def status(election_date):
"""
The route /<election_date>/status will return the status of a given
election date test, including the current position in the hopper, the
playback speed, and the path of the file that will be served at the current
position.
"""
if request.args.get('national', None):
LEVEL='local'
if request.args['national'].lower() == 'true':
LEVEL = 'national'
else:
return json.dumps({
'error': True,
'message': 'must specify national=true or national=false'
})
election_key = 'AP_DEJAVU_%s' % election_date
hopper = sorted(glob.glob('%s%s/%s/*' % (DATA_DIR, election_date, LEVEL)), key=lambda x:x.split('recording-')[1])
position = int(r_conn.get(election_key + '_POSITION') or 0)
playback = int(r_conn.get(election_key + '_PLAYBACK') or 1)
errormode = utils.to_bool(r_conn.get(election_key + '_ERRORMODE') or 'False')
ratelimited = utils.to_bool(r_conn.get(election_key + '_RATELIMITED') or 'False')
return json.dumps({
'playback': playback,
'position': position,
'errormode': errormode,
'ratelimited': ratelimited,
'file': hopper[position-1],
'level': LEVEL
})
def execute(h2o, params, config):
frame_id = config.get('frame_id')
df = h2o.get_frame(frame_id)
target_column = params.get("target_column")
analyzer = params.get("analyzer")
if len(analyzer) > 0:
url = params.get("url")
df_token = df[target_column].tokenize(
f'tokenize:elasticsearch:{url}?analyzer={analyzer}_analyzer')
else:
df_token = df[target_column].tokenize(params.get('regex'))
if to_bool(params.get('lower_case')):
df_token = df_token.tolower()
min_word_len = int(params.get('min_word_len'))
if min_word_len > 0:
df_token = df_token[(df_token.nchar() >= min_word_len) |
(df_token.isna()), :]
if to_bool(params.get('use_stop_words')):
df_token = df_token[(df_token.isna()) |
(~df_token.isin(STOP_WORDS)), :]
dest_frame_id = append_frame_id(frame_id, params.get('suffix'))
h2o.assign(df_token, dest_frame_id)
return {'frame_id': dest_frame_id}
def index():
"""
Will match directories named like the following:
2015-09-10
09-10-2015
"""
context = utils.build_context()
context['elections'] = []
elections = sorted([a.split('/')[-1] for a in glob.glob('%s/*' % DATA_DIR) if re.match('(\d{2,4}[-]\d{2,4}[-]\d{2,4})', a.split('/')[-1])], key=lambda x:x)
for e in elections:
for level in ['local', 'national']:
national = False
if level == 'national':
national = True
e_dict = {}
election_key = 'AP_DEJAVU_%s' % e
e_dict['election_date'] = e
e_dict['national'] = national
e_dict['level'] = level
e_dict['title'] = "%s [%s]" % (e, level)
e_dict['position'] = int(r_conn.get(election_key + '_POSITION') or 0)
e_dict['total_positions'] = len(glob.glob('%s%s/%s/*' % (DATA_DIR, e, level)))
e_dict['playback'] = int(r_conn.get(election_key + '_PLAYBACK') or 1)
e_dict['errormode'] = utils.to_bool(r_conn.get(election_key + '_ERRORMODE') or 'False')
e_dict['ratelimited'] = utils.to_bool(r_conn.get(election_key + '_RATELIMITED') or 'False')
context['elections'].append(e_dict)
return render_template('index.html', **context)
def on_participate(self, request):
#获取分组列表
alts = request.args.getlist('alternatives')
# alt_fractions = request.args.getlist('alt_fractions')
experiment_name = request.args.get('experiment') #实验名称
force = request.args.get('force')
#获取record_force参数，并进行to_bool判断若为[y,ture,yes],返回Ture,否则False
record_force = to_bool(request.args.get('record_force', 'false'))
client_id = request.args.get('client_id') #获取client_id
traffic_fraction = request.args.get('traffic_fraction') #获取参与比例
if traffic_fraction is not None:
traffic_fraction = float(
traffic_fraction) #若traffic_fraction存在则转为float()
prefetch = to_bool(request.args.get(
'prefetch',
'false')) #获取prefetch，并进行to_bool判断若为[y,ture,yes],返回Ture,否则Fals
if client_id is None or experiment_name is None or alts is None: #三个必要参数，否则返回400
return json_error({'message': 'missing arguments'}, request, 400)
dt = None
if request.args.get(
"datetime"): #获取datetime参数，转化为datetime.datetime()类型，没有则默认None
dt = dateutil.parser.parse(request.args.get("datetime"))
try:
if should_exclude_visitor(
request): #判断user_agent是爬虫或者IP为指定排除的ip,则拒绝参与
exp = Experiment.find(experiment_name, redis=self.redis)
if exp.winner is not None: #判断是否已经胜出
alt = exp.winner
else:
alt = exp.control #没有胜出，返回分组中的第一个赋值给alt
else:
alt = participate(
experiment_name,
alts,
client_id,
force=force,
record_force=record_force,
traffic_fraction=traffic_fraction,
# alt_fraction=alt_fractions,
prefetch=prefetch,
datetime=dt,
redis=self.redis)
except ValueError as e:
return json_error({'message': str(e)}, request, 400)
resp = {
'alternative': {
'name': alt.name
},
'experiment': {
'name': alt.experiment.name,
},
'client_id': client_id,
'status': 'ok'
}
return json_success(resp, request)
def execute(h2o, params, config):
frame_id = config.get('frame_id')
model_id = config.get('model_id')
df = h2o.get_frame(frame_id)
column_header = params.get('column_header')
if len(column_header) > 0:
df_head = df[:int(column_header)]
df = df[int(column_header):]
pred_model = h2o.get_model(model_id)
df_pred = pred_model.predict(df)
df_pred.columns = [x[len('reconstr_'):] for x in df_pred.columns]
dest_frame_id = append_frame_id(frame_id, params.get('suffix'))
if to_bool(params.get('topn_output')):
df_topn = get_topN(df_pred, int(params.get('topn_percent')))
if df_head is not None:
df_topn = df_head.cbind(df_topn)
h2o.assign(df_topn, dest_frame_id)
h2o.remove(str(df_pred.frame_id))
else:
h2o.assign(df_pred, dest_frame_id)
return {'frame_id': dest_frame_id}
def execute(h2o, params, config):
frame_id = config.get('frame_id')
df = h2o.get_frame(frame_id)
from h2o.estimators import H2OWord2vecEstimator
w2v_model = H2OWord2vecEstimator(
epochs=int(params.get('epochs')),
init_learning_rate=float(params.get('init_learning_rate')),
max_runtime_secs=float(params.get('max_runtime_secs')),
min_word_freq=int(params.get('min_word_freq')),
sent_sample_rate=float(params.get('sent_sample_rate')),
vec_size=int(params.get('vec_size')),
window_size=int(params.get('window_size')))
w2v_model.train(training_frame=df)
save_model(params, w2v_model.model_id)
is_transform = params.get("is_transform")
if is_transform is not None and to_bool(is_transform):
df_vecs = w2v_model.transform(
df, aggregate_method=params.get('aggregate_method'))
dest_frame_id = append_frame_id(frame_id,
params.get('transform_suffix'))
h2o.assign(df_vecs, dest_frame_id)
else:
dest_frame_id = frame_id
return {'frame_id': dest_frame_id, 'model_id': w2v_model.model_id}
def on_participate(self, request):
alts = request.args.getlist('alternatives')
experiment_name = request.args.get('experiment')
force = request.args.get('force')
record_force = to_bool(request.args.get('record_force', 'false'))
client_id = request.args.get('client_id')
traffic_fraction = request.args.get('traffic_fraction')
if traffic_fraction is not None:
traffic_fraction = float(traffic_fraction)
prefetch = to_bool(request.args.get('prefetch', 'false'))
if client_id is None or experiment_name is None or alts is None:
return json_error({'message': 'missing arguments'}, request, 400)
dt = None
if request.args.get("datetime"):
dt = dateutil.parser.parse(request.args.get("datetime"))
try:
if should_exclude_visitor(request):
exp = Experiment.find(experiment_name, redis=self.redis)
if exp.winner is not None:
alt = exp.winner
else:
alt = exp.control
else:
alt = participate(experiment_name, alts, client_id,
force=force, record_force=record_force,
traffic_fraction=traffic_fraction,
prefetch=prefetch, datetime=dt, redis=self.redis)
except ValueError as e:
return json_error({'message': str(e)}, request, 400)
resp = {
'alternative': {
'name': alt.name
},
'experiment': {
'name': alt.experiment.name,
},
'client_id': client_id,
'status': 'ok'
}
return json_success(resp, request)
def on_participate(self, request):
alts = request.args.getlist('alternatives')
experiment_name = request.args.get('experiment')
force = request.args.get('force')
record_force = to_bool(request.args.get('record_force', 'false'))
client_id = request.args.get('client_id')
traffic_fraction = request.args.get('traffic_fraction')
if traffic_fraction is not None:
traffic_fraction = float(traffic_fraction)
prefetch = to_bool(request.args.get('prefetch', 'false'))
if client_id is None or experiment_name is None or alts is None:
return json_error({'message': 'missing arguments'}, request, 400)
dt = None
if request.args.get("datetime"):
dt = dateutil.parser.parse(request.args.get("datetime"))
try:
if should_exclude_visitor(request):
exp = Experiment.find(experiment_name, redis=self.redis)
if exp.winner is not None:
alt = exp.winner
else:
alt = exp.control
else:
alt = participate(experiment_name, alts, client_id,
force=force, record_force=record_force,
traffic_fraction=traffic_fraction,
prefetch=prefetch, datetime=dt, redis=self.redis)
except ValueError as e:
return json_error({'message': str(e)}, request, 400)
resp = {
'alternative': {
'name': alt.name
},
'experiment': {
'name': alt.experiment.name,
},
'client_id': client_id,
'status': 'ok'
}
return json_success(resp, request)
def index():
"""
Will match directories named like the following:
2015-09-10
09-10-2015
"""
context = utils.build_context()
context['elections'] = []
elections = sorted([a.split('/')[-1] for a in glob.glob('%s/*' % DATA_DIR) if re.match('(\d{2,4}[-]\d{2,4}[-]\d{2,4})', a.split('/')[-1])], key=lambda x:x)
for e in elections:
e_dict = {}
election_key = 'AP_DEJAVU_%s' % e
e_dict['election_date'] = e
e_dict['position'] = int(os.environ.get(election_key + '_POSITION', '0'))
e_dict['total_positions'] = len(glob.glob('%s%s/*' % (DATA_DIR, e)))
e_dict['playback'] = int(os.environ.get(election_key + '_PLAYBACK', '1'))
e_dict['errormode'] = utils.to_bool(os.environ.get(election_key + '_ERRORMODE', 'False'))
e_dict['ratelimited'] = utils.to_bool(os.environ.get(election_key + '_RATELIMITED', 'False'))
context['elections'].append(e_dict)
return render_template('index.html', **context)
def status(year, election_date):
"""
The route /<election_date>/status will return the status of a given
election date test, including the current position in the hopper, the
playback speed, and the path of the file that will be served at the current
position.
