def simulated_annealing(scenario, time_limit=20000):
# simulated annealing (described here: # https://en.wikipedia.org/wiki/Simulated_annealing)
# is generally an optimization algorithm that tweaks the state by a little each time
# until reaching optimum. in order not to get stuck in local optima, the process sometimes
# chooses tweaks with high energy (bad score) in order to shake things up and allow for new pathways
# as the optimization proceeds, the "temperature" lowers and high energy tweaks become less and less
# likely.
# the basic version only mutates the route using swaps, AKA a swap between two consecutive nodes
# time_limit: amount of allowed computation time in milliseconds
start_time = time_stamp()
scenario_len = scenario.shape[0]
# normalizer that makes the scale (width and height) of the scenario irrelevant
distance_normalizer = get_average_edge_length(scenario)
# initiate a route randomly
route = scenario.copy()
np.random.shuffle(route)
# score of each possible swap
energy_deltas = get_energy_deltas(route, distance_normalizer,
np.arange(scenario_len))
while time_stamp() < start_time + time_limit:
# figure the current temperature
stamp = time_stamp()
temperature = get_temperature(stamp - start_time, time_limit)
# get the thresholds for all swaps
thresholds = get_thresholds(energy_deltas, temperature)
# get a random number
random_num = np.random.random()
# get a random swap with threshold above the random number
available_swaps = np.arange(
thresholds.shape[0])[thresholds > random_num]
if available_swaps.size > 0:
winner_swap = np.random.choice(available_swaps, 1)[0]
# swap the nodes
i, j = swap_index_to_node_indices(winner_swap, scenario_len)
route[[i, j]] = route[[j, i]]
# recalculate the affected energies
affected_indices = get_indices_affected_by_swap(
winner_swap, scenario_len)
new_energy_deltas = get_energy_deltas(route, distance_normalizer,
affected_indices)
energy_deltas[affected_indices] = new_energy_deltas
# turn the coordinates route to indices route each pointing
# to its corresponding node in the original scenario array.
return xy_route_to_indices_route(scenario, route)
def simulated_annealing(scenario, time_limit=40000, use_flips=True, use_pops=True):
# simulated annealing (described here: # https://en.wikipedia.org/wiki/Simulated_annealing)
# is generally an optimization algorithm that tweaks the state by a little each time
# until reaching optimum. in order not to get stuck in local optima, the process sometimes
# chooses tweaks with high energy (bad score) in order to shake things up and allow for new pathways
# as the optimization proceeds, the "temperature" lowers and high energy tweaks become less and less
# likely.
# the advanced version is basically an improvement over local_search, it checks for both segment
# flips and node pops.
# time_limit: amount of allowed computation time in milliseconds
start_time = time_stamp()
counter = 0
# normalizer that makes the scale (width and height) of the scenario irrelevant
distance_normalizer = get_average_edge_length(scenario)
# initiate a route randomly
route = scenario.copy()
np.random.shuffle(route)
while time_stamp() < start_time + time_limit:
# figure the current temperature
stamp = time_stamp()
temperature = get_temperature(stamp - start_time, time_limit)
# get a random threshold, route length changes below that threshold will be accepted
random_threshold = get_random_threshold(temperature, distance_normalizer)
if use_flips:
# the two endpoints of the segment to flip
segment_start, segment_end = find_segment_flip(route, threshold=random_threshold)
# if succeeded to find a flip below the threshold
if segment_start is not None:
route[segment_start + 1: segment_end] = route[segment_end - 1: segment_start: -1]
counter += 1
if use_pops:
pop_from, place_at = find_pop(route, threshold=random_threshold)
# if succeeded to find a pop below the threshold
if pop_from is not None:
# simulating a python list pop() and insert() using math
if pop_from > place_at:
route[place_at:pop_from + 1] = route[np.r_[pop_from, place_at:pop_from]]
else:
route[pop_from:place_at] = route[np.r_[pop_from + 1:place_at, pop_from]]
continue
# turn the coordinates route to indices route each pointing
# to its corresponding node in the original scenario array.
return xy_route_to_indices_route(scenario, route)
def drinker_pay(self, drinker_name, amount):
"""
Drinker ``drinker_name`` pays some amount ``amount``.
