def main():
c: Configuration = Configuration()
# set configuration params
c.simulator_write_debug_level = DebugLevel.HIGH
main_runner = MainRunner(c)
analyzer = Analyzer(main_runner.main_services)
analyzer.analyze_algorithms()
def main():
opt = OptionParser()
opt.parser.add_argument('target', help='target file path', type=str)
opt_args = opt.get_commandline_arguments()
LOG.reset_logger(opt_args)
target = opt_args.target
person_names = []
is_debug = opt_args.debug
analyzer = Analyzer()
analyzer.execute(target, person_names, is_debug)
print('>> FINISHED!!')
return 0
def analyze_movie(file):
analyzer = Analyzer(file)
video = Video.objects.create(file=file)
for t, frame in enumerate(analyzer.get_histogram()):
r, g, b = frame
print("Frame:" + str(t))
for num_bin in range(0, len(r)):
Histogram.objects.create(video=video, t=t, num_bin=num_bin, bin_value=r[num_bin], color=Histogram.RED)
Histogram.objects.create(video=video, t=t, num_bin=num_bin, bin_value=g[num_bin], color=Histogram.GREEN)
Histogram.objects.create(video=video, t=t, num_bin=num_bin, bin_value=b[num_bin], color=Histogram.BLUE)
print("Finished")
def run(self):
if self.main_services.settings.generator:
Generator.main(self)
elif self.main_services.settings.trainer:
Trainer.main(self)
elif self.main_services.settings.analyzer:
Analyzer.main(self)
elif self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
if self.main_services.settings.clear_cache:
self.main_services.resources.cache_dir.clear()
def __init__(self, gdrive_certificat_path):
self.server = Flask(__name__,
static_url_path='',
static_folder='frontend/static')
self.socketio = SocketIO(self.server)
self.session_maker = run_orm()
self.analyzer = Analyzer()
self.drive = Drive(gdrive_certificat_path)
self.server.route('/', methods=['GET'])(self.index)
self.server.route('/upload', methods=['POST'])(self.upload_task)
self.server.route('/end_check', methods=['POST'])(self.end_check)
def main(argv):
print("#" * 64)
print("##" + " " * 23 + "KICKASS DISASS" + " " * 23 + "##")
print("##" + " " * 60 + "##")
print("##" + "Powered by Capstone {:d}.{:d} ##".format(
CS_API_MAJOR, CS_API_MINOR).rjust(62))
print("#" * 64 + "\n")
if len(argv) <= 1:
raise Exception("Usage: python3 main.py binary")
loaders = load_modules()
# open file and check for loaders that are able to disass the file
with open(argv[1], 'rb') as f:
valid_loaders = [l for l in loaders if l.accept_file(f)]
if not valid_loaders:
print("+ No loaders recognize the file")
sys.exit(-1)
if len(valid_loaders) >= 2:
loader = prompt_user_for_loader(valid_loaders)
else:
loader = valid_loaders[0]
# launch the actual analysis
anal = Analyzer(loader, f)
def _analyze_and_output(self, src: Story, person_names: list,
is_debug: bool) -> ResultData:
# serialize and compile as text
mode = CompileMode.NOVEL_TEXT
fmode = FormatMode.DEFAULT
LOG.info('Serialize for Analyzer')
result = assertion.is_instance(Serializer().execute(src, mode),
ResultData)
if not result.is_succeeded:
return result
tmp = assertion.is_instance(result.data, CodeList)
LOG.info('Validate for Analyzer')
result = assertion.is_instance(Validater().execute(tmp), ResultData)
if not result.is_succeeded:
return result
tmp = assertion.is_instance(result.data, CodeList)
LOG.info('Compile for Analyzer')
result = assertion.is_instance(
Compiler().execute(tmp, mode, {}, False, False), ResultData)
if not result.is_succeeded:
return result
tmp = assertion.is_instance(result.data, RawData)
LOG.info('Format for Analyzer')
result = assertion.is_instance(Formatter().execute(tmp, fmode),
ResultData)
if not result.is_succeeded:
return result
tmp = assertion.is_instance(result.data, TextList)
LOG.info('RUN: call Analyzer')
result = Analyzer().execute(tmp, person_names, is_debug)
return ResultData([], True, None)
def _test(_dir):
sample_names = get_sample_names(_dir)
monitor_data = []
first_http_response = None
for sample_name in sample_names:
http_response = HttpResponse(open("%s/%s" % (_dir, sample_name)).read())
if first_http_response == None:
first_http_response = http_response
parsed_data = parser.parse_ms_monitor_result(http_response.body)
parsed_data["sample_time"] = datetime.fromtimestamp(int(sample_name.split(".")[0]))
monitor_data.append(parsed_data)
az = Analyzer()
result = az.do_analyze(monitor_data[0]["sample_time"], first_http_response, monitor_data[1:])
return result
def test_00():
url = "http://10.33.0.57/tag_live_monitor/tvie/zhso/xfw"
stream_id = 0
_map = {}
analyzer = Analyzer()
http_client = MediaServerAdminQueryer(analyzer, stream_id, url, _map)
asyncore.loop(map=_map)
print http_client.response.body
def _test(_dir):
sample_names = get_sample_names(_dir)
monitor_data = []
first_http_response = None
for sample_name in sample_names:
http_response = HttpResponse(
open("%s/%s" % (_dir, sample_name)).read())
if first_http_response == None:
first_http_response = http_response
parsed_data = parser.parse_ms_monitor_result(http_response.body)
parsed_data['sample_time'] = datetime.fromtimestamp(
int(sample_name.split('.')[0]))
monitor_data.append(parsed_data)
az = Analyzer()
result = az.do_analyze(monitor_data[0]['sample_time'], first_http_response,
monitor_data[1:])
return result
def __init__(self):
# Initialising fields
self.parameters = {}
# Loading config
self.config = self.loadConfig()