"""
if request.args.get('national', None):
LEVEL = 'local'
if request.args['national'].lower() == 'true':
LEVEL = 'national'
else:
return json.dumps({
'error':
True,
'message':
'must specify national=true or national=false'
})
election_key = 'AP_DEJAVU_%s' % election_date
hopper = sorted(glob.glob('%s%s/%s/*' % (DATA_DIR, election_date, LEVEL)),
key=lambda x: x)
position = int(r_conn.get(election_key + '_POSITION') or 0)
playback = int(r_conn.get(election_key + '_PLAYBACK') or 1)
errormode = utils.to_bool(
r_conn.get(election_key + '_ERRORMODE') or 'False')
ratelimited = utils.to_bool(
r_conn.get(election_key + '_RATELIMITED') or 'False')
return json.dumps({
'playback': playback,
'position': position,
'errormode': errormode,
'ratelimited': ratelimited,
'file': hopper[position - 1],
'level': LEVEL
})
def execute(h2o, params, config):
frame_id = config.get('frame_id')
df = h2o.get_frame(frame_id)
column = params.get('column')
ascending = to_bool(params.get('ascending'))
df_sort = df.sort(by=[column], ascending=[ascending])
dest_frame_id = append_frame_id(frame_id, params.get('suffix'))
h2o.assign(df_sort, dest_frame_id)
return {'frame_id': dest_frame_id}
def add_template(self):
"""Add a new template in JSON format to `pandoc_templates.json`.
"""
if DELIMITER in self.arg:
key, value = self._parse_query(self.arg)
if key == 'Name':
self._store_template_info('name', value)
elif key == 'Defaults':
bool_v = utils.to_bool(value)
self._store_template_info('use_defaults', bool_v)
elif key == 'Command':
clean_cmd = self._parse_template(value)
self._store_template_info('options', clean_cmd)
return 'Template successfully created!'
def add_template(self):
"""Add a new template in JSON format to `pandoc_templates.json`.
"""
if DELIMITER in self.arg:
key, value = self._parse_query(self.arg)
if key == 'Name':
self._store_template_info('name', value)
elif key == 'Defaults':
bool_v = utils.to_bool(value)
self._store_template_info('use_defaults', bool_v)
elif key == 'Command':
clean_cmd = self._parse_template(value)
self._store_template_info('options', clean_cmd)
return 'Template successfully created!'
def status(election_date):
"""
The route /<election_date>/status will return the status of a given
election date test, including the current position in the hopper, the
playback speed, and the path of the file that will be served at the current
position.
"""
election_key = 'AP_DEJAVU_%s' % election_date
hopper = sorted(glob.glob('%s%s/*' % (DATA_DIR, election_date)), key=lambda x:x.split('recording-')[1])
position = int(os.environ.get(election_key + '_POSITION', '0'))
playback = int(os.environ.get(election_key + '_PLAYBACK', '1'))
errormode = utils.to_bool(os.environ.get(election_key + '_ERRORMODE', 'False'))
ratelimited = utils.to_bool(os.environ.get(election_key + '_RATELIMITED', 'False'))
return json.dumps({
'playback': playback,
'position': position,
'errormode': errormode,
'ratelimited': ratelimited,
'file': hopper[position-1]
})
def index(year):
"""
Will match directories named like the following:
2015-09-10
09-10-2015
"""
context = utils.build_context()
context['elections'] = []
context['year'] = year
elections = sorted([
a.split('/')[-1] for a in glob.glob('%s/*' % DATA_DIR)
if re.match('(\d{2,4}[-]\d{2,4}[-]\d{2,4})',
a.split('/')[-1])
],
key=lambda x: x)
for e in elections:
for level in ['national']:
national = True
e_dict = {}
election_key = 'AP_DEJAVU_%s' % e
e_dict['election_date'] = e
e_dict['national'] = national
e_dict['level'] = level
e_dict['title'] = "%s [%s]" % (e, level)
e_dict['position'] = int(
r_conn.get(election_key + '_POSITION') or 0)
e_dict['total_positions'] = len(
glob.glob('%s%s/%s/*' % (DATA_DIR, e, level)))
e_dict['playback'] = int(
r_conn.get(election_key + '_PLAYBACK') or 1)
e_dict['errormode'] = utils.to_bool(
r_conn.get(election_key + '_ERRORMODE') or 'False')
e_dict['ratelimited'] = utils.to_bool(
r_conn.get(election_key + '_RATELIMITED') or 'False')
context['elections'].append(e_dict)
return render_template('index.html', **context)
def execute(h2o, params, config):
frame_id = config.get('frame_id')
df = h2o.get_frame(frame_id)
column_header = params.get('column_header')
if len(column_header) > 0:
df = df[int(column_header):]
from h2o.estimators.glrm import H2OGeneralizedLowRankEstimator
glrm_model = H2OGeneralizedLowRankEstimator(
expand_user_y=to_bool(params.get('expand_user_y')),
gamma_x=float(params.get('gamma_x')),
gamma_y=float(params.get('gamma_y')),
ignore_const_cols=to_bool(params.get('ignore_const_cols')),
impute_original=to_bool(params.get('impute_original')),
init=str(params.get('init')),
init_step_size=float(params.get('init_step_size')),
k=int(params.get('k')),
loss=str(params.get('loss')),
max_iterations=int(params.get('max_iterations')),
max_runtime_secs=float(params.get('max_runtime_secs')),
max_updates=int(params.get('max_updates')),
min_step_size=float(params.get('min_step_size')),
multi_loss=str(params.get('multi_loss')),
period=int(params.get('period')),
recover_svd=to_bool(params.get('recover_svd')),
regularization_x=str(params.get('regularization_x')),
regularization_y=str(params.get('regularization_y')),
score_each_iteration=to_bool(params.get('score_each_iteration')),
seed=int(params.get('seed')),
svd_method=str(params.get('svd_method')))
glrm_model.train(training_frame=df)
glrm_model.show()
save_model(params, glrm_model.model_id)
return {'frame_id': frame_id, 'model_id': glrm_model.model_id}
def update_mlt_params(user_mlt_params, params):
for k, v in user_mlt_params.items():
if k not in USER_MLT_PARAMS:
continue
if not v.strip():
v = None # use ES defaults
elif k == 'fields' or k == 'stop_words':
v = map(string.strip, v.split(','))
elif k == 'analyzer':
pass
elif k in ('boost_terms'):
v = float(v)
elif k == 'include':
v = utils.to_bool(v)
elif k == 'minimum_should_match':
v = str(v)
else:
v = int(v)
params[k] = v
def execute(h2o, params, config):
frame_id = config.get('frame_id')
df = h2o.get_frame(frame_id)
columns = params.get('columns')
if columns is not None and len(columns) > 2:
columns = json.loads(columns)
df = df[columns]
use_value = params.get('use')
if use_value is not None and len(use_value) == 0:
use_value = None
df_cor = df.cor(na_rm=to_bool(params.get('na_rm')),
use=use_value,
method=params.get('method'))
dest_frame_id = append_frame_id(frame_id, params.get('suffix'))
h2o.assign(df_cor, dest_frame_id)
return {'frame_id': dest_frame_id}
def notify_failure(container: Container):
if not to_bool(container.labels.get(FAILURE_NOTIFY)):
log(f"<8133c8b6> ({container.name}) Skip send container failure message because "
+ f"of 'failure_notify'={container.labels.get(FAILURE_NOTIFY)}")
return
time = datetime.now(timezone.utc)
failure_time = container.attrs["State"]["Health"]["Log"][-1]["End"]
healthcheck_response = container.attrs["State"]["Health"]["Log"][-1][
"Output"]
if type(healthcheck_response) is str:
healthcheck_response = healthcheck_response.strip()
latest_send = None
container_sends = list(
sorted(filter(lambda x: x.container_name == container.name,
sended_emails),
key=lambda x: x.send_time))
if len(container_sends) > 0:
latest_send = container_sends[-1]
send_timeout = config.default_send_timeout_min
if FAILURE_NOTIFY_TIMEOUT in container.labels.keys():
send_timeout = int(container.labels[FAILURE_NOTIFY_TIMEOUT])
if not latest_send or (time -
latest_send.send_time).seconds / 60 > send_timeout:
default_address = config.default_receiver_address
addresses = [default_address]
if FAILURE_NOTIFY_EMAIL in container.labels.keys():
addresses = container.labels[FAILURE_NOTIFY_EMAIL].split(',')
log(f"<ad25da5b> ({container.name}) Send container failure message to: {addresses}"
)
for address in addresses:
send_email(
Email(address=address,
container_name=container.name,
failure_time=failure_time,
healthcheck_response=healthcheck_response))
else:
log(f"<8133c8b6> ({container.name}) Skip send container failure message because "
+ f"of send timeout: {send_timeout} minutes")
def create(self, request):
"""
POST
Insertar los tribunales a partir de una comision
:param request:
:return :
{status: True/False, data:{datos de la comision insertada o de todas las comisiones}
"""
try:
params = utils.get_params(request)
self.logger.info('INICIO WS - COMISIONEVALUACIONVIEW POST del usuario: %s con parametros: %s' %
(request.user.email if hasattr(request.user, 'email') else request.user.username, params))
if request.user.has_perm('comisiones_evaluacion.comision.create') or request.user.is_admin:
comisiones = utils.to_bool(params.get('comisiones'))
comision = Comision(request.user, params.get('convocatoria'), params.get('anio'),
params.get('titulacion'), comisiones=comisiones)
if not comisiones:
comision.tutores_comisiones()
resul = comision.asig_tfgs()
while comision.reintentar:
comision = Comision(request.user, params.get('convocatoria'), comisiones=comisiones)
comision.asig_tfgs()
if resul['status']:
resul_status = status.HTTP_200_OK
else:
resul = dict(message=resul['message'])
resul_status = status.HTTP_400_BAD_REQUEST
else:
resul = dict(message="Sin privilegios")
resul_status = status.HTTP_405_METHOD_NOT_ALLOWED
self.logger.info('FIN WS - COMISIONEVALUACIONVIEW POST del usuario: %s con resultado: %s' %
(request.user.email if hasattr(request.user, 'email') else request.user.username, resul))
return Response(resul, status=resul_status)
except Exception as e:
resul = dict(status=False, message="Error en la llamada")
