'cores', directory)
print('load dataset')
with open(dataset_pickle, 'rb') as f:
X, y = pickle.load(f)
print('start experiment')
chronos = np.zeros((len(models), len(cores), repeat))
stop_times = np.zeros((len(models), len(cores), repeat), dtype=int)
for r in range(repeat):
for c_idx, core in enumerate(cores):
for m_idx, model in enumerate(models):
p = model(core)
print("{} - cores {} - repeat {}".format(p, core, r))
m = LogisticParallelSGD(p)
timing, epoch, iteration, losses = m.fit_until(X, y, num_features=X.shape[1], num_samples=X.shape[0],
baseline=baseline)
chronos[m_idx, c_idx, r] = timing
stop_times[m_idx, c_idx, r] = epoch * X.shape[0] + iteration
pickle_it(chronos, 'chronos', directory)
pickle_it(stop_times, 'stop_times', directory)
pickle_it(chronos, 'chronos', directory)
pickle_it(stop_times, 'stop_times', directory)
print('results saved in "{}"'.format(directory))
def main(args):
with open(args.data_dir+'/ptb.vocab.json', 'r') as file:
vocab = json.load(file)
# required to map between integer-value sentences and real sentences
w2i, i2w = vocab['w2i'], vocab['i2w']
# make sure our models for the VAE and Actor exist
if not os.path.exists(args.load_vae):
raise FileNotFoundError(args.load_vae)
model = SentenceVAE(
vocab_size=len(w2i),
sos_idx=w2i['<sos>'],
eos_idx=w2i['<eos>'],
pad_idx=w2i['<pad>'],
unk_idx=w2i['<unk>'],
max_sequence_length=args.max_sequence_length,
embedding_size=args.embedding_size,
rnn_type=args.rnn_type,
hidden_size=args.hidden_size,
word_dropout=args.word_dropout,
embedding_dropout=args.embedding_dropout,
latent_size=args.latent_size,
num_layers=args.num_layers,
bidirectional=args.bidirectional
)
model.load_state_dict(
torch.load(args.load_vae, map_location=lambda storage, loc: storage))
model.eval()
print("vae model loaded from %s"%(args.load_vae))
# to run in constraint mode, we need the trained generator
if args.constraint_mode:
if not os.path.exists(args.load_actor):
raise FileNotFoundError(args.load_actor)
actor = Actor(
dim_z=args.latent_size, dim_model=2048, num_labels=args.n_tags)
actor.load_state_dict(
torch.load(args.load_actor, map_location=lambda storage, loc:storage))
actor.eval()
print("actor model loaded from %s"%(args.load_actor))
if torch.cuda.is_available():
model = model.cuda()
if args.constraint_mode:
actor = actor.cuda() # TODO: to(self.devices)
if args.sample:
print('*** SAMPLE Z: ***')
# get samples from the prior
sample_sents, z = model.inference(n=args.num_samples)
sample_sents, sample_tags = get_sents_and_tags(sample_sents, i2w, w2i)
pickle_it(z.cpu().numpy(), 'samples/z_sample_n{}.pkl'.format(args.num_samples))
pickle_it(sample_sents, 'samples/sents_sample_n{}.pkl'.format(args.num_samples))
pickle_it(sample_tags, 'samples/tags_sample_n{}.pkl'.format(args.num_samples))
print(sample_sents, sep='\n')
if args.constraint_mode:
print('*** SAMPLE Z_PRIME: ***')
# get samples from the prior, conditioned via the actor
all_tags_sample_prime = []
all_sents_sample_prime = {}
all_z_sample_prime = {}
for i, condition in enumerate(LABELS):
# binary vector denoting each of the PHRASE_TAGS
labels = torch.Tensor(condition).repeat(args.num_samples, 1).cuda()
# take z and manipulate using the actor to generate z_prime
z_prime = actor.forward(z, labels)
sample_sents_prime, z_prime = model.inference(
z=z_prime, n=args.num_samples)
sample_sents_prime, sample_tags_prime = get_sents_and_tags(
sample_sents_prime, i2w, w2i)
print('conditoned on: {}'.format(condition))
print(sample_sents_prime, sep='\n')
all_tags_sample_prime.append(sample_tags_prime)
all_sents_sample_prime[LABEL_NAMES[i]] = sample_sents_prime
all_z_sample_prime[LABEL_NAMES[i]] = z_prime.data.cpu().numpy()