self.analyzer = Analyzer(self.config)
global analyzer_global
analyzer_global = self.analyzer
# Selecting mode
if len(sys.argv) > 1:
self.parse_args()
self.analyzer.error_occurred.connect(self.print_error)
filename = self.parameters["input"]
if not os.path.exists(filename):
self.print_error("File {} does not exist".format(filename))
sys.exit()
try:
text = self.analyzer.load_file(self.parameters["input"])
if not self.analyzer.isTextValid(text):
self.print_error(
"File {} does not contain valid text".format(filename))
sys.exit()
except Exception as e:
self.print_error("Error loading file {}:\n{}".format(
filename, e))
sys.exit()
result = self.analyzer.analyze(text, self.parameters)
if result is None:
self.print_error("Unknown error occurred")
sys.exit()
result.save_to_file(self.parameters["output"],
self.parameters["threshold"],
n_digits=5)
sys.exit(0)
else:
show_splashscreen()
self.ui = GUI(analyzer=self.analyzer, config=self.config)
def main(argv):
if len(argv) < 2:
raise Exception("Usage: python3 main.py filename (mode)")
loaders = load_modules()
# open file and check for loaders that are able to disass the file
with open(argv[1], 'rb') as f:
valid_loaders = [l for l in loaders if l.accept_file(f)]
# TODO check whether elf or not
exe = ELFExecutable(argv[1])
exe._extract_symbol_table()
if not valid_loaders:
print("+ No loaders recognize the file")
sys.exit(-1)
if len(valid_loaders) >= 2:
loader = prompt_user_for_loader(valid_loaders)
else:
loader = valid_loaders[0]
fname = os.path.basename(f.name)
if len(argv) >= 3:
os.system("python3 init_main.py " + fname + " disasm_all_mode")
else:
os.system("python3 init_main.py " + fname)
if not os.path.isdir("input/" + fname):
print("Input file(input/" + fname + ") does not exist")
sys.exit()
"""
print("#" * 64)
print("##" + " " * 23 + "KICKASS DISASS" + " " * 23 + "##")
print("##" + " " * 60 + "##")
print(
"##" +
"Powered by Capstone {:d}.{:d} ##".format(
CS_API_MAJOR,
CS_API_MINOR).rjust(62))
print("#" * 64 + "\n")
"""
if len(argv) >= 3:
anal = Analyzer(loader, exe, f, argv[2])
else:
anal = Analyzer(loader, exe, f)
anal.start()
class MainApp(Tk):
# analyzer is completely isolated from UI, check the '/analyzer' module
analyzer = Analyzer()
def __init__(self):
super(MainApp, self).__init__()
self.setup()
self.top_frame = TopFrame(self)
self.body_frame = BodyFrame(self)
def setup(self):
self.title('GitHub User Performance Analyzer')
self.configure(bg='white')
self.minsize(0, 500)
self.resizable(0, 0)
self.grid_columnconfigure(0, weight=1)
self.grid_rowconfigure(1, weight=1)
@staticmethod
def confirm(question, icon='question'):
result = messagebox.askquestion("Confirmation", question, icon=icon)
return result == 'yes'
def setUp(self):
self.analyzer = Analyzer()
self.supervisor = Supervisor()
self.test_data = read_test_data()
class PlagiarismTest(unittest.TestCase):
def setUp(self):
self.analyzer = Analyzer()
self.supervisor = Supervisor()
self.test_data = read_test_data()
def tearDown(self):
pass
def test_find_similarity_groups_by_rows(self):
similarity_matrix = np.asarray([[0, 6, 12, 8, 1], [5, 0, 4, 14, 9],
[11, 5, 0, 7, 3], [7, 12, 8, 0, 14],
[2, 10, 4, 12, 0]])
result = self.analyzer.find_similarity_groups_by_rows(
similarity_matrix)
self.assertEqual(result[1], 1)
self.assertEqual(result[2], 2)
self.assertEqual(result[3], 1)
self.assertEqual(result[4], 3)
def test_find_similarity_groups_by_mean_group_similarity(self):
similarity_matrix = np.asarray([[0, 6, 12, 8, 1], [5, 0, 4, 14, 9],
[11, 5, 0, 7, 3], [7, 12, 8, 0, 14],
[2, 10, 4, 12, 0]])
result = self.analyzer.find_similarity_groups_by_mean_group_similarity(
similarity_matrix)
self.assertEqual(result[1], 1)
self.assertEqual(result[2], 2)
self.assertEqual(result[3], 1)
self.assertEqual(result[4], 1)
def test_analyze_test_via_supervisor(self):
text = self.test_data["observer_test_text"]
result = self.supervisor.markup(text)
self.assertEqual(result.text, text)
self.assertEqual(result.num_letters, 377)
self.assertEqual(result.num_sentences, 10)
self.assertEqual(result.num_words, 62)
self.assertEqual(len(result.tokens), 62)
self.assertEqual(len(result.morph_tokens), 62)
def test_create_plagiarism_matrix(self):
essays = self.test_data["plagiarism_test_essays"]
essays = [self.supervisor.markup(essay) for essay in essays]
matrix, coincidences = plagiarism.create_plagiarism_matrix(essays)
self.assertEqual((matrix.shape == np.array([len(essays),
len(essays)])).all(), True)
# для полюсного эссе номер 1
self.assertEqual(matrix[0, 1] == 100,
True) # эссе имеет сходство по 100% предолжений
self.assertEqual(matrix[0, 2] > 33 and matrix[0, 2] < 65,
True) # эссе имеет сходство по 50% предолжений
self.assertEqual(matrix[0, 3] == 0, True)
# для полюсного эссе номер 2
self.assertEqual(matrix[3, 0] == 0, True)
self.assertEqual(matrix[3, 1] == 0,
True) # эссе имеет сходство по 0% предолжений
self.assertEqual(matrix[3, 2] > 35 and matrix[3, 2] < 65,
True) # эссе имеет сходство по 50% предолжений
def test_similarity_matrix(self):
essays = self.test_data["similarity_test_essays"]
essays = [self.supervisor.markup(essay) for essay in essays]
result = similarity.create_similarity_matrix(essays)
self.assertEqual((result.shape == np.array([len(essays),
len(essays)])).all(), True)
# для эссе номер 1
self.assertEqual(result[0, 1] > 13, True) # эссе на ту же тему