self.logger.critical('COMISIONEVALUACIONVIEW POST: %s %s' % (resul, e))
return Response(resul, status=status.HTTP_400_BAD_REQUEST)
def execute(h2o, params, config):
frame_id = config.get('frame_id')
df = h2o.get_frame(frame_id)
input_columns = params.get("input_columns")
if input_columns is None or len(input_columns) == 0:
input_columns = df.col_names
else:
import json
input_columns = json.loads(input_columns)
from h2o.estimators import H2OKMeansEstimator
kmeans_model = H2OKMeansEstimator(
categorical_encoding=params.get("categorical_encoding"),
estimate_k=to_bool(params.get("estimate_k")),
fold_assignment=params.get("fold_assignment"),
ignore_const_cols=to_bool(params.get("ignore_const_cols")),
init=params.get("init"),
k=int(params.get("k")),
keep_cross_validation_fold_assignment=to_bool(
params.get("keep_cross_validation_fold_assignment")),
keep_cross_validation_models=to_bool(
params.get("keep_cross_validation_models")),
keep_cross_validation_predictions=to_bool(
params.get("keep_cross_validation_predictions")),
max_iterations=int(params.get("max_iterations")),
max_runtime_secs=float(params.get("max_runtime_secs")),
nfolds=int(params.get("nfolds")),
score_each_iteration=to_bool(params.get("score_each_iteration")),
seed=int(params.get("seed")),
standardize=to_bool(params.get("standardize")))
kmeans_model.train(x=input_columns, training_frame=df)
kmeans_model.show()
save_model(params, kmeans_model.model_id)
return {'frame_id': frame_id, 'model_id': kmeans_model.model_id}
def get_body_data(self, body):
self.correlation_id = body.get('CorrelationId')
self.success = to_bool(body.get('Success'))
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError(
'SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "localhost"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections': int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '15')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses': os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path': os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
def __init__(self, arguments, model, train_fn, valid_fn,
train_sets, valid_sets):
self.logger = logging.getLogger(__name__)
self.model = model
self.train_fn = train_fn
self.valid_fn = valid_fn
self.train_sets = train_sets
self.valid_sets = valid_sets
#################### parse configs ####################
self.resume = False
if "resume" in arguments:
self.resume = utils.to_bool(arguments["resume"])
self.wdir = arguments["wdir"]
self.output_file = arguments["output_file"]
self.learn_rate = 0.1
if "learn_rate" in arguments:
self.learn_rate = float(arguments["learn_rate"])
self.halving_factor = 0.5
if "halving_factor" in arguments:
self.halving_factor = float(arguments["halving_factor"])
self.max_iters = 20
if "max_iters" in arguments:
self.max_iters = int(arguments["max_iters"])
self.min_iters = 0
if "min_iters" in arguments:
self.min_iters = int(arguments["min_iters"])
self.keep_lr_iters = 15
if "keep_lr_iters" in arguments:
self.keep_lr_iters = int(arguments["keep_lr_iters"])
self.start_halving_impr=0.01
if "start_halving_impr" in arguments:
self.start_halving_impr = float(arguments["start_halving_impr"])
self.end_halving_impr=0.001
if "end_halving_impr" in arguments:
self.end_halving_impr = float(arguments["end_halving_impr"])
self.continue_with_rate = False
if "continue_with_rate" in arguments:
self.continue_with_rate = utils.to_bool(arguments["continue_with_rate"])
self.halving_criteria = "loss"
if "halving_criteria" in arguments:
self.halving_criteria = arguments["halving_criteria"]
criteria_list = ["loss", "frame_err"]
if self.halving_criteria not in criteria_list:
raise Exception("invalid halving criteria. must be one of " + str(criteria_list))
# batch_size and momentum
self.batch_size=256
if arguments.has_key('batch_size'):
self.batch_size = int(arguments['batch_size'])
self.momentum=0.5
self.momentum_start = 1
if arguments.has_key('momentum'):
self.momentum = float(arguments['momentum'])
if 'momentum_start' in arguments:
self.momentum_start = int(arguments['momentum_start'])
# other stuff
if self.resume:
if not os.path.exists(self.wdir):
raise Exception("wdir must exist if resume=True")
else:
if not os.path.exists(self.wdir):
os.makedirs(self.wdir)
else:
self.logger.info("Directory already exists...")
out = StringIO.StringIO()
print >>out, "\n********** Trainer **********"
print >>out, "resume", self.resume
print >>out, "wdir", self.wdir
print >>out, "output_file", self.output_file
print >>out, "learn_rate", self.learn_rate
print >>out, "halving_factor", self.halving_factor
print >>out, "max_iters", self.max_iters
print >>out, "min_iters", self.min_iters
print >>out, "keep_lr_iters", self.keep_lr_iters
print >>out, "start_halving_impr", self.start_halving_impr
print >>out, "end_halving_impr", self.end_halving_impr
print >>out, "continue_with_rate", self.continue_with_rate
print >>out, "halving_criteria", self.halving_criteria
print >>out, "batch_size", self.batch_size
print >>out, "momentum", self.momentum
print >>out, "momentum_start", self.momentum_start
self.logger.info(out.getvalue())
self.mlp_init = self.wdir + "/mlp_init"
if not self.resume: # brand new
save(self.model, self.mlp_init)
# runtime state
self.iter = 0
self.done = False
self.loss = sys.float_info.max
self.rate = self.learn_rate
self.mlp_best = self.mlp_init
self.halving = False
self.wasAccepted = True
if self.resume:
if os.path.isfile(self.wdir+"/trainer_state"):
self._load_state()
def replay(election_date):
"""
The route `/<election_date>` will replay the election files found in the folder
`/<DATA_DIR>/<election_date>/`. The files should be named such that the first file
will be sorted first in a list by `glob.glob()`, e.g., a higher letter (a) or lower
number (0). Incrementing UNIX timestamps (such as those captured by Elex) would be
ideal.
This route takes two optional control parameters. Once these have been passed to set
up an election test, the raw URL will obey the instructions below until the last
file in the hopper has been reached.
* `position` will return the file at this position in the hopper. So, for example,
`position=0` would set the pointer to the the first file in the hopper and return it.
* `playback` will increment the position by itself until it is reset. So, for example,
`playback=5` would skip to every fifth file in the hopper.
When the last file in the hopper has been reached, it will be returned until the Flask
app is restarted OR a new pair of control parameters are passed.
Example: Let's say you would like to test an election at 10x speed. You have 109
files in your `/<DATA_DIR>/<election_date>/` folder named 001.json through 109.json
* Request 1: `/<election_date>?position=0&playback=10` > 001.json
* Request 2: `/<election_date>` > 011.json
* Request 3: `/<election_date>` > 021.json
* Request 4: `/<election_date>` > 031.json
* Request 5: `/<election_date>` > 041.json
* Request 6: `/<election_date>` > 051.json
* Request 7: `/<election_date>` > 061.json
* Request 8: `/<election_date>` > 071.json
* Request 9: `/<election_date>` > 081.json
* Request 10: `/<election_date>` > 091.json
* Request 11: `/<election_date>` > 101.json
* Request 12: `/<election_date>` > 109.json
* Request 13 - ???: `/<election_date>` > 109.json
Requesting /<election_date>?position=0&playback=1 will reset to the default position
and playback speeds, respectively.
"""
election_key = 'AP_DEJAVU_%s' % election_date
hopper = sorted(glob.glob('%s%s/*' % (DATA_DIR, election_date)), key=lambda x:x.split('recording-')[1])
position = int(os.environ.get(election_key + '_POSITION', '0'))
playback = int(os.environ.get(election_key + '_PLAYBACK', '1'))
errormode = utils.to_bool(os.environ.get(election_key + '_ERRORMODE', 'False'))
ratelimited = utils.to_bool(os.environ.get(election_key + '_RATELIMITED', 'False'))
if request.args.get('errormode', None):
if request.args.get('errormode', None) == 'true':
os.environ[election_key + '_ERRORMODE'] = 'True'
errormode = True
if request.args.get('errormode', None) == 'false':
os.environ[election_key + '_ERRORMODE'] = 'False'
errormode = False
if request.args.get('ratelimited', None):
if request.args.get('ratelimited', None) == 'true':
os.environ[election_key + '_RATELIMITED'] = 'True'
ratelimited = True
if request.args.get('ratelimited', None) == 'false':
os.environ[election_key + '_RATELIMITED'] = 'False'
ratelimited = False
if request.args.get('playback', None):
try:
playback = abs(int(request.args.get('playback', None)))
except ValueError:
return json.dumps({
'error': True,
'error_type': 'ValueError',
'message': 'playback must be an integer greater than 0.'
})
if request.args.get('position', None):
try:
position = abs(int(request.args.get('position', None)))
except ValueError:
return json.dumps({
'error': True,
'error_type': 'ValueError',
'message': 'position must be an integer greater than 0.'
})
os.environ[election_key + '_PLAYBACK'] = str(playback)
if request.args.get('ratelimited', None) or request.args.get('errormode', None):
return json.dumps({"success": True})
else:
if ratelimited:
return make_response((RATELIMITED_STRING, 403, RATELIMITED_HEADERS))
if errormode:
if random.randrange(1,3) % 2 == 0:
return make_response(json.dumps({"status": 500, "error": True}), 500, ERRORMODE_HEADERS)
if position + playback < (len(hopper) - 1):
"""
Needs the if statement here to set the position truly to zero if it's specified
in the url params.
