def main ():
manager = Manager()
# Range Sensor Distance
distance = manager.list([0, 0, 0, 0, 0, 0])
# DataBase
db = manager.list()
# Command from Keyboard
keyboard_cmd = manager.dict()
keyboard_cmd['vld'] = 0
keyboard_cmd['cmd'] = ''
# GoBack Ready
goback_cmd = manager.dict()
goback_cmd['vld'] = 0
#Step 1: Initialzation
initialization.initialize()
#Step 2: Departure
operations.take_off()
#Step 3: Multi threading start
to_thr = Process(target = timeout, args = (runtime, ))
key_thr = Process(target = keyboard_thread.keyboard_thread_wrapper_function, args = (keyboard_cmd, ))
cmd_thr = Process(target = command_thread.command_thread_wrapper_function, args = (distance, keyboard_cmd, db, goback_cmd, ))
rs_thr = Process(target = range_sensor_thread.range_sensor_wrapper_function, args = (distance, ))
gb_thr = Process(target = goback_thread.goback_wrapper_function, args = (goback_cmd, ))
cmd_thr.start()
rs_thr.start()
to_thr.start()
key_thr.start()
gb_thr.start()
#Step 4: Multi threading end
to_thr.join()
cmd_thr.terminate()
rs_thr.terminate()
key_thr.terminate()
gb_thr.terminate()
operations.landing()
time.sleep(1)
cleanup.cleanup()
def fit_model(structure, data_matrix, old_root=None, gibbs_steps=200):
if old_root is None:
X = data_matrix.sample_latent_values(np.zeros((data_matrix.m, data_matrix.n)), 1.)
old_root = GaussianNode(X, 'scalar', 1.)
root = initialization.initialize(data_matrix, old_root, old_root.structure(), structure, num_iter=gibbs_steps)
model = models.get_model(structure, fixed_noise_variance=data_matrix.fixed_variance())
models.align(root, model)
dumb_samplers.sweep(data_matrix, root, num_iter=gibbs_steps)
dumb_samplers.sweep(data_matrix, root, maximize=True, num_iter=1)
return root
def main(dev_run: ("Runs all the training with 1 iteration only", 'option',
'd'),
version="0.1"):
# dev_run is a flag, which sets all the training iterations to 1,
# for the purposes of testing the assembly process
if dev_run:
development_run = True
else:
development_run = False
VERSION = version
# converting the resources into JSON files
from conversion import convert_resources
convert_resources()
print("\nResources converted\n")
# initializing the model
from initialization import initialize
initialize()
print("\nModel initialized\n")
# training the components, this may naturally take long time
# for training them separately at different times, just run the respective scripts
from tagger_training import train_tagger
train_tagger(development_run)
print("\nTagger trained\n")
from parser_training import train_parser
train_parser(development_run)
print("\nParser trained\n")
from ner_training import train_ner
train_ner(development_run)
print("\nNER component trained\n")
# assembling the model
from assembly import assemble_model
assemble_model(VERSION)
print("\nModel assembled\n")
def main ():
loop = 0
manager = Manager()
distance = manager.list([0, 0, 0, 0, 0, 0])
# distance2 = manager.list([0, 0, 0, 0, 0, 0])
#Step 1: Initialzation
initialization.initialize()
#Step 2: Departure
operations.take_off()
#Step 3: Multi threading start
timeout_thread = Process(target = timeout, args = (runtime, ))
p1 = Process(target = command_thread.command_thread_wrapper_function, args = (distance, ))
p2 = Process(target = range_sensor_thread.range_sensor_wrapper_function, args = (distance, ))
p1.start()
p2.start()
timeout_thread.start()
#Step 4: Multi threading end
timeout_thread.join()
p1.terminate()
p2.terminate()
time.sleep(1)
cleanup.cleanup()
def main ():
loop = 0
#Step 1: Initialzation
initialization.initialize()
#Step 2: Departure
operations.take_off()
#while True :
while (loop < 300): #time for a loop: 0.15s
loop += 1
#Step 2: Detect Range
# for i in range(0, range_sensor.num_directions):
# distance[i] = range_sensor.detect_range(i)
distance[4] = range_sensor.detect_range(4)
time.sleep(0.05)
#Step 3: Analyze Range
# if (distance[4] > 180):
# #operations.hover()
# operations.move_backward()
# elif (distance[4] < 180):
# #operations.hover()
# operations.move_backward()
# else:
# #operations.hover()
# operations.move_backward()
#Step 4: Generate Next Command
if (loop < 50):
operations.hover()
elif (loop < 100):
operations.move_backward()
elif (loop < 150):
operations.hover()
elif (loop < 200):
operations.move_backward()
elif (loop < 250):
operations.drop()
else :
operations.drop()
cleanup.cleanup()
def pmm(data, k, max_iterations=20, threshold=1e-4, verbose=False):
"""
Fits a poisson mixture model to the data with k latent classes.
Args:
- data: data to assign to different mixtures, shape (num_samples, dim)
- k: number of latent classes
Return Values:
- means: values of means params, shape (k, dim)
- phis: probs of different classes, shape (k)
- log_prob: last computed log probability of data
- w: responsibilities of different classes for each sample, shape (num_samples, k)
"""
# initialize the mean parameters for each class
means, assignments, _ = initialization.initialize(data, k, num_runs=15)
phis = np.zeros(k)
for assignment in assignments:
phis[assignment] += 1
phis /= len(assignments)
if verbose:
utils.plot_1d_data_assigments(data, means, assignments)
# repeatedly run e-step and m-step until convergence
# initialize some values to reuse or return
prev_log_prob, log_prob