self.assertEqual(result[0, 2] > 4 and result[0, 2] < 10,
True) # эссе на пересекающуюся тему
self.assertEqual(result[0, 3] < 4, True) # эссе на обособленную тему
def test_read_from_file(self):
expected_text = self.test_data["lecture_reader_expected_text"].strip()
presentation_text = lecture_reader.read_from_file(
TEST_PRESENTATION_PATH).strip()
self.assertEqual(presentation_text, expected_text)
def test_analyzer(self):
lecture = Presentation(TEST_PRESENTATION_PATH)
lecture_text = read_from_presentation(lecture)
essays = self.test_data["analyzer_test_essays"]
pd_essays = pd.DataFrame(data=essays, columns=["text"])
report = self.analyzer.analyze(lecture_text, pd_essays)
self.assert_lecture(self.test_data["lecture_reader_expected_text"],
report.lecture, 176)
self.assert_essay(essays[0], report.essays[0], GradeType.FAIL, 1,
[LabelType.FAIL, LabelType.LECTURE_PLAGIARISM], 302)
self.assert_essay(essays[1], report.essays[1], GradeType.SUCCESS, 1,
[LabelType.SUCCESS], 338)
self.assert_essay(essays[2], report.essays[2], GradeType.FAIL, 2,
[LabelType.FAIL], 246)
def assert_lecture(self, text, lecture, num_words):
self.assertEqual(lecture.text, text)
self.assertEqual(lecture.statistic.num_letters, len(text))
self.assertEqual(lecture.statistic.num_sentences,
len(sent_tokenize(text)))
self.assertEqual(lecture.statistic.num_words, num_words)
def assert_essay(self, text, essay, grade, group, labels, num_words):
self.assertEqual(essay.text, text)
self.assertEqual(essay.statistic.num_letters, len(text))
self.assertEqual(essay.statistic.num_sentences,
len(sent_tokenize(text)))
self.assertEqual(essay.statistic.num_words, num_words)
self.assertEqual(essay.grade, grade)
self.assertEqual(essay.group, group)
self.assertEqual(len(labels), len(essay.labels))
for label in essay.labels:
if label.type in labels:
labels.remove(label.type)
else:
self.assertEqual(True, False)
self.assertEqual(len(labels), 0)
def test_analyzer_incorrect_essay_list(self):
lecture = Presentation(TEST_PRESENTATION_PATH)
essays = self.test_data["analyzer_test_essays"]
pd_essays = pd.DataFrame(data=essays, columns=["roles"])
self.assertRaises(NotFoundEssayColumn, self.analyzer.analyze, lecture,
pd_essays)
def __init__(self, strategy_data):
self.strategy_data = strategy_data
self.analyzer = Analyzer(self)
self.email_service = EmailService()
class Strategy:
def __init__(self, strategy_data):
self.strategy_data = strategy_data
self.analyzer = Analyzer(self)
self.email_service = EmailService()
def initialize(self, context):
context.algo_id = self.strategy_data.get('algo_id')
context.live_trading = self.strategy_data.get('live_trading')
self.strategy_data.get('initialize')(context)
if self.strategy_data.get('live_trading', False) is False:
self.analyzer.initialize()
else:
self.email_service.initialize()
def SendMessage(self, subject, message):
if self.strategy_data.get('live_trading', False) is True:
self.email_service.SendMessage(subject, message)
def handle_data(self, context, data):
self.strategy_data.get('handle_data')(context, data)
if self.strategy_data.get('live_trading', False) is False:
self.analyzer.handle_data(context)
def analyze(self, context, data):
print("Analyse method got called")
self.strategy_data.get('analyze')(context, data)
if self.strategy_data.get('live_trading', False) is False:
self.analyzer.finalize()
else:
algo_id = self.strategy_data.get('algo_id')
db_engine = create_engine('sqlite:///{}'.format(
os.path.join(str(Path.home()), 'algodb.db')))
prev_date_sql = "select date from prev_run_date where algo_id={}".format(
algo_id)
prev_run_date = pd.read_sql(prev_date_sql, db_engine)['date'][0]
run_date = str(context.datetime.date())
prev_pos_sql = "select holding_name, quantity, buy_price, last_price from daily_holdings " \
"where algo_id={} and date='{}'".format(algo_id, prev_run_date)
prev_pos = pd.read_sql(prev_pos_sql, db_engine)
if prev_pos.empty:
prev_pos_list = []
else:
prev_pos_list = list(prev_pos['holding_name'])
prev_pos.set_index('holding_name', inplace=True)
curr_positions = context.portfolio.positions.values()
stock_email_columns = [
'Holding', 'Shares', 'Buy Price', 'Yest Price',
'Current Price', 'Dollar Gain Today', 'Pct Gain Today',
'Dollar Gain Net', 'Pct Gain Net', 'Market Value'
]
stock_email = pd.DataFrame(columns=stock_email_columns)
sold_list = []
for position in list(curr_positions):
if position.sid.symbol in prev_pos_list:
if position.amount == 0:
sold_list.append(position)
continue
prev_stock_pos = prev_pos.loc[position.sid.symbol]
gain_today = position.last_sale_price - prev_stock_pos[
'last_price']
pct_gain_today = str(
round(
(gain_today / prev_stock_pos['last_price']) * 100,
4)) + ' %'
gain_total = position.last_sale_price - position.cost_basis
pct_gain_total = str(
round((gain_total / position.cost_basis) * 100,
4)) + ' %'
stock_email.loc[position.asset] = [
position.asset.symbol, position.amount,
round(position.cost_basis, 4),
prev_stock_pos['last_price'], position.last_sale_price,
gain_today, pct_gain_today, gain_total, pct_gain_total,
position.amount * position.last_sale_price
]
else:
stock_email.loc[position.asset] = [
position.asset.symbol, position.amount,
round(position.cost_basis,
4), '-', position.last_sale_price, '-', '-', '-',
'-', position.amount * position.last_sale_price