"""
if request.args.get('position', None) or request.args.get('playback', None) or request.args.get('ratelimited', None) or request.args.get('errormode', None):
os.environ[election_key + '_POSITION'] = str(position)
else:
os.environ[election_key + '_POSITION'] = str(position + playback)
else:
os.environ[election_key + '_POSITION'] = str(len(hopper))
with open(hopper[position - 1], 'r') as readfile:
payload = str(readfile.read())
return payload
from dotenv import load_dotenv
from utils import get_required_env, to_bool, log
dotenv_path = os.path.join(os.path.dirname(__file__), '.env')
if os.path.exists(dotenv_path):
load_dotenv(dotenv_path)
default_send_timeout_min: int = int(os.getenv("DEFAULT_SEND_TIMEOUT_MIN", 15))
default_receiver_address: str = get_required_env("DEFAULT_RECEIVER_ADDRESS")
email_from: str = get_required_env("EMAIL_FROM")
email_host: str = get_required_env("EMAIL_HOST")
email_port: int = int(os.getenv("EMAIL_PORT", 465))
email_login: Optional[str] = os.getenv("EMAIL_LOGIN")
email_password: Optional[str] = os.getenv("EMAIL_PASSWORD")
email_enable_tls: bool = to_bool(os.getenv("EMAIL_ENABLE_TLS", False))
email_use_ssl: bool = to_bool(os.getenv("EMAIL_USE_SSL", False))
container_label: str = os.getenv("AUTOHEAL_CONTAINER_LABEL", "autoheal")
container_stop_timeout: int = int(
os.getenv("AUTOHEAL_DEFAULT_STOP_TIMEOUT", 10))
container_interval: int = int(os.getenv("AUTOHEAL_INTERVAL", 5))
container_start_period: int = int(os.getenv("AUTOHEAL_START_PERIOD", 0))
container_debounce_time: int = int(os.getenv("AUTOHEAL_DEBOUNCE_TIME", 0))
clean_period: int = int(os.getenv("CLEAN_PERIOD", 24 * 60))
docker_base_url: str = os.getenv("DOCKER_BASE_URL",
"unix://var/run/docker.sock")
docker_timeout: int = int(os.getenv("DOCKER_CONNECTION_TIMEOUT", 60))
log(f"""Configuration:
DEFAULT_SEND_TIMEOUT_MIN={default_send_timeout_min}
DEFAULT_RECEIVER_ADDRESS={default_receiver_address}
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError(
'SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "127.0.0.1"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections': int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '15')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses': os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path': os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
def __init__(self,
mod,
title,
version,
doc,
usage_data=None,
avail_modes=None,
supported_ui_toolkits=None,
run_as_root_ok=False,
quiet=False):
self.mod = mod
self.title = title
self.version = version
self.doc = doc
self.usage_data = usage_data
os.umask(0037)
log.set_module(mod)
self.args = []
self.quiet = quiet
self.lock_file = None
prop.prog = sys.argv[0]
if os.getenv("HPLIP_DEBUG"):
log.set_level('debug')
self.avail_modes = avail_modes
if supported_ui_toolkits is not None:
self.supported_ui_toolkits = supported_ui_toolkits
self.num_supported_ui_toolkits = len(self.supported_ui_toolkits)
else:
self.supported_ui_toolkits = []
self.num_supported_ui_toolkits = 0
self.default_ui_toolkit = sys_conf.get('configure', 'ui-toolkit',
'qt4')
self.num_installed_ui_toolkits = 0
self.installed_ui_toolkits = []
if utils.to_bool(sys_conf.get('configure', 'qt3', '0')):
self.installed_ui_toolkits.append(UI_TOOLKIT_QT3)
self.num_installed_ui_toolkits += 1
if utils.to_bool(sys_conf.get('configure', 'qt4', '0')):
self.installed_ui_toolkits.append(UI_TOOLKIT_QT4)
self.num_installed_ui_toolkits += 1
self.default_mode = INTERACTIVE_MODE
self.num_valid_modes = 0
if self.avail_modes is not None:
if GUI_MODE in self.avail_modes and prop.gui_build and self.installed_ui_toolkits:
self.num_valid_modes += 1
if INTERACTIVE_MODE in self.avail_modes:
self.num_valid_modes += 1
if NON_INTERACTIVE_MODE in self.avail_modes:
self.num_valid_modes += 1
if self.avail_modes is not None:
if INTERACTIVE_MODE in self.avail_modes:
self.default_mode = INTERACTIVE_MODE
elif NON_INTERACTIVE_MODE in self.avail_modes:
self.default_mode = NON_INTERACTIVE_MODE
if self.supported_ui_toolkits is not None and prop.gui_build and self.installed_ui_toolkits:
if self.default_ui_toolkit == 'qt3' and UI_TOOLKIT_QT4 in self.supported_ui_toolkits and \
UI_TOOLKIT_QT3 not in self.supported_ui_toolkits and INTERACTIVE_MODE in self.avail_modes:
# interactive + qt4 and default is qt3 --> set to interactive (if avail) (e.g., hp-align)
self.default_mode = INTERACTIVE_MODE
self.default_ui_toolkit = 'none'
elif (UI_TOOLKIT_QT4 in self.supported_ui_toolkits and self.default_ui_toolkit == 'qt4' and UI_TOOLKIT_QT4 in self.installed_ui_toolkits) or \
(UI_TOOLKIT_QT3 in self.supported_ui_toolkits and self.default_ui_toolkit == 'qt3' and UI_TOOLKIT_QT3 in self.installed_ui_toolkits):
self.default_mode = GUI_MODE
elif self.default_ui_toolkit == 'qt3' and UI_TOOLKIT_QT3 not in self.supported_ui_toolkits:
if UI_TOOLKIT_QT4 in self.supported_ui_toolkits and UI_TOOLKIT_QT4 in self.installed_ui_toolkits: # (e.g, hp-linefeedcal?)
self.default_ui_toolkit = 'qt4'
self.default_mode = GUI_MODE
elif INTERACTIVE_MODE in self.avail_modes:
self.default_mode = INTERACTIVE_MODE
elif NON_INTERACTIVE_MODE in self.avail_modes:
self.default_mode = NON_INTERACTIVE_MODE
else:
log.error("%s cannot be run using Qt3 toolkit." % self.mod)
sys.exit(1)
elif self.default_ui_toolkit == 'qt4' and UI_TOOLKIT_QT4 not in self.supported_ui_toolkits:
if UI_TOOLKIT_QT3 in self.supported_ui_toolkits and UI_TOOLKIT_QT3 in self.installed_ui_toolkits: # (e.g., hp-unload)
self.default_ui_toolkit = 'qt3'
self.default_mode = GUI_MODE
elif INTERACTIVE_MODE in self.avail_modes:
self.default_mode = INTERACTIVE_MODE
elif NON_INTERACTIVE_MODE in self.avail_modes:
self.default_mode = NON_INTERACTIVE_MODE
else:
log.error("%s cannot be run using Qt4 toolkit." % self.mod)
sys.exit(1)
self.mode = self.default_mode
#log.debug("Default ui-toolkit: %s" % self.default_ui_toolkit)
#log.debug("Default mode: %s" % self.default_mode)
if os.getuid() == 0 and not run_as_root_ok:
log.warn("%s should not be run as root/superuser." % mod)
def replay(year, election_date):
"""
The route `/<election_date>` will replay the election files found in the folder
`/<DATA_DIR>/<election_date>/`. The files should be named such that the first file
will be sorted first in a list by `glob.glob()`, e.g., a higher letter (a) or lower
number (0). Incrementing UNIX timestamps (such as those captured by Elex) would be
ideal.
This route takes two optional control parameters. Once these have been passed to set
up an election test, the raw URL will obey the instructions below until the last
file in the hopper has been reached.
* `position` will return the file at this position in the hopper. So, for example,
`position=0` would set the pointer to the the first file in the hopper and return it.
* `playback` will increment the position by itself until it is reset. So, for example,
`playback=5` would skip to every fifth file in the hopper.
When the last file in the hopper has been reached, it will be returned until the Flask
app is restarted OR a new pair of control parameters are passed.
Example: Let's say you would like to test an election at 10x speed. You have 109
files in your `/<DATA_DIR>/<election_date>/` folder named 001.json through 109.json
* Request 1: `/<election_date>?position=0&playback=10` > 001.json
* Request 2: `/<election_date>` > 011.json
* Request 3: `/<election_date>` > 021.json
* Request 4: `/<election_date>` > 031.json
* Request 5: `/<election_date>` > 041.json
* Request 6: `/<election_date>` > 051.json
* Request 7: `/<election_date>` > 061.json
* Request 8: `/<election_date>` > 071.json
* Request 9: `/<election_date>` > 081.json
* Request 10: `/<election_date>` > 091.json
* Request 11: `/<election_date>` > 101.json
* Request 12: `/<election_date>` > 109.json
* Request 13 - ???: `/<election_date>` > 109.json
Requesting /<election_date>?position=0&playback=1 will reset to the default position
and playback speeds, respectively.
"""
LEVEL = 'national'
if request.args.get('national', None):
if request.args['national'].lower() == 'false':
LEVEL = 'local'
# if request.args.get('national', None) and request.args.get('level', None):
# LEVEL = 'local'
# if request.args['national'].lower() == 'true':
# LEVEL = 'national'
# if request.args['level'] == 'district':
# LEVEL = 'districts'
# else:
# return json.dumps({
# 'error': True,
# 'message': 'must specify national=true or national=false and level=ru or level=district'
# })
election_key = 'AP_DEJAVU_%s' % election_date
hopper = sorted(glob.glob('%s%s/%s/*' % (DATA_DIR, election_date, LEVEL)),
key=lambda x: x)
position = int(r_conn.get(election_key + '_POSITION') or 0)
playback = int(r_conn.get(election_key + '_PLAYBACK') or 1)
errormode = utils.to_bool(
r_conn.get(election_key + '_ERRORMODE') or 'False')
ratelimited = utils.to_bool(
r_conn.get(election_key + '_RATELIMITED') or 'False')
if request.args.get('errormode', None):
if request.args.get('errormode', None) == 'true':
r_conn.set(election_key + '_ERRORMODE', 'True')
errormode = True
if request.args.get('errormode', None) == 'false':
r_conn.set(election_key + '_ERRORMODE', 'False')
errormode = False
if request.args.get('ratelimited', None):
if request.args.get('ratelimited', None) == 'true':
r_conn.set(election_key + '_RATELIMITED', 'True')
ratelimited = True
if request.args.get('ratelimited', None) == 'false':
r_conn.set(election_key + '_RATELIMITED', 'False')
ratelimited = False
if request.args.get('playback', None):
try:
playback = abs(int(request.args.get('playback', None)))
except ValueError:
return json.dumps({
'error':
True,
'error_type':
'ValueError',
'message':
'playback must be an integer greater than 0.'