This function would be called via RPC somehow.
:param str drinker_name:
:param Decimal|int|str amount:
:return: updated Drinker
:rtype: Drinker
"""
amount = Decimal(amount)
print("%s: %s pays %s %s." %
(time_stamp(), drinker_name, amount, self.currency))
with self.lock:
drinker = self.get_drinker(drinker_name)
drinker.credit_balance += amount
if drinker.credit_balance >= 0:
# Reset counts in this case.
drinker.buy_item_counts.clear()
self._save_drinker(drinker)
# We want to have a Git commit right after (after the lock release), so enforce this now.
self._add_git_commit_drinkers_task(wait_time=0)
self.admin_cash_position.cash_position += amount
self._save_admin_cash_position()
for cb in self.update_drinker_callbacks:
cb(drinker_name)
return drinker
def drinker_buy_item(self, drinker_name, item_name, amount=1):
"""
:param str drinker_name:
:param str item_name: intern name
:param int amount: can be negative, to undo drinks
:return: updated Drinker
:rtype: Drinker
"""
print("%s: %s drinks %s (amount: %i)." %
(time_stamp(), drinker_name, item_name, amount))
assert isinstance(amount, int)
with self.lock:
drinker = self.get_drinker(drinker_name)
item = self._get_buy_item_by_intern_name(item_name)
drinker.buy_item_counts.setdefault(item_name, 0)
drinker.buy_item_counts[item_name] += amount
drinker.total_buy_item_counts.setdefault(item_name, 0)
drinker.total_buy_item_counts[item_name] += amount
drinker.credit_balance -= item.price * amount
self._save_drinker(drinker)
if amount != 1:
# We want to have a Git commit right after (after the lock release), so enforce this now.
self._add_git_commit_drinkers_task(wait_time=0)
for cb in self.update_drinker_callbacks:
cb(drinker_name)
return drinker
def __init__(self, s_dim, t_dim, a_dim, architecture, type):
super().__init__()
self.architecture = architecture
self._s_dim = s_dim
self._t_dim = t_dim
self._a_dim = a_dim
self._type = type
assert type in ['multitask'], 'Invalid agent type.'
self._id = time_stamp()
def create_token(self, special: bool = False):
salt = random_string(4)
if special:
h = _hash(salt, self.__hw_pin, None)
return salt + ":" + b2a_base64_trimed(h)
else:
tm = str(time_stamp())
h = _hash(salt, self.__dev_sec, tm)
return salt + ":" + b2a_base64_trimed(h) + ":" + tm
def _get_data(self, b):
self.__handler_data = { #NOSONAR
'd': self.name,
'tm': time_stamp(),
's': [{
'n': 'alert',
'v': 1
}]
}
def __init__(self, concept_architecture, first_level_agent, s_dim, a_dim, t_dim):
super().__init__()
self.concept_architecture = concept_architecture
self.first_level_agent = first_level_agent
freeze(self.first_level_agent)
self._s_dim = s_dim
self._a_dim = a_dim
self._t_dim = t_dim
self._id = time_stamp()
def local_search(scenario,
initiate_greedy=True,
use_segment_flip=True,
use_pop=True,
time_limit=np.inf):
stop_time = time_stamp() + time_limit
# initiate a route
if initiate_greedy:
route = scenario[greedy(scenario)]
else:
route = scenario[random_walk(scenario)]
while time_stamp() < stop_time:
if use_segment_flip:
# the two endpoints of the segment to flip
segment_start, segment_end = find_segment_flip(route)
if segment_start is not None:
route[segment_start +
1:segment_end] = route[segment_end - 1:segment_start:-1]
continue
if use_pop:
pop_from, place_at = find_pop(route)
# if found a pop that shortens the route
if pop_from is not None:
# simulating a python list pop() and insert() using math
if pop_from > place_at:
route[place_at:pop_from +
1] = route[np.r_[pop_from, place_at:pop_from]]
else:
route[pop_from:place_at] = route[np.r_[pop_from +