]
portfolio = context.portfolio
# stock_email = stock_email.join(pd.DataFrame(portfolio.current_portfolio_weights, columns=['Weightage']))
# stock_email['Weightage'] = round(stock_email['Weightage'] * 100, 4).astype(str) + ' %'
port_email = pd.Series([
round(portfolio.portfolio_value, 4),
round(portfolio.pnl, 4),
str(
round(
portfolio.pnl /
(portfolio.portfolio_value - portfolio.pnl), 4)) +
' %',
round(portfolio.cash, 4),
round(portfolio.positions_value, 4)
],
index=[
'Portfolio Value', 'Net Gain',
'Percent Net Gain', 'Cash Value',
'Position Value'
])
message = "<p><h3>Holdings Summary</h3></p>" + stock_email.to_html(index=False) \
+ "<p><h3>Portfolio Summary</h3></p>" + pd.DataFrame(port_email).T.to_html(index=False)
subject = '{} : Daily Summary - {}'.format(
self.strategy_data.get('algo_name'), run_date)
self.email_service.SendNotifications(subject, message)
prev_run_update_sql = "update prev_run_date set date='{}' where algo_id={}".format(
run_date, algo_id)
with db_engine.connect() as connection:
try:
for position in list(curr_positions):
if position in sold_list:
continue
insert_holding_sql = "Insert into daily_holdings (date, algo_id, holding_name, quantity, " \
"buy_price, last_price) values ('{}',{},'{}',{},{},{})"\
.format(run_date, algo_id, position.sid.symbol, position.amount,
round(position.cost_basis, 4), position.last_sale_price)
connection.execute(insert_holding_sql)
connection.execute(prev_run_update_sql)
except Exception as e:
print(e)
self.strategy_data.get('after_trading_end')(context, data)
def before_trading_start(self, context, data):
self.strategy_data.get('before_trading_start')(context, data)
if self.strategy_data.get('live_trading', False) is False:
self.analyzer.before_trading_start()
def run_algorithm(self):
live_trading = self.strategy_data.get('live_trading', False)
# If live_trading true, trade with Virtual broker using database prices, else run normal backtest
# database prices are updated from master_algo
kwargs = {
'start': self.strategy_data.get('start'),
'end': self.strategy_data.get('end'),
'initialize': self.initialize,
'handle_data': self.handle_data,
'analyze': self.analyze,
'before_trading_start': self.before_trading_start,
'bundle': 'quandl',
'capital_base': self.strategy_data.get('capital_base'),
'tws_uri': self.strategy_data.get('tws_uri'),
'live_trading': live_trading
}
run_algo_thread = threading.Thread(target=run_algorithm, kwargs=kwargs)
run_algo_thread.start()
if self.strategy_data.get('live_trading', False) is False:
self.analyzer.show_plot()
sys.exit(self.analyzer.app.exec_())
run_algo_thread.join()
threshold = self.main_window.opt_bar.threshold.value()
result.save_to_file(filename=filename, threshold=threshold, n_digits=5)
except Exception as e:
self.error_occurred.emit("Не удалось экспортировать результат")
def on_analysis_completed(self, result: Predict):
params = self.main_window.opt_bar.options_to_dict()
self.main_window.result_widget.show_output(result, params)
def show_splashscreen():
splash = qw.QSplashScreen(QPixmap(":/Splash_email_v2.png"), qc.Qt.WindowStaysOnTopHint)
time = qc.QTime()
splash.show()
time.start()
while time.elapsed() <= 3000:
pass
splash.finish(None)
if __name__ == '__main__':
from PyQt5.QtWidgets import QApplication
import sys
from configparser import ConfigParser
a = QApplication(sys.argv)
c = ConfigParser()
c.read("../config.ini")
g = GUI(config=c, analyzer=Analyzer(c))
a.exec()
def run(argv):
if argv == 'downloader':
Downloader.run()
elif argv == 'analyzer':
Analyzer().run()
import datetime
from eye.eye import Eye
from analyzer.analyzer import Analyzer
from sender.sender import Sender
from config.config import Config
configuration = Config()
eye = Eye()
analyzer = Analyzer()
sender = Sender()
frame = eye.capture()
while True:
data = analyzer.analyze(frame)
sender.send(data)
frame = eye.capture()
def get_analyzer(holiday_plan_id):
if HolidayPlan.query.filter_by(id=holiday_plan_id).scalar() is None:
return Analyzer()
analyzer_as_json = \
HolidayPlan.query.filter_by(id=holiday_plan_id).first().analyzer_as_json
return Analyzer().from_json(analyzer_as_json)
def run_multiple(self):
#Two options, generator
if self.main_services.settings.generator and self.main_services.settings.trainer:
Generator.main(self)
Trainer.main(self)
elif self.main_services.settings.generator and self.main_services.settings.analyzer:
Generator.main(self)
Analyzer.main(self)
elif self.main_services.settings.generator and self.main_services.settings.load_simulator:
Generator.main(self)
simulator: Simulator = Simulator(self.main_services)
simulator.start()
#Three options, generator
elif self.main_services.settings.generator and self.main_services.settings.trainer and self.main_services.settings.analyzer:
Generator.main(self)
Trainer.main(self)
Analyzer.main(self)
elif self.main_services.settings.generator and self.main_services.settings.trainer and self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
Generator.main(self)
Trainer.main(self)
#Four options
elif self.main_services.settings.generator and self.main_services.settings.trainer and self.main_services.settings.analyzer and self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
Generator.main(self)
Trainer.main(self)
Analyzer.main(self)
#Trainer