})
if request.args.get('position', None):
try:
position = abs(int(request.args.get('position', None)))
except ValueError:
return json.dumps({
'error':
True,
'error_type':
'ValueError',
'message':
'position must be an integer greater than 0.'
})
r_conn.set(election_key + '_PLAYBACK', str(playback))
if request.args.get('ratelimited', None) or request.args.get(
'errormode', None):
return json.dumps({"success": True})
else:
if ratelimited:
return make_response(
(RATELIMITED_STRING, 403, RATELIMITED_HEADERS))
if errormode:
if random.randrange(1, 3) % 2 == 0:
return make_response(
json.dumps({
"status": 500,
"error": True
}), 500, ERRORMODE_HEADERS)
if position + playback < (len(hopper) - 1):
"""
Needs the if statement here to set the position truly to zero if it's specified
in the url params.
"""
if request.args.get('position', None) or request.args.get(
'playback', None) or request.args.get(
'ratelimited', None) or request.args.get(
'errormode', None):
r_conn.set(election_key + '_POSITION', str(position))
else:
r_conn.set(election_key + '_POSITION', str(position + playback))
else:
r_conn.set(election_key + '_POSITION', str(len(hopper)))
with open(hopper[position - 1], 'r') as readfile:
payload = str(readfile.read())
return payload
def __init__(self,
inputsize,
taskcla,
use_processor=True,
processor_feats=(32, 128),
emb_size=100,
use_stem=1,
use_concat=False,
use_combination=True,
use_dropout=False,
gate="sigmoid"):
super(Net, self).__init__()
# safty checks
if use_stem is not None and use_stem >= 5:
raise ValueError(
"The value of use_stem ({}) is larger than the number of layers (5)!"
.format(use_stem))
# set internal values
ncha, size, _ = inputsize
self.taskcla = taskcla # contains tasks with number of classes
self.use_stem = use_stem # defines the number of stem layers to use (or None for none)
self.use_processor = utils.to_bool(
use_processor
) # defines if the pre-processor should be applied (or simply use task embeddings)
self.is_linear_processor = False if use_stem is None else use_stem > 3
self.use_combination = utils.to_bool(
use_combination
) # defines if attention masks should be generated through combination (to save weights)
self.use_concat = utils.to_bool(
use_concat
) # defines if input to the pre-processor should be concated
self.use_dropout = utils.to_bool(use_dropout)
self.emb_size = len(
taskcla
) if emb_size is None or use_processor is False else emb_size
# create all relevant convolutions (either native as stem or dwa masked)
self.mask_layers = torch.nn.ModuleList()
self.mask_shapes = []
self.c1, s, psize1 = self._create_conv(ncha, 64, size // 8, size, 1,
use_stem, inputsize)
self.c2, s, psize2 = self._create_conv(64, 128, size // 10, s, 2,
use_stem, psize1)
self.c3, s, psize3 = self._create_conv(128, 256, 2, s, 3, use_stem,
psize2)
self.smid = s
self.maxpool = torch.nn.MaxPool2d(2)
self.relu = torch.nn.ReLU()
# check if dropout sould be added or skipped (through identity)
if use_dropout is True:
self.drop1 = torch.nn.Dropout(0.2)
self.drop2 = torch.nn.Dropout(0.5)
else:
self.drop1 = torch.nn.Identity()
self.drop2 = torch.nn.Identity()
self.fc1, psize4 = self._create_linear(256 * self.smid * self.smid,
2048, 4, use_stem, psize3)
self.fc2, psize5 = self._create_linear(2048, 2048, 5, use_stem, psize4)
# define the names of the masks
self.mask_names = [
"c1.weight", "c2.weight", "c3.weight", "fc1.weight", "fc2.weight"
]
if use_stem is not None:
self.mask_names = self.mask_names[use_stem:]
# generate task processor
# all context processor stuff should start with 'p'
if use_processor is True:
# params
f_bn, f_out = processor_feats
self.processor_size = psize5
# adjust layers if input from FC
if self.is_linear_processor:
self.pfc1 = torch.nn.Linear(self.processor_size[0], f_bn)
self.pfc2 = torch.nn.Linear(f_bn, f_out)
self.pfc3 = torch.nn.Linear(f_out, self.emb_size)
#self.pfc3 = torch.nn.Embedding(100 * len(taskcla), self.emb_size)
else:
# check for input size and minimize
if self.processor_size[1] >= 14:
self.pc_min = torch.nn.MaxPool2d(2)
c, w, h = self.processor_size
self.processor_size = (c, w // 2, h // 2)
else:
self.pc_min = torch.nn.Identity()
# compute processor
self.pc1 = torch.nn.Conv2d(self.processor_size[0], f_bn,
(1, 1), (1, 1), 0)
self.pc2 = torch.nn.Conv2d(f_bn, f_out, (3, 3), (2, 2), 1)
cin = int(np.ceil(self.processor_size[1] / 2))
self.pfc1 = torch.nn.Linear(cin * cin * f_out, self.emb_size)
#self.pfc1 = torch.nn.Embedding(100 * len(taskcla), self.emb_size)
else:
self.task_eye = torch.eye(len(taskcla), requires_grad=False).cuda()
# generate all possible heads (and put them in list - list is needed for torch to properly detect layers)
self.last = torch.nn.ModuleList()
for t, n in self.taskcla:
self.last.append(torch.nn.Linear(2048, n))
# gates for this approach
if gate == "tanh":
self.gate = torch.nn.Tanh()
elif gate == "sigmoid":
self.gate = torch.nn.Sigmoid()
else:
print("ERROR: given gate {} is unkown, using sigmoid")
self.gate = torch.nn.Sigmoid()
return
def __init__(self, mod, title, version, doc,
usage_data=None, avail_modes=None,
supported_ui_toolkits=None,
run_as_root_ok=False, quiet=False):
self.mod = mod
self.title = title
self.version = version
self.doc = doc
self.usage_data = usage_data
os.umask(0037)
log.set_module(mod)
self.args = []
self.quiet = quiet
self.lock_file = None
prop.prog = sys.argv[0]
if os.getenv("HPLIP_DEBUG"):
log.set_level('debug')
self.avail_modes = avail_modes
if supported_ui_toolkits is not None:
self.supported_ui_toolkits = supported_ui_toolkits
self.num_supported_ui_toolkits = len(self.supported_ui_toolkits)
else:
self.supported_ui_toolkits = []
self.num_supported_ui_toolkits = 0
self.default_ui_toolkit = sys_conf.get('configure', 'ui-toolkit', 'qt4')
self.num_installed_ui_toolkits = 0
self.installed_ui_toolkits = []
if utils.to_bool(sys_conf.get('configure', 'qt3', '0')):
self.installed_ui_toolkits.append(UI_TOOLKIT_QT3)
self.num_installed_ui_toolkits += 1
if utils.to_bool(sys_conf.get('configure', 'qt4', '0')):
self.installed_ui_toolkits.append(UI_TOOLKIT_QT4)
self.num_installed_ui_toolkits += 1
self.default_mode = INTERACTIVE_MODE
self.num_valid_modes = 0
if self.avail_modes is not None:
if GUI_MODE in self.avail_modes and prop.gui_build and self.installed_ui_toolkits:
self.num_valid_modes += 1
if INTERACTIVE_MODE in self.avail_modes:
self.num_valid_modes += 1
if NON_INTERACTIVE_MODE in self.avail_modes:
self.num_valid_modes += 1
if self.avail_modes is not None:
if INTERACTIVE_MODE in self.avail_modes:
self.default_mode = INTERACTIVE_MODE
elif NON_INTERACTIVE_MODE in self.avail_modes:
self.default_mode = NON_INTERACTIVE_MODE
if self.supported_ui_toolkits is not None and prop.gui_build and self.installed_ui_toolkits:
if self.default_ui_toolkit == 'qt3' and UI_TOOLKIT_QT4 in self.supported_ui_toolkits and \
UI_TOOLKIT_QT3 not in self.supported_ui_toolkits and INTERACTIVE_MODE in self.avail_modes:
# interactive + qt4 and default is qt3 --> set to interactive (if avail) (e.g., hp-align)
self.default_mode = INTERACTIVE_MODE
self.default_ui_toolkit = 'none'
elif (UI_TOOLKIT_QT4 in self.supported_ui_toolkits and self.default_ui_toolkit == 'qt4' and UI_TOOLKIT_QT4 in self.installed_ui_toolkits) or \
(UI_TOOLKIT_QT3 in self.supported_ui_toolkits and self.default_ui_toolkit == 'qt3' and UI_TOOLKIT_QT3 in self.installed_ui_toolkits):
self.default_mode = GUI_MODE
elif self.default_ui_toolkit == 'qt3' and UI_TOOLKIT_QT3 not in self.supported_ui_toolkits:
if UI_TOOLKIT_QT4 in self.supported_ui_toolkits and UI_TOOLKIT_QT4 in self.installed_ui_toolkits: # (e.g, hp-linefeedcal?)