1:place_at,
pop_from]]
continue
break
return xy_route_to_indices_route(scenario, route)
def __init__(self, n_actions, second_level_architecture, first_level_actor, noop_action, temporal_ratio=5):
super().__init__()
self.second_level_architecture = second_level_architecture
self.first_level_actor = first_level_actor
freeze(self.first_level_actor)
self._n_actions = n_actions + int(noop_action)
self._first_level_a_dim = self.first_level_actor._a_dim
self._noop = noop_action
self._temporal_ratio = temporal_ratio
self._id = time_stamp()
def store_database(database, n_parts):
part_size = len(database.buffer) // n_parts
DB_ID = time_stamp()
os.makedirs(SAVE_PATH + DB_ID)
for i in range(0, n_parts):
PATH = SAVE_PATH + DB_ID + '/SAC_training_level2_database_part_' + str(i) + '.p'
if (i+1) < n_parts:
pickle.dump(list(itertools.islice(database.buffer, part_size*i, part_size*(i+1))), open(PATH, 'wb'))
else:
pickle.dump(list(itertools.islice(database.buffer, part_size*i, None)), open(PATH, 'wb'))
async def async_sensor_values(self):
'''
异步模式获取值
'''
from uasyncio import sleep_ms
if self.__prepare != None:
self.__prepare()
await sleep_ms(self.__delay)
l = []
for s in self.__sensors:
l.append(s.value)
v = {}
v['d'] = self.__name
v['s'] = l
v['tm'] = time_stamp()
return v
#%%
print("Importing modules")
import pandas as pd
import os
import datetime
from datetime import date, timedelta
from forex_python.bitcoin import BtcConverter
from forex_python.converter import CurrencyRates
from sales import caluclate_sales
from utils import time_stamp, consolidate_data, label_df, recategorise, add_cuts
#%%
TIME_STAMP = time_stamp()
SOURCE_PATH = "DATA/INPUT/SCRAPED"
SAVE_PATH = "DATA/OUTPUT"
#%%
print("Consolidating scraped data")
# Pulling raw scraped data from CSVs, I really need to get into SQL
df = consolidate_data(SOURCE_PATH, SAVE_PATH, TIME_STAMP)
# Consistent records only exist from Feb 4th
START_DATE = datetime.datetime(2020, 2, 4)
END_DATE = pd.to_datetime(datetime.date.today())
df = df[(df.time_stamp >= START_DATE) & (df.time_stamp < END_DATE)]
# Stripping out and reordering columns
df = df[[
"source",
def __init__(self, s_dim, latent_dim, n_concepts, noisy=True, lr=1e-4):
super().__init__(s_dim, latent_dim, n_concepts, noisy, lr)
self._id = time_stamp()
arg('-n', '--n-img', type=int, default=20),
# pipeline
arg('-is', '--sch', action='store_true'),
arg('-rs', '--rsch', action='store_true'),
arg('-pred', action='store_true'),
# use saved
arg('-load-urls'),
arg('-load-preds'),
)
name = opts.name + '__' if is_(opts.name) else ''
RESULT_PREFIX = osp.join('reverse-img-final-preds',
'%s_to_%s' % (opts.src, opts.target),
name + time_stamp())
mkdir_p(RESULT_PREFIX)
fh = init_logging(file=osp.join(RESULT_PREFIX, 'log.log'), stdout=True)
LOGGER = get_logger(__name__, main=True)
from nlp_utils import get_words
from image_search import image_search
from reverse_image_search import reverse_search_urls
queries = []
if is_(opts.file):
queries.extend(get_words(opts.file, i=opts.start, j=opts.stop))
if is_(opts.query):
queries.extend(opts.query)
def setup(
self,
fetch_cmds=None,
pass_conds=None,
cry_cmds=None,
cry_retries=None,
cry_pretries=None,
celeb_cmds=None,
pass_logs=None,
csv_file_path=None,
log_while_waiting=None,
ongoing=None,
wait_after_celeb=None,
**kwargs,
):
"""
sets up class parameters:
fetch_mds: commands to run when pass values are invalid. These might be establidhsing connection
pass_conds: pass conditions as list of texts. A pass_cond=True always passes.