elif self.main_services.settings.trainer and self.main_services.settings.analyzer:
Trainer.main(self)
Analyzer.main(self)
elif self.main_services.settings.trainer and self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
Trainer.main(self)
elif self.main_services.settings.trainer and self.main_services.settings.analyzer and self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
Trainer.main(self)
Analyzer.main(self)
#Analyzer
elif self.main_services.settings.analyzer and self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
Analyzer.main(self)
#Singles
elif self.main_services.settings.generator:
Generator.main(self)
elif self.main_services.settings.trainer:
Trainer.main(self)
elif self.main_services.settings.analyzer:
Analyzer.main(self)
elif self.main_services.settings.load_simulator:
simulator: Simulator = Simulator(self.main_services)
simulator.start()
if self.main_services.settings.clear_cache:
self.main_services.resources.cache_dir.clear()
args = vars(parser.parse_args())
print("\n\n")
print(datetime.datetime.now().strftime('%d %b %G %I:%M%p'))
print("\n")
params = dict()
params['index_by'] = 'titles'
params['enable_filtering'] = False
params['resources'] = "./res/"
params['similarity_threshold'] = .4
params['server_mode'] = False
# parse argument
if args['server']:
params['server_mode'] = True
def print_configuration(params):
print("Configuration:")
print("\tIndex by: %s" % params['index_by'])
print("\tFiltering enabled: %s" % repr(params['enable_filtering']))
print("\tResource folder: %s" % params['resources'])
print("\tSimilarity threshold: %.2f" % params['similarity_threshold'])
print("\tServer mode enabled: %s" % repr(params['server_mode']))
print_configuration(params)
a = Analyzer(params)
a.begin()
import sys
from parser.parser import Parser
from compiler.compiler import Compiler
from analyzer.analyzer import Analyzer
BUILTINS_FILE = "src/runtime/builtins_signatures"
if __name__ == '__main__':
if len(sys.argv) != 3:
puts(
"Error! Provide source and target files as argunents to this script"
)
exit(1)
code_file = sys.argv[1]
target_file = sys.argv[2]
builtins = BUILTINS_FILE
ast = Parser(code_file, [builtins]).parse()
if Analyzer().validate(ast):
Compiler(target_file).compile(ast)
def scan_file(path):
path = os.path.normpath(path)
tree = build_ast(path)
analyzer = Analyzer(path)
analyzer.visit(tree)
analyzer.report()
class EstimatorServer():
def __init__(self, gdrive_certificat_path):
self.server = Flask(__name__,
static_url_path='',
static_folder='frontend/static')
self.socketio = SocketIO(self.server)
self.session_maker = run_orm()
self.analyzer = Analyzer()
self.drive = Drive(gdrive_certificat_path)
self.server.route('/', methods=['GET'])(self.index)
self.server.route('/upload', methods=['POST'])(self.upload_task)
self.server.route('/end_check', methods=['POST'])(self.end_check)
def item_callback(self, current_item_id, count_items):
self.socketio.emit(
'changed-report-status',
json.dumps({
'status':
'handling',
'description':
f'Обработано ответов из архива {current_item_id} из {count_items}'
}))
def load_lecture_file(self, file):
extension = re.findall(r'\.\w+$', file.filename)[0]
if extension == '.pptx':
lecture = Presentation(file)
return read_from_presentation(lecture)
if extension == '.txt':
wrapper = io.TextIOWrapper(file, encoding='utf-8')
return wrapper.read()
if extension == '.docx':
lecture = docx.Document(file)
fullText = []
for para in lecture.paragraphs:
fullText.append(para.text)
return ' '.join(fullText)
raise NotSupportLectureExtensionType(extension)
def load_essays_file(self, file):
extension = re.findall(r'\.\w+$', file.filename)[0]
if extension == '.csv':
return pd.read_csv(file)
if extension == '.xlsx':
return pd.read_excel(file)
if extension == '.zip':
archive = zipfile.ZipFile(file, 'r')
return download_archive(self.drive, archive, self.item_callback)
raise NotSupportEssayExtensionType(extension)
def index(self):
return self.server.send_static_file("index.html")
def upload_task(self):
try:
if 'lecture' not in request.files or 'essays' not in request.files:
return redirect(request.url)
lecture = request.files['lecture']
essays = request.files['essays']
if lecture.filename == '' or essays.filename == '':
return redirect(request.url)
self.socketio.emit(
'changed-report-status',
json.dumps({
'status': 'handling',
"description": "Обработка файлов"
}))
lecture = self.load_lecture_file(lecture)
essays = self.load_essays_file(essays)
essays = essays.dropna(axis=0)
self.socketio.emit(
'changed-report-status',
json.dumps({
'status': 'handling',
"description": "Оценка эссе"
}))
report = self.analyzer.analyze(lecture, essays)
report_schema = ReportSchema()
session = self.session_maker()
session.add(report)
session.commit()
self.socketio.emit('changed-report-status',
json.dumps({'status': 'handled'}))
return report_schema.dump(report)
except Exception as e:
print(e)
print(traceback.print_exc())
if type(e) == NotFoundEssayColumn or type(e) == NotSupportEssayExtensionType\
or type(e) == NotSupportLectureExtensionType:
return json.dumps({"status": "error", "text": str(e)}), 500
else:
return json.dumps({
"status": "error",
"text": "Ошибка оценки загруженных эссе"
}), 500
def end_check(self):
try:
session = self.session_maker()
report_schema = ReportSchema()
json_data = json.loads(request.data)
report = report_schema.load(data=json_data, session=session)
if report.lecture is None:
raise Exception("Received incorrect data. Report not found")
session.commit()
return json.dumps({
"status":
"success",
"text":
"Результаты проверки эссе успешно сохранены"
})
except Exception as e:
print(e)
print(traceback.print_exc())
session.close()
return json.dumps({
"status": "error",
"text": "Ошибка сохранения резултатов проверки"
}), 500
def start(self):
self.socketio.run(self.server, host='localhost', port=5000)
def main():
parser = argparse.ArgumentParser(description="Gaussian Input Output HMM")
parser.add_argument('--data',
type=str,
default='./dataset/ptb/',
help='location of the data corpus')
parser.add_argument('--batch', type=int, default=256)
parser.add_argument('--optim', choices=['sgd', 'adam'], default='adam')
parser.add_argument('--lr', type=float, default=0.001)
parser.add_argument('--lr_decay',
type=float,
default=0.999995,
help='Decay rate of learning rate')
parser.add_argument('--amsgrad', action='store_true', help='AMD Grad')
parser.add_argument('--weight_decay',
type=float,
default=0.001,
help='weight for l2 norm decay')
parser.add_argument('--warmup_steps',
type=int,
default=0,
metavar='N',
help='number of steps to warm up (default: 0)')
parser.add_argument('--var_scale', type=float, default=1.0)
parser.add_argument('--log_dir',
type=str,
default='./output/' +
datetime.datetime.now().strftime("%Y-%m-%d_%H%M%S") +
"/")
parser.add_argument('--dim', type=int, default=10)
parser.add_argument('--gpu', action='store_true')
parser.add_argument('--random_seed', type=int, default=10)
parser.add_argument('--in_mu_drop', type=float, default=0.0)
parser.add_argument('--in_cho_drop', type=float, default=0.0)
parser.add_argument('--t_mu_drop', type=float, default=0.0)
parser.add_argument('--t_cho_drop', type=float, default=0.0)
parser.add_argument('--out_mu_drop', type=float, default=0.0)
parser.add_argument('--out_cho_drop', type=float, default=0.0)
parser.add_argument('--trans_cho_method',
type=str,
choices=['random', 'wishart'],
default='random')
parser.add_argument(
'--input_cho_init',
type=float,
default=0.0,
help=
'init method of input cholesky matrix. 0 means random. The other score means constant'
)
parser.add_argument(
'--trans_cho_init',
type=float,
default=1.0,
help='init added scale of random version init_cho_init')
parser.add_argument(
'--output_cho_init',
type=float,
default=0.0,
help=
'init method of output cholesky matrix. 0 means random. The other score means constant'
)
# i_comp_num = 1, t_comp_num = 1, o_comp_num = 1, max_comp = 1,
parser.add_argument('--input_comp_num',
type=int,
default=1,
help='input mixture gaussian component number')
parser.add_argument('--tran_comp_num',
type=int,
default=1,
help='transition mixture gaussian component number')
parser.add_argument('--output_comp_num',
type=int,
default=1,
help='output mixture gaussian component number')
parser.add_argument(
'--threshold',
type=float,
default=1.0,
help=
'pruning hyper-parameter, greater than 1 is max component, less than 1 is max value'
)
parser.add_argument('--unk_replace',
type=float,
default=0.0,
help='The rate to replace a singleton word with UNK')
parser.add_argument('--tran_weight', type=float, default=0.0001)
parser.add_argument('--input_weight', type=float, default=0.0)
parser.add_argument('--output_weight', type=float, default=0.0)
parser.add_argument('--emission_cho_grad', type=bool, default=False)
parser.add_argument('--transition_cho_grad', type=bool, default=True)
parser.add_argument('--decode_cho_grad', type=bool, default=False)
parser.add_argument('--gaussian_decode', action='store_true')
parser.add_argument('--analysis', action='store_true')
parser.add_argument('--sep_normalize', type=float, default=0.01)
args = parser.parse_args()
np.random.seed(args.random_seed)
torch.manual_seed(args.random_seed)
random.seed(args.random_seed)
log_dir = args.log_dir
# setting optimizer
optim = args.optim
batch_size = args.batch
# optim = 'sgd'
lr = args.lr
lr_decay = args.lr_decay
warmup_steps = args.warmup_steps
amsgrad = args.amsgrad
weight_decay = args.weight_decay
root = args.data
in_mu_drop = args.in_mu_drop
in_cho_drop = args.in_cho_drop
t_mu_drop = args.t_mu_drop
t_cho_drop = args.t_cho_drop
out_mu_drop = args.out_mu_drop
out_cho_drop = args.out_cho_drop
tran_cho_method = args.trans_cho_method
input_cho_init = args.input_cho_init
trans_cho_init = args.trans_cho_init
output_cho_init = args.output_cho_init
input_num_comp = args.input_comp_num
tran_num_comp = args.tran_comp_num
output_num_comp = args.output_comp_num
threshold = args.threshold
unk_replace = args.unk_replace
normalize_weight = [
args.tran_weight, args.input_weight, args.output_weight
]
gaussian_decode = args.gaussian_decode
sep_normalize = args.sep_normalize
analysis = args.analysis
EMISSION_CHO_GRAD = args.emission_cho_grad
TRANSITION_CHO_GRAD = args.transition_cho_grad
DECODE_CHO_GRAD = args.decode_cho_grad
if not os.path.exists(log_dir):
os.makedirs(log_dir)
# save parameter
save_parameter_to_json(log_dir, vars(args))
logger = get_logger('Sequence-Labeling')
change_handler(logger, log_dir)
# logger = LOGGER
logger.info(args)
device = torch.device('cuda') if args.gpu else torch.device('cpu')
# Loading data
logger.info('Load PTB data....')