self.default_ui_toolkit = 'qt4'
self.default_mode = GUI_MODE
elif INTERACTIVE_MODE in self.avail_modes:
self.default_mode = INTERACTIVE_MODE
elif NON_INTERACTIVE_MODE in self.avail_modes:
self.default_mode = NON_INTERACTIVE_MODE
else:
log.error("%s cannot be run using Qt3 toolkit." % self.mod)
sys.exit(1)
elif self.default_ui_toolkit == 'qt4' and UI_TOOLKIT_QT4 not in self.supported_ui_toolkits:
if UI_TOOLKIT_QT3 in self.supported_ui_toolkits and UI_TOOLKIT_QT3 in self.installed_ui_toolkits: # (e.g., hp-unload)
self.default_ui_toolkit = 'qt3'
self.default_mode = GUI_MODE
elif INTERACTIVE_MODE in self.avail_modes:
self.default_mode = INTERACTIVE_MODE
elif NON_INTERACTIVE_MODE in self.avail_modes:
self.default_mode = NON_INTERACTIVE_MODE
else:
log.error("%s cannot be run using Qt4 toolkit." % self.mod)
sys.exit(1)
self.mode = self.default_mode
#log.debug("Default ui-toolkit: %s" % self.default_ui_toolkit)
#log.debug("Default mode: %s" % self.default_mode)
if os.getuid() == 0 and not run_as_root_ok:
log.warn("%s should not be run as root/superuser." % mod)
def __init__(self, arguments, model, train_sets):
self.logger = logging.getLogger(__name__)
self.model = model
self.train_sets = train_sets
#################### parse configs ####################
self.resume = False
if "resume" in arguments:
self.resume = utils.to_bool(arguments["resume"])
self.wdir = arguments["wdir"]
self.output_file = arguments["output_file"]
self.max_iters = 20
if "max_iters" in arguments:
self.max_iters = int(arguments["max_iters"])
#self.max_iters_without_impr = 3
#if "max_iters_without_impr" in arguments:
# self.max_iters_without_impr = int(arguments["max_iters_without_impr"])
self.first_layer_to_train = 0
if "first_layer_to_train" in arguments:
self.first_layer_to_train = int(arguments["first_layer_to_train"])
self.last_layer_to_train = model.n_layers - 1 # number hidden layers - 1
if "last_layer_to_train" in arguments:
self.last_layer_to_train = int(arguments["last_layer_to_train"])
# other stuff
if self.resume:
if not os.path.exists(self.wdir):
raise Exception("wdir must exist if resume=True")
else:
if not os.path.exists(self.wdir):
os.makedirs(self.wdir)
else:
self.logger.info("Directory already exists...")
out = StringIO.StringIO()
print >> out, "\n********** LayerwiseTrainer **********"
print >> out, "resume", self.resume
print >> out, "wdir", self.wdir
print >> out, "output_file", self.output_file
print >> out, "max_iters", self.max_iters
print >> out, "first_layer_to_train", self.first_layer_to_train
print >> out, "last_layer_to_train", self.last_layer_to_train
self.logger.info(out.getvalue())
self.mlp_init = self.wdir + "/mlp_init"
if not self.resume: # brand new
save(self.model, self.mlp_init)
# runtime state
self.layer_index = self.first_layer_to_train
self.iter = 0
self.loss = sys.float_info.max
self.mlp_best = self.mlp_init
self.mlp_crrnt = self.mlp_init
self.iters_without_impr = 0
if self.resume:
if os.path.isfile(self.wdir + "/layerwisetrainer_state"):
self._load_state()
def __init__(self, dataset_args, n_ins):
# stats
self.mean = None
self.std = None
if 'train_stat' in dataset_args.keys():
train_stat = dataset_args['train_stat']
featureStats = stats.FeatureStats()
featureStats.Load(train_stat)
self.mean = featureStats.GetMean()
self.std = featureStats.GetInvStd()
# open lstfile
file_path = dataset_args["lst_file"]
if file_path.endswith('.gz'):
file_read = gzip.open(file_path, 'r')
else:
file_read = open(file_path, 'r')
separate_lines = False
if "separate_lines" in dataset_args:
separate_lines = to_bool(dataset_args["separate_lines"])
self.has_labels = True
if "has_labels" in dataset_args:
self.has_labels = to_bool(dataset_args["has_labels"])
# parse it, file_lst is a list of (featureFile, labelFile) pairs in the input set
lines = [ln.strip() for ln in file_read]
lines = [ln for ln in lines if ln != "" ]
if self.has_labels:
if separate_lines:
if len(lines) % 2 != 0:
print("List has mis-matched number of feature files and label files")
sys.exit(1)
self.orig_file_lst = []
for i in xrange(0, len(lines), 2):
self.orig_file_lst.append((lines[i], lines[i+1]))
else:
self.orig_file_lst = []
for i in xrange(len(lines)):
pair = re.compile("\s+").split(lines[i])
if len(pair) != 2:
print(lines[i])
print("Each line in the train and eval lists must contain feature file and label file separated by space character")
sys.exit(1)
self.orig_file_lst.append(pair)
else:
# no labels
self.orig_file_lst = []
for i in xrange(0, len(lines), 1):
self.orig_file_lst.append((lines[i], None))
# save arguments
self.n_ins = n_ins
self.file_format = dataset_args['file_format']
self.file_format = "htk"
if 'file_format' in dataset_args:
self.file_format = dataset_args['file_format']
self.offsetLabels = False
if 'offset_labels' in dataset_args:
self.offsetLabels = to_bool(dataset_args['offset_labels'])
self.chunk_size = 32768
if 'gpu_chunk' in dataset_args:
self.chunk_size = int(dataset_args['gpu_chunk'])
self.maxFeats = 0
if "max_feats" in dataset_args:
self.maxFeats = int(dataset_args["max_feats"])
if self.maxFeats == 0:
self.maxFeats = sys.maxint
self.shuffle = True
if 'shuffle' in dataset_args:
self.shuffle = to_bool(dataset_args['shuffle'])
self.seed = None
if "seed" in dataset_args:
self.seed = int(dataset_args["seed"])
if int("_split_id" in dataset_args) + int("_num_splits" in dataset_args) == 1:
raise Exception("_split_id must be used with _num_splits")
self.num_splits = 0
if "_num_splits" in dataset_args:
self.num_splits = int(dataset_Args["_num_splits"])
self.split_id = dataset_args["_split_id"]
# internal state
self.split_parts = False
self.by_matrix = False
self.x = numpy.zeros((self.chunk_size, self.n_ins), dtype=numpy.float32)
if self.has_labels:
self.y = numpy.zeros((self.chunk_size,), dtype=numpy.int32)
else:
self.y = None
self.numpy_rng = numpy.random.RandomState(self.seed)
#self.make_shared()
self.initialize_read()
def __init__(self, dataset_args, n_ins):
# stats
self.mean = None
self.std = None
if 'train_stat' in dataset_args.keys():
train_stat = dataset_args['train_stat']
featureStats = stats.FeatureStats()
featureStats.Load(train_stat)
self.mean = featureStats.GetMean()
self.std = featureStats.GetInvStd()
# open lstfile
file_path = dataset_args["lst_file"]
if file_path.endswith('.gz'):
file_read = gzip.open(file_path, 'r')
else:
file_read = open(file_path, 'r')
separate_lines = False
if "separate_lines" in dataset_args:
separate_lines = to_bool(dataset_args["separate_lines"])
self.has_labels = True
if "has_labels" in dataset_args:
self.has_labels = to_bool(dataset_args["has_labels"])
# parse it, file_lst is a list of (featureFile, labelFile) pairs in the input set
lines = [ln.strip() for ln in file_read]
lines = [ln for ln in lines if ln != ""]
if self.has_labels:
if separate_lines:
if len(lines) % 2 != 0:
print("List has mis-matched number of feature files and label files")
sys.exit(1)
self.orig_file_lst = []
for i in xrange(0, len(lines), 2):
self.orig_file_lst.append((lines[i], lines[i + 1]))
else:
self.orig_file_lst = []
for i in xrange(len(lines)):
pair = re.compile("\s+").split(lines[i])
if len(pair) != 2:
print(lines[i])
print(
"Each line in the train and eval lists must contain feature file and label file separated by space character")
sys.exit(1)
self.orig_file_lst.append(pair)
else:
# no labels
self.orig_file_lst = []
for i in xrange(0, len(lines), 1):
self.orig_file_lst.append((lines[i], None))
# save arguments
self.n_ins = n_ins
self.file_format = dataset_args['file_format']
self.file_format = "htk"
if 'file_format' in dataset_args:
self.file_format = dataset_args['file_format']
self.offsetLabels = False
if 'offset_labels' in dataset_args:
self.offsetLabels = to_bool(dataset_args['offset_labels'])
self.chunk_size = 32768
if 'gpu_chunk' in dataset_args:
self.chunk_size = int(dataset_args['gpu_chunk'])
self.maxFeats = 0
if "max_feats" in dataset_args:
self.maxFeats = int(dataset_args["max_feats"])
if self.maxFeats == 0:
self.maxFeats = sys.maxint
self.shuffle = True
if 'shuffle' in dataset_args:
self.shuffle = to_bool(dataset_args['shuffle'])
self.seed = None
if "seed" in dataset_args:
self.seed = int(dataset_args["seed"])
if int("_split_id" in dataset_args) + int("_num_splits" in dataset_args) == 1:
raise Exception("_split_id must be used with _num_splits")
self.num_splits = 0
if "_num_splits" in dataset_args:
self.num_splits = int(dataset_Args["_num_splits"])
self.split_id = dataset_args["_split_id"]
# internal state
self.split_parts = False
self.by_matrix = False
self.x = numpy.zeros((self.chunk_size, self.n_ins), dtype=numpy.float32)
if self.has_labels:
self.y = numpy.zeros((self.chunk_size,), dtype=numpy.int32)
else:
self.y = None
self.numpy_rng = numpy.random.RandomState(self.seed)
# self.make_shared()
self.initialize_read()
import yaml
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError('SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "localhost"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections': int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '15')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses':os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path':os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
default=2500,
)
ap.add_argument("-st",
"--scroll_time",
help="How much time should I take to scroll?",
default=8)
args = vars(ap.parse_args())
print(args)
# ---------------------------------------------------------
# Global Variables
# ---------------------------------------------------------
# whether to download photos or not
download_uploaded_photos = utils.to_bool(args["uploaded_photos"])
download_friends_photos = utils.to_bool(args["friends_photos"])
# whether to download the full image or its thumbnail (small size)
# if small size is True then it will be very quick else if its false then it will open each photo to download it
# and it will take much more time
friends_small_size = utils.to_bool(args["friends_small_size"])
photos_small_size = utils.to_bool(args["photos_small_size"])
total_scrolls = int(args["total_scrolls"])
scroll_time = int(args["scroll_time"])
current_scrolls = 0
old_height = 0