In case of [] or None or not passing, cry_cmds will be used to create verification condition:
"=" in statements will be replaced by "==". non-statement commands will be ignored.
"""
if kwargs:
# merge with existing
self.kwargs = {**self.kwargs, **kwargs}
if wait_after_celeb:
self.wait_after_celeb = wait_after_celeb
if ongoing:
self.ongoing = ongoing
# every one of cmds and conds pass this point,
# so its best to esxpand them here
self.fetch_cmds = expand_pmac_stats(
fetch_cmds if fetch_cmds else self.fetch_cmds, **self.kwargs)
self.fetch_cmds_parsed = parse_cmds(self.fetch_cmds)
self.cry_cmds = expand_pmac_stats(
cry_cmds if cry_cmds else self.cry_cmds, **self.kwargs)
self.cry_cmds_parsed = parse_cmds(self.cry_cmds)
self.celeb_cmds = expand_pmac_stats(
celeb_cmds if celeb_cmds else self.celeb_cmds, **self.kwargs)
self.celeb_cmds_parsed = parse_cmds(self.celeb_cmds)
if cry_retries:
self.cry_retries = cry_retries
if cry_pretries:
self.cry_pretries = cry_pretries
# TODO fix: this makes an arbitrary pass_cond when cry_cmds are changed
# via setup, i.e. no NEW pass_conds are supplied.
# dirty fix is to check if there are only pass_conds parsed, then
# don't change them.
if (not self.pass_conds) and (not pass_conds) and (cry_cmds):
# an empty pass-cond (but not a None) means: check for all of the command statements:
pass_conds = stats_to_conds(cry_cmds)
self.pass_conds = expand_pmac_stats(
pass_conds if pass_conds else self.pass_conds, **self.kwargs)
self.pass_conds_parsed = parse_stats(self.pass_conds)
# pass_logs_parsed need to be fetched with pass_conds_parsed, stored, and logged at celeb.
self.pass_logs = expand_pmac_stats(
pass_logs if pass_logs else self.pass_logs, **self.kwargs)
self.pass_logs_parsed = parse_stats(self.pass_logs)
if csv_file_path:
self.csv_file_name = csv_file_path
# setup the headers, they get written when (and only if) the first set of readings are ready
if self.pass_logs_parsed:
headers = ["Time"] + list(list(zip(*self.pass_logs_parsed))[2])
# remove and reshape special caharacters headers
headers = [normalise_header(header) for header in headers]
self.csvcontent = ",".join(map(str, headers)) + "\n"
if self.csvcontent and self.csv_file_name:
# time_stamp the filename
self.csv_file_stamped = utils.time_stamp(
self.csv_file_name)
# if file exists, make a backup of the existing file
# do not leave until the file doesn't exist!
n_copies = 0
while os.path.exists(self.csv_file_stamped):
name, ext = os.path.splitext(self.csv_file_stamped)
modif_time_str = time.strftime(
"%y%m%d_%H%M",
time.localtime(
os.path.getmtime(self.csv_file_stamped)),
)
n_copies_str = f"({n_copies})" if n_copies > 0 else ""
try:
os.rename(
self.csv_file_stamped,
f"{name}_{modif_time_str}{n_copies_str}{ext}",
)
except FileExistsError:
# forget it... the file is already archived...
# TODO or you need to be too fussy and break the execution for this?
n_copies += 1
open(self.csv_file_stamped, "w+")
else:
# self.log_vals = []
self.csvcontent = None
if log_while_waiting:
self.log_while_waiting = log_while_waiting
# TODO change this crap[ solution
# floating digits used for == comparison
self.ndigits = 6
super().setup(**self.kwargs)
return self
def publish_op_log(self, p, c, ret):
x = {'p': p, 'c': c, 'r': ret, 'tm': time_stamp()}
if ENCRYPTED_OUTPUT:
st = Sec()
x = st.enc_paylaod(x)
return self.publish(x, topic = OP_LOG_TOPIC)