alphabet_path = os.path.join(root, 'alphabets')
train_path = os.path.join(root, 'train.conllu')
dev_path = os.path.join(root, 'dev.conllu')
test_path = os.path.join(root, 'test.conllu')
word_alphabet, char_alphabet, pos_alphabet, type_alphabet = conllx_data.create_alphabets(
alphabet_path,
train_path,
data_paths=[dev_path, test_path],
embedd_dict=None,
max_vocabulary_size=1e5,
min_occurrence=1)
train_dataset = conllx_data.read_bucketed_data(train_path, word_alphabet,
char_alphabet, pos_alphabet,
type_alphabet)
num_data = sum(train_dataset[1])
dev_dataset = conllx_data.read_data(dev_path, word_alphabet, char_alphabet,
pos_alphabet, type_alphabet)
test_dataset = conllx_data.read_data(test_path, word_alphabet,
char_alphabet, pos_alphabet,
type_alphabet)
logger.info("Word Alphabet Size: %d" % word_alphabet.size())
logger.info("Character Alphabet Size: %d" % char_alphabet.size())
logger.info("POS Alphabet Size: %d" % pos_alphabet.size())
ntokens = word_alphabet.size()
nlabels = pos_alphabet.size()
# init analyzer
if analysis:
analyzer = Analyzer(word_alphabet=word_alphabet,
pos_alphabet=pos_alphabet)
else:
analyzer = None
# build model
if threshold >= 1.0:
model = MixtureGaussianSequenceLabeling(
dim=args.dim,
ntokens=ntokens,
nlabels=nlabels,
t_cho_method=tran_cho_method,
t_cho_init=trans_cho_init,
in_cho_init=input_cho_init,
out_cho_init=output_cho_init,
in_mu_drop=in_mu_drop,
in_cho_drop=in_cho_drop,
t_mu_drop=t_mu_drop,
t_cho_drop=t_cho_drop,
out_mu_drop=out_mu_drop,
out_cho_drop=out_cho_drop,
i_comp_num=input_num_comp,
t_comp_num=tran_num_comp,
o_comp_num=output_num_comp,
max_comp=int(threshold),
gaussian_decode=gaussian_decode)
else:
model = ThresholdPruningMGSL(dim=args.dim,
ntokens=ntokens,
nlabels=nlabels,
t_cho_method=tran_cho_method,
t_cho_init=trans_cho_init,
in_cho_init=input_cho_init,
out_cho_init=output_cho_init,
in_mu_drop=in_mu_drop,
in_cho_drop=in_cho_drop,
t_mu_drop=t_mu_drop,
t_cho_drop=t_cho_drop,
out_mu_drop=out_mu_drop,
out_cho_drop=out_cho_drop,
i_comp_num=input_num_comp,
t_comp_num=tran_num_comp,
o_comp_num=output_num_comp,
threshold=threshold,
gaussian_decode=gaussian_decode)
# model = RNNSequenceLabeling("LSTM", ntokens=ntokens, nlabels=nlabels, ninp=args.dim, nhid=args.dim, dropout=in_mu_drop)
# model = WeightIOHMM(vocab_size=ntokens, nlabel=nlabels, num_state=100)
model.to(device)
logger.info('Building model ' + model.__class__.__name__ + '...')