driver = None
def update(self, tfg, validated_data):
try:
# comprobando titulo
if 'titulo' in validated_data.keys():
if validated_data.get('titulo') == '' or not utils.is_string(validated_data.get('titulo')):
raise NameError("Titulo incorrecto")
else:
tfg.titulo = validated_data.get('titulo')
# comprobando tipo
if 'tipo' in validated_data.keys():
if validated_data.get('tipo') == '' or not utils.is_string(validated_data.get('tipo')):
raise NameError("Tipo incorrecto")
else:
tfg.tipo = validated_data.get('tipo')
# comprobando n_alumnos
if 'n_alumnos' in validated_data.keys():
if ( int(validated_data.get('n_alumnos')) <= 0) or ( int(validated_data.get('n_alumnos')) > 3):
raise NameError("Numero de alumnos incorrecto")
else:
tfg.n_alumnos = validated_data.get('n_alumnos')
# comprobando descripcion
if 'descripcion' in validated_data.keys():
if validated_data.get('descripcion') == '' or not utils.is_string(validated_data.get('descripcion')):
raise NameError("Descripcion incorrecta")
else:
tfg.descripcion = validated_data.get('descripcion')
# comprobando conocimientos_previos
if 'conocimientos_previos' in validated_data.keys():
if validated_data.get('conocimientos_previos') == '' or \
not utils.is_string(validated_data.get('conocimientos_previos')):
raise NameError("Conocimientos Previos incorrectos")
else:
tfg.conocimientos_previos = validated_data.get('conocimientos_previos')
# comprobando hard_soft
if 'hard_soft' in validated_data.keys():
if validated_data.get('hard_soft') == '' or not utils.is_string(validated_data.get('hard_soft')):
raise NameError("Hard/Soft incorrectos")
else:
tfg.hard_soft = validated_data.get('hard_soft')
# comprobando tutor
if 'tutor' in validated_data.keys():
try:
tutor = Profesor.objects.get(email=validated_data.get('tutor'))
except:
raise NameError('El tutor no existe')
# if not isinstance(tutor, Profesor) or tutor.groups.filter(name='Profesores').exists():
# raise NameError("Tutor incorrecto")
# else:
tfg.tutor = tutor
# comprobando cotutor
if 'cotutor' in validated_data.keys():
if validated_data.get('cotutor') != '':
try:
cotutor = Profesor.objects.get(email=validated_data.get('cotutor'))
except:
raise NameError("El Cotutor no existe")
# if not isinstance(cotutor, Profesor) or not cotutor.groups.filter(name='Profesores').exists():
# raise NameError("Cotutor incorrecto")
# else:
tfg.cotutor = cotutor
else:
tfg.cotutor = None
# comprobando titulacion
if 'titulacion' in validated_data.keys():
try:
titulacion = Titulacion.objects.get(codigo=validated_data.get('titulacion'))
except:
raise NameError('La Titulacion no existe')
if not isinstance(titulacion, Titulacion):
raise NameError("Titulacion incorrecta")
else:
tfg.titulacion = titulacion
# comprobando publicado
if 'publicado' in validated_data.keys():
if validated_data.get('publicado') == '' or not utils.is_bool(validated_data.get('publicado')):
raise NameError("Valor Publicado invalido")
else:
tfg.publicado = utils.to_bool(validated_data.get('publicado'))
# comprobando validado
if 'validado' in validated_data.keys():
if validated_data.get('validado') == '' or not utils.is_bool(validated_data.get('validado')):
raise NameError("Valor Validado invalido")
else:
tfg.validado = utils.to_bool(validated_data.get('validado'))
# comprobando asignado
if 'asignado' in validated_data.keys():
if validated_data.get('asignado') == '' or not utils.is_bool(validated_data.get('asignado')):
raise NameError("Valor Publicado invalido")
else:
try:
tfg_asig = Tfg_Asig.objects.get(tfg=tfg)
except:
raise NameError('El Tfg asignado no existe')
if validated_data.get('asignado'):
tfg.asignado = utils.to_bool(validated_data.get('asignado'))
else:
try:
resul = self.serializer_class(tfg_asig).delete
tfg.asignado = utils.to_bool(validated_data.get('asignado'))
except:
raise NameError('Error al eliminar la asignacion del tfg')
tfg.save()
return dict(status=True, data=Tfg.objects.get(titulo=tfg.titulo))
except NameError as e:
return dict(status=False, message=e.message)
def __init__(self,
model,
nepochs=200,
sbatch=32,
lr=0.075,
lr_min=1e-4,
lr_factor=3,
lr_patience=5,
warmup=[10, 750],
clipgrad=10000,
curriculum="linear:100:0.2",
log_path=None,
sparsity=0.2,
bin_sparsity=False,
alpha=1.0,
lamb_loss=[10, 0.05],
lamb_reg=500,
delta=1,
stiff=None,
use_anchor_first=False,
scale_att_loss=False,
use_task_loss=False,
use_apex=False):
super().__init__(model, nepochs, sbatch, lr, lr_min, lr_factor,
lr_patience, warmup, clipgrad, curriculum, log_path,
AMP_READY and use_apex)
# set parameters
print(
"Setting Parameters to:\n\tsparsity: {}{}\n\talpha: {}\n\tdelta: {}\n\tlambda: {} / {}\n\tanchor (first task): {}\n\tscale att: {}"
.format(sparsity, " (bin)" if bin_sparsity is True else "", alpha,
delta, lamb_loss, lamb_reg, use_anchor_first,
scale_att_loss))
self.sparsity = sparsity
self.alpha = alpha
if isinstance(lamb_loss, list) or isinstance(lamb_loss, tuple):
self.lamb_loss = lamb_loss
else:
self.lamb_loss = (lamb_loss, lamb_loss)
self.lamb_reg = lamb_reg
self.delta = delta
self.bin_sparse = utils.to_bool(bin_sparsity)
self.use_anchor_first = utils.to_bool(use_anchor_first)
self.scale_attention = utils.to_bool(scale_att_loss)
self.use_task_loss = utils.to_bool(use_task_loss)
self.stiff = stiff
# define constants used over training
self.fisher = None
self.anchor_neg = None
self.anchor_pos = None
self.anchor_task = None
self.anchor_store = [None] * len(model.taskcla)
# some anchor settings
self.anchor_thres = 0.4 # complexity threshold for anchor data (not use to high complexity to avoid confusion)
self.anchor_batches = 10 # number of batches to use for anchor training
self.max_layers = 5
# helper to improve time on sparsity
self.sparsity_rates = {}
# generate all task losses
if self.use_task_loss is True:
print("INFO: Generating task loss")
num_tasks = len(model.taskcla)
emb_size = model.emb_size
elements = emb_size / num_tasks
ite = np.zeros((num_tasks, emb_size), np.float32)
for i in range(num_tasks):
ite[i, int(i * elements):int((i + 1 * elements))] = 1
self.ideal_task_embs = torch.from_numpy(ite).cuda()
self.emb_mse_criterion = torch.nn.MSELoss(reduction='none')
return
def get_model(dataset):
verbose = utils.to_bool(request.args.get('verbose', VERBOSE))
explain = utils.to_bool(request.args.get('explain', EXPLAIN['enable']))
method = request.args.get('method', DEFAULT_METHOD)
return models.QueryIndex.build_query_index(
dataset, verbose=verbose, explain=explain, method=method)
import yaml
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError('SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "localhost"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '15')),
'enable_whiplash': to_bool(os.environ.get('SIXPACK_CONFIG_WHIPLASH', 'False')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses':os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path':os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
import yaml
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError('SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '11024')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "pub-redis-11024.us-east-1-4.3.ec2.garantiadata.com"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', "Lsbtz4NIlAeKu5J5"),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections': int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '0')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses':os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path':os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError(
'SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "localhost"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections': int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '15')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses': os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path': os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
def __init__(self, arguments, model, train_fn, valid_fn, train_sets,
valid_sets):
self.logger = logging.getLogger(__name__)
self.model = model
self.train_fn = train_fn
self.valid_fn = valid_fn
self.train_sets = train_sets
self.valid_sets = valid_sets
#################### parse configs ####################
self.resume = False
if "resume" in arguments:
self.resume = utils.to_bool(arguments["resume"])
self.wdir = arguments["wdir"]
self.output_file = arguments["output_file"]
self.learn_rate = 0.1
if "learn_rate" in arguments:
self.learn_rate = float(arguments["learn_rate"])
self.halving_factor = 0.5
if "halving_factor" in arguments:
self.halving_factor = float(arguments["halving_factor"])
self.max_iters = 20
if "max_iters" in arguments:
self.max_iters = int(arguments["max_iters"])
self.min_iters = 0
if "min_iters" in arguments:
self.min_iters = int(arguments["min_iters"])
self.keep_lr_iters = 15
if "keep_lr_iters" in arguments:
self.keep_lr_iters = int(arguments["keep_lr_iters"])
self.start_halving_impr = 0.01
if "start_halving_impr" in arguments:
self.start_halving_impr = float(arguments["start_halving_impr"])
self.end_halving_impr = 0.001
if "end_halving_impr" in arguments:
self.end_halving_impr = float(arguments["end_halving_impr"])
self.continue_with_rate = False
if "continue_with_rate" in arguments:
self.continue_with_rate = utils.to_bool(
arguments["continue_with_rate"])
self.halving_criteria = "loss"
if "halving_criteria" in arguments:
self.halving_criteria = arguments["halving_criteria"]
criteria_list = ["loss", "frame_err"]
if self.halving_criteria not in criteria_list:
raise Exception("invalid halving criteria. must be one of " +
str(criteria_list))
# batch_size and momentum
self.batch_size = 256
if arguments.has_key('batch_size'):
self.batch_size = int(arguments['batch_size'])
self.momentum = 0.5
self.momentum_start = 1
if arguments.has_key('momentum'):
self.momentum = float(arguments['momentum'])
if 'momentum_start' in arguments:
self.momentum_start = int(arguments['momentum_start'])
# other stuff
if self.resume:
if not os.path.exists(self.wdir):
raise Exception("wdir must exist if resume=True")
else:
if not os.path.exists(self.wdir):
os.makedirs(self.wdir)
else:
self.logger.info("Directory already exists...")