pickle_it(all_tags_sample_prime, 'samples/tags_sample_prime_n{}.pkl'.format(args.num_samples))
pickle_it(all_sents_sample_prime, 'samples/sents_sample_prime_n{}.pkl'.format(args.num_samples))
pickle_it(all_z_sample_prime, 'samples/z_sample_prime_n{}.pkl'.format(args.num_samples))
if args.interpolate:
# get random samples from the latent space
z1 = torch.randn([args.latent_size]).numpy()
z2 = torch.randn([args.latent_size]).numpy()
z = to_var(torch.from_numpy(interpolate(start=z1, end=z2, steps=args.num_samples-2)).float())
print('*** INTERP Z: ***')
interp_sents, _ = model.inference(z=z)
interp_sents, interp_tags = get_sents_and_tags(interp_sents, i2w, w2i)
pickle_it(z.cpu().numpy(), 'samples/z_interp_n{}.pkl'.format(args.num_samples))
pickle_it(interp_sents, 'samples/sents_interp_n{}.pkl'.format(args.num_samples))
pickle_it(interp_tags, 'samples/tags_interp_n{}.pkl'.format(args.num_samples))
print(interp_sents, sep='\n')
if args.constraint_mode:
print('*** INTERP Z_PRIME: ***')
all_tags_interp_prime = []
all_sents_interp_prime = {}
all_z_interp_prime = {}
for i, condition in enumerate(LABELS):
# binary vector denoting each of the PHRASE_TAGS
labels = torch.Tensor(condition).repeat(args.num_samples, 1).cuda()
# z prime conditioned on this particular binary variable
z_prime = actor.forward(z, labels)
interp_sents_prime, z_prime = model.inference(
z=z_prime, n=args.num_samples)
interp_sents_prime, interp_tags_prime = get_sents_and_tags(
interp_sents_prime, i2w, w2i)
print('conditoned on: {}'.format(condition))
print(interp_sents_prime, sep='\n')
all_tags_interp_prime.append(interp_tags_prime)
all_sents_interp_prime[LABEL_NAMES[i]] = interp_sents_prime
all_z_interp_prime[LABEL_NAMES[i]] = z_prime.data.cpu().numpy()
pickle_it(all_tags_interp_prime, 'samples/tags_interp_prime_n{}.pkl'.format(args.num_samples))
pickle_it(all_sents_interp_prime, 'samples/sents_interp_prime_n{}.pkl'.format(args.num_samples))
pickle_it(all_z_interp_prime, 'samples/z_interp_prime_n{}.pkl'.format(args.num_samples))
import IPython; IPython.embed()
def home_parser(self, load=True):
if load:
data = pickle_it(action='load', filename='specter_home.pkl')
if data != 'file not found':
return (data)
url = self.base_url + 'about'
metadata = {}
try:
session = self.init_session()
page = session.get(url, verify=False)
session.close()
except Exception as e:
metadata['error'] = str(e)
return (metadata)
metadata['specter_health'] = pickle_it('load', 'specter_health.pkl')
if metadata['specter_health'] == 'file not found':
metadata['specter_health'] = 'Failed Check'
soup = BeautifulSoup(page.text, 'html.parser')
# Get Specter Version
try:
metadata['version'] = (soup.find(
text=re.compile('Specter Version')).parent()[0].get_text())
except Exception as e:
metadata['version'] = f'Error: {e}'
# Get Bitcoin Core Data
# Check Sync Status
metadata['bitcoin_core_is_syncing'] = False
# This is the text that will be searched on Specter
# page to signal that Core is still synching
txt_sc = 'Bitcoin Core is still syncing'
if txt_sc in page.text:
metadata['bitcoin_core_is_syncing'] = True
# Same but for updating Wallets Check
txt_sc = 'Updating wallets data'
metadata['wallets_is_syncing'] = False
if txt_sc in page.text:
metadata['wallets_is_syncing'] = True
try:
div_id = 'bitcoin_core_info'
data = soup.find("div", {"id": div_id})
data = data.find('table')
data = data.find_all('tr')
bitcoin_core_data = {}
for element in data:
cols = element.find_all('td')
bitcoin_core_data[cols[0].get_text().split(':')
[0]] = cols[1].get_text()
metadata['bitcoin_core_data'] = bitcoin_core_data
# Format blocks count
try:
metadata['bitcoin_core_data']['Blocks count'] = (
"{0:,.0f}".format(
float(metadata['bitcoin_core_data']['Blocks count'])))
metadata['bitcoin_core_data']['Difficulty'] = (