# optimizer = optim.Adam(model.parameters(), lr=lr)
parameters_need_update = filter(lambda p: p.requires_grad,
model.parameters())
optimizer, scheduler = get_optimizer(parameters_need_update,
optim,
lr,
amsgrad,
weight_decay,
lr_decay=lr_decay,
warmup_steps=warmup_steps)
# depend on dev ppl
best_epoch = (-1, 0.0, 0.0)
num_batches = num_data // batch_size + 1
def train(best_epoch, thread=6):
epoch = 0
while epoch - best_epoch[0] <= thread:
epoch_loss = 0
num_back = 0
num_words = 0
num_insts = 0
model.train()
for step, data in enumerate(
iterate_data(train_dataset,
batch_size,
bucketed=True,
unk_replace=unk_replace,
shuffle=True)):
# for j in tqdm(range(math.ceil(len(train_dataset) / batch_size))):
optimizer.zero_grad()
# samples = train_dataset[j * batch_size: (j + 1) * batch_size]
words, labels, masks = data['WORD'].to(device), data['POS'].to(
device), data['MASK'].to(device)
loss = 0.0
if threshold >= 1.0:
# sentences, labels, masks, revert_order = standardize_batch(samples)
loss = model.get_loss(words,
labels,
masks,
normalize_weight=normalize_weight,
sep_normalize=sep_normalize)
else:
for i in range(batch_size):
loss += model.get_loss(
words[i],
labels[i],
masks[i],
normalize_weight=normalize_weight,
sep_normalize=sep_normalize)
# loss = model.get_loss(words, labels, masks)
loss.backward()
optimizer.step()
scheduler.step()
epoch_loss += (loss.item()) * words.size(0)
num_words += torch.sum(masks).item()
num_insts += words.size()[0]
if step % 10 == 0:
torch.cuda.empty_cache()
sys.stdout.write("\b" * num_back)
sys.stdout.write(" " * num_back)
sys.stdout.write("\b" * num_back)
curr_lr = scheduler.get_lr()[0]
log_info = '[%d/%d (%.0f%%) lr=%.6f] loss: %.4f (%.4f)' % (
step, num_batches, 100. * step / num_batches, curr_lr,
epoch_loss / num_insts, epoch_loss / num_words)
sys.stdout.write(log_info)
sys.stdout.flush()
num_back = len(log_info)
logger.info('Epoch ' + str(epoch) + ' Loss: ' +
str(round(epoch_loss / num_insts, 4)))
if threshold >= 1.0:
acc, _ = evaluate(dev_dataset, batch_size, model, device)
else:
acc, _ = evaluate(dev_dataset, 1, model, device)
logger.info('\t Dev Acc: ' + str(round(acc * 100, 3)))
if analysis:
analyse(model,
dev_dataset,
batch_size,
device,
analyzer,
log_dir + '/dev_' + str(epoch),
buckted=False)
analyse(model,
test_dataset,
batch_size,
device,
analyzer,
log_dir + '/test_' + str(epoch),
buckted=False)
if best_epoch[1] < acc:
test_acc, _ = evaluate(test_dataset, batch_size, model, device)
logger.info('\t Test Acc: ' + str(round(test_acc * 100, 3)))
best_epoch = (epoch, acc, test_acc)
patient = 0
else:
patient += 1
epoch += 1
if patient > 4:
print('reset optimizer momentums')
scheduler.reset_state()
patient = 0
logger.info("Best Epoch: " + str(best_epoch[0]) + " Dev ACC: " +
str(round(best_epoch[1] * 100, 3)) + "Test ACC: " +
str(round(best_epoch[2] * 100, 3)))
return best_epoch
best_epoch = train(best_epoch, thread=10)
# logger.info("After tunning mu. Here we tunning variance")
# # flip gradient
#
# for parameter in model.parameters():
# # flip
# parameter.requires_grad = not parameter.requires_grad
# best_epoch = train(best_epoch)
with open(log_dir + '/' + 'result.json', 'w') as f:
final_result = {
"Epoch": best_epoch[0],
"Dev": best_epoch[1] * 100,
"Test": best_epoch[2] * 100
}
json.dump(final_result, f)
class ATC:
section = "AvailableOptions"
def __init__(self):
# Initialising fields
self.parameters = {}
# Loading config
self.config = self.loadConfig()
self.analyzer = Analyzer(self.config)
global analyzer_global
analyzer_global = self.analyzer
# Selecting mode
if len(sys.argv) > 1:
self.parse_args()
self.analyzer.error_occurred.connect(self.print_error)
filename = self.parameters["input"]
if not os.path.exists(filename):
self.print_error("File {} does not exist".format(filename))
sys.exit()
try:
text = self.analyzer.load_file(self.parameters["input"])
if not self.analyzer.isTextValid(text):
self.print_error(
"File {} does not contain valid text".format(filename))
sys.exit()
except Exception as e:
self.print_error("Error loading file {}:\n{}".format(
filename, e))
sys.exit()
result = self.analyzer.analyze(text, self.parameters)
if result is None:
self.print_error("Unknown error occurred")
sys.exit()
result.save_to_file(self.parameters["output"],
self.parameters["threshold"],
n_digits=5)
sys.exit(0)
else:
show_splashscreen()
self.ui = GUI(analyzer=self.analyzer, config=self.config)
def parse_args(self):
description = "Automated Text Classifier for VINITI. Чтобы запустить графический сеанс, " \
"запустите программу без аргументов"
argparser = ArgumentParser(prog="ATC", description=description)
formats = self.config.get(self.section, "formats").split(", ")
languages = self.config.get(self.section, "languages").split(", ")
norm_options = self.config.get(self.section,
"norm_predict").split(", ")
argparser.add_argument("-i",
"--input",
help="полный путь к файлу с текстом",
required=True)
# type=unescaped_str
argparser.add_argument(
"-o",
"--output",
help="полный путь к файлу, в который будет записан результат",
required=True)
argparser.add_argument("-id",
"--rubricator-id",
help="идентификатор рубрикатора",
required=True)
argparser.add_argument("-f",
"--format",
help="формат входного файла",
choices=formats,
required=False)
argparser.add_argument("-l",
"--language",
help="язык входного текста",
choices=languages,
required=True)
argparser.add_argument("-t",
"--threshold",
help="пороговое значение вероятности. " +
"Ответы классификатора с вероятностью ниже " +
"заданной выведены не будут",
default=0.0,
type=float,
required=False)
argparser.add_argument(
"-n",
"--normalize",
help="нормировать ли предсказание классификатора",
choices=norm_options,
required=False,
default="not")
subparsers = argparser.add_subparsers(help="Commands")
# Creating server command
server_parser = subparsers.add_parser("server",
help="запустить режим сервера")
server_parser.add_argument(
"port",
help="номер порта, на котором запустить сервер",
action=LaunchServer,
type=int)
self.parameters = vars(argparser.parse_args())
@staticmethod
def print_error(error_msg: str):
print(error_msg, file=sys.stderr)
@staticmethod
def loadConfig():
parser = ConfigParser()
parser.read([os.path.join(os.path.dirname(__file__), "config.ini")],
encoding="utf-8")
return parser