out = StringIO.StringIO()
print >> out, "\n********** Trainer **********"
print >> out, "resume", self.resume
print >> out, "wdir", self.wdir
print >> out, "output_file", self.output_file
print >> out, "learn_rate", self.learn_rate
print >> out, "halving_factor", self.halving_factor
print >> out, "max_iters", self.max_iters
print >> out, "min_iters", self.min_iters
print >> out, "keep_lr_iters", self.keep_lr_iters
print >> out, "start_halving_impr", self.start_halving_impr
print >> out, "end_halving_impr", self.end_halving_impr
print >> out, "continue_with_rate", self.continue_with_rate
print >> out, "halving_criteria", self.halving_criteria
print >> out, "batch_size", self.batch_size
print >> out, "momentum", self.momentum
print >> out, "momentum_start", self.momentum_start
self.logger.info(out.getvalue())
self.mlp_init = self.wdir + "/mlp_init"
if not self.resume: # brand new
save(self.model, self.mlp_init)
# runtime state
self.iter = 0
self.done = False
self.loss = sys.float_info.max
self.rate = self.learn_rate
self.mlp_best = self.mlp_init
self.halving = False
self.wasAccepted = True
if self.resume:
if os.path.isfile(self.wdir + "/trainer_state"):
self._load_state()
import yaml
import os
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError('SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled': to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port': int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host': os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "localhost"),
'redis_password': os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix': os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout': os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name': os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections': int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db': int(os.environ.get('SIXPACK_CONFIG_REDIS_DB', '15')),
'robot_regex': os.environ.get('SIXPACK_CONFIG_ROBOT_REGEX', "$^|trivial|facebook|MetaURI|butterfly|google|"
"amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|"
"Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|"
"slurp|java|fetch|spider|url|crawl|oneriot|abby|"
"commentreader|twiceler"),
'ignored_ip_addresses':os.environ.get('SIXPACK_CONFIG_IGNORE_IPS', "").split(","),
'asset_path':os.environ.get('SIXPACK_CONFIG_ASSET_PATH', "gen"),
from utils import to_bool
config_path = os.environ.get('SIXPACK_CONFIG', None)
if config_path:
try:
CONFIG = yaml.safe_load(open(config_path, 'r'))
except IOError:
raise RuntimeError(
'SIXPACK_CONFIG - {0} - is an invalid path'.format(config_path))
except yaml.YAMLError, exc:
raise RuntimeError('Error in configuration file: {0}'.format(str(exc)))
else:
CONFIG = {
'enabled':
to_bool(os.environ.get('SIXPACK_CONFIG_ENABLED', 'True')),
'redis_port':
int(os.environ.get('SIXPACK_CONFIG_REDIS_PORT', '6379')),
'redis_host':
os.environ.get('SIXPACK_CONFIG_REDIS_HOST', "localhost"),
'redis_password':
os.environ.get('SIXPACK_CONFIG_REDIS_PASSWORD', None),
'redis_prefix':
os.environ.get('SIXPACK_CONFIG_REDIS_PREFIX', "sxp"),
'redis_socket_timeout':
os.environ.get('SIXPACK_CONFIG_REDIS_SOCKET_TIMEOUT', None),
'redis_sentinel_service_name':
os.environ.get('SIXPACK_CONFIG_REDIS_SENTINEL_SERVICE_NAME', None),
'redis_max_connections':
int(os.environ.get('SIXPACK_CONFIG_REDIS_MAX_CONNECTIONS', '0')),
'redis_db':
def update(self, tfg, validated_data):
try:
# comprobando titulo
if 'titulo' in validated_data.keys():
if validated_data.get('titulo') == '' or not utils.is_string(
validated_data.get('titulo')):
raise NameError("Titulo incorrecto")
else:
tfg.titulo = validated_data.get('titulo')
# comprobando tipo
if 'tipo' in validated_data.keys():
if validated_data.get('tipo') == '' or not utils.is_string(
validated_data.get('tipo')):
raise NameError("Tipo incorrecto")
else:
tfg.tipo = validated_data.get('tipo')
# comprobando n_alumnos
if 'n_alumnos' in validated_data.keys():
if (int(validated_data.get('n_alumnos')) <= 0) or (int(
validated_data.get('n_alumnos')) > 3):
raise NameError("Numero de alumnos incorrecto")
else:
tfg.n_alumnos = validated_data.get('n_alumnos')
# comprobando descripcion
if 'descripcion' in validated_data.keys():
if validated_data.get(
'descripcion') == '' or not utils.is_string(
validated_data.get('descripcion')):
raise NameError("Descripcion incorrecta")
else:
tfg.descripcion = validated_data.get('descripcion')
# comprobando conocimientos_previos
if 'conocimientos_previos' in validated_data.keys():
if validated_data.get('conocimientos_previos') == '' or \
not utils.is_string(validated_data.get('conocimientos_previos')):
raise NameError("Conocimientos Previos incorrectos")
else:
tfg.conocimientos_previos = validated_data.get(
'conocimientos_previos')
# comprobando hard_soft
if 'hard_soft' in validated_data.keys():
if validated_data.get(
'hard_soft') == '' or not utils.is_string(
validated_data.get('hard_soft')):
raise NameError("Hard/Soft incorrectos")
else:
tfg.hard_soft = validated_data.get('hard_soft')
# comprobando tutor
if 'tutor' in validated_data.keys():
try:
tutor = Profesor.objects.get(
email=validated_data.get('tutor'))
except:
raise NameError('El tutor no existe')
# if not isinstance(tutor, Profesor) or tutor.groups.filter(name='Profesores').exists():
# raise NameError("Tutor incorrecto")
# else:
tfg.tutor = tutor
# comprobando cotutor
if 'cotutor' in validated_data.keys():
if validated_data.get('cotutor') != '':
try:
cotutor = Profesor.objects.get(
email=validated_data.get('cotutor'))
except:
raise NameError("El Cotutor no existe")
# if not isinstance(cotutor, Profesor) or not cotutor.groups.filter(name='Profesores').exists():
# raise NameError("Cotutor incorrecto")
# else:
tfg.cotutor = cotutor
else:
tfg.cotutor = None
# comprobando titulacion
if 'titulacion' in validated_data.keys():
try:
titulacion = Titulacion.objects.get(
codigo=validated_data.get('titulacion'))
except:
raise NameError('La Titulacion no existe')
if not isinstance(titulacion, Titulacion):
raise NameError("Titulacion incorrecta")
else:
tfg.titulacion = titulacion
# comprobando publicado
if 'publicado' in validated_data.keys():
if validated_data.get('publicado') == '' or not utils.is_bool(
validated_data.get('publicado')):
raise NameError("Valor Publicado invalido")
else:
tfg.publicado = utils.to_bool(
validated_data.get('publicado'))
# comprobando validado
if 'validado' in validated_data.keys():
if validated_data.get('validado') == '' or not utils.is_bool(
validated_data.get('validado')):
raise NameError("Valor Validado invalido")
else:
tfg.validado = utils.to_bool(
validated_data.get('validado'))
# comprobando asignado
if 'asignado' in validated_data.keys():
if validated_data.get('asignado') == '' or not utils.is_bool(
validated_data.get('asignado')):
raise NameError("Valor Publicado invalido")
else:
try:
tfg_asig = Tfg_Asig.objects.get(tfg=tfg)
except:
raise NameError('El Tfg asignado no existe')
if validated_data.get('asignado'):
tfg.asignado = utils.to_bool(
validated_data.get('asignado'))
else:
try:
resul = self.serializer_class(tfg_asig).delete
tfg.asignado = utils.to_bool(
validated_data.get('asignado'))
except:
raise NameError(
'Error al eliminar la asignacion del tfg')
tfg.save()
return dict(status=True, data=Tfg.objects.get(titulo=tfg.titulo))
except NameError as e:
return dict(status=False, message=e.message)
import os
from flask import Flask
from flask import jsonify
from utils import to_bool
POD_NAME = os.environ.get("POD_NAME")
POD_IP = os.environ.get("POD_IP")
NODE_NAME = os.environ.get("NODE_NAME")
app = Flask(__name__)
app.config['JSONIFY_PRETTYPRINT_REGULAR'] = True
@app.route('/')
def meta():
return jsonify({
'META': 'Here is some info about this pod',
'POD_NAME': POD_NAME,
'POD_IP': POD_IP,
'NODE_NAME': NODE_NAME,
})
if __name__ == '__main__':
app.run(
debug=to_bool(os.environ.get('DEBUG')),
host='0.0.0.0'
)
def __init__(self, arguments, model, train_sets):
self.logger = logging.getLogger(__name__)
self.model = model
self.train_sets = train_sets
#################### parse configs ####################
self.resume = False
if "resume" in arguments:
self.resume = utils.to_bool(arguments["resume"])
self.wdir = arguments["wdir"]
self.output_file = arguments["output_file"]
self.max_iters = 20
if "max_iters" in arguments:
self.max_iters = int(arguments["max_iters"])
#self.max_iters_without_impr = 3
#if "max_iters_without_impr" in arguments:
# self.max_iters_without_impr = int(arguments["max_iters_without_impr"])
self.first_layer_to_train = 0
if "first_layer_to_train" in arguments:
self.first_layer_to_train = int(arguments["first_layer_to_train"])
self.last_layer_to_train = model.n_layers - 1 # number hidden layers - 1
if "last_layer_to_train" in arguments:
self.last_layer_to_train = int(arguments["last_layer_to_train"])
# other stuff
if self.resume:
if not os.path.exists(self.wdir):
raise Exception("wdir must exist if resume=True")
else:
if not os.path.exists(self.wdir):
os.makedirs(self.wdir)
else:
self.logger.info("Directory already exists...")
out = StringIO.StringIO()
print >>out, "\n********** LayerwiseTrainer **********"
print >>out, "resume", self.resume
print >>out, "wdir", self.wdir
print >>out, "output_file", self.output_file
print >>out, "max_iters", self.max_iters
print >>out, "first_layer_to_train", self.first_layer_to_train
print >>out, "last_layer_to_train", self.last_layer_to_train
self.logger.info(out.getvalue())
self.mlp_init = self.wdir + "/mlp_init"
if not self.resume: # brand new
save(self.model, self.mlp_init)
# runtime state
self.layer_index = self.first_layer_to_train
self.iter = 0
self.loss = sys.float_info.max
self.mlp_best = self.mlp_init
self.mlp_crrnt = self.mlp_init
self.iters_without_impr = 0
if self.resume:
if os.path.isfile(self.wdir+"/layerwisetrainer_state"):
self._load_state()