"{0:,.0f}".format(
float(metadata['bitcoin_core_data']['Difficulty'])))
except Exception:
pass
except Exception as e:
metadata['bitcoin_core_html'] = (
f"<span class='text-warning'>Error: {str(e)}</span>")
metadata['bitcoin_core_data'] = {'error': str(e)}
# Get Wallet Names
wallet_dict = {}
wallet_alias = []
try:
div_id = "wallets_list"
data = soup.find("div", {"id": div_id})
data = data.find_all('a', href=True)
for element in data:
try:
link = element['href']
if 'wallets/new_wallet/' in link:
continue
alias = list(filter(None, link.split('/')))[-1]
wallet_alias.append(alias)
wallet_dict[alias] = {}
wallet_dict[alias]['url'] = self.base_url[:-1] + link
find_class = 'grow'
wallet_info = element.findAll("div",
{"class": find_class})[0]
wallet_info = wallet_info.get_text().split('\n')
wallet_info = list(filter(None, wallet_info))
wallet_dict[alias]['name'] = wallet_info[0].lstrip()
wallet_dict[alias]['keys'] = wallet_info[1]
except Exception as e:
pass
metadata['alias_list'] = wallet_alias
metadata['wallet_dict'] = wallet_dict
except Exception as e:
metadata['alias_list'] = None
metadata['wallet_dict'] = {'error': str(e)}
# Get Device Names
device_dict = {}
device_list = []
try:
div_id = "devices_list"
data = soup.find("div", {"id": div_id})
data = data.find_all('a', href=True)
for element in data:
try:
link = element['href']
if 'devices/new_device_type' in link:
continue
alias = list(filter(None, link.split('/')))[-1]
device_list.append(alias)
device_dict[alias] = {}
device_dict[alias]['url'] = self.base_url[:-1] + link
device_dict[alias][
'image'] = self.base_url[:-1] + element.find(
'img').get('src')
device_info = element.get_text().split('\n')
device_dict[alias]['name'] = list(filter(
None, device_info))[0].lstrip()
device_dict[alias]['keys'] = list(filter(
None, device_info))[1]
except Exception:
pass
metadata['device_list'] = device_list
metadata['device_dict'] = device_dict
metadata['last_update'] = datetime.now().strftime(
'%m/%d/%Y, %H:%M:%S')
except Exception as e:
metadata['device_list'] = None
metadata['device_dict'] = {'error': str(e)}
# Save to pickle file
pickle_it(action='save', filename='specter_home.pkl', data=metadata)
return (metadata)
regularizer=1 / n,
estimate='mean',
qsgd_s=2**4))
params.append(
Parameters(name="qsgd-2bit",
num_epoch=NUM_EPOCH,
lr_type='bottou',
initial_lr=10.,
regularizer=1 / n,
estimate='mean',
qsgd_s=2**2))
def run_logistic(param):
m = LogisticSGD(param)
res = m.fit(X, y)
print('{} - score: {}'.format(param, m.score(X, y)))
return res
if not os.path.exists(RESULT_DIR):
os.makedirs(RESULT_DIR)
pickle_it(params, 'params', RESULT_DIR)
print('start experiment')
with mp.Pool(len(params)) as pool:
results = pool.map(run_logistic, params)
pickle_it(results, 'results', RESULT_DIR)
print('results saved in "{}"'.format(RESULT_DIR))
print('load dataset')
with open(args.dataset_file, 'rb') as f:
X, y = pickle.load(f)
n, d = X.shape
model = Parameters(num_epoch=args.num_epochs,
lr_type=args.lr,
initial_lr=args.initial_lr,
n_cores=args.num_cores,
regularizer=1. / n,
take_k=args.k,
take_top=args.model is 'top',
with_memory=with_memory,
estimate='final')
print('start experiment')
m = LogisticParallelSGD(model)
iters, timers, losses = m.fit(X,
y,
num_features=X.shape[1],
num_samples=X.shape[0])
pickle_it(iters, 'iters', directory)
pickle_it(timers, 'timers', directory)
pickle_it(losses, 'losses', directory)
print('results saved in "{}"'.format(directory))
MAIN_SETTINGS = {'HEAT_THRESHOLD': 0.8,
'TEMP_TOLERANCE': 1,
'HUM_TOLERANCE': 10,
'MAIN_SENSOR': 'mock_location1'}
if __name__ == '__main__':
from utils import pickle_it
pickle_it(MAIN_SETTINGS, "main_settings")
print("pickled: " + str(MAIN_SETTINGS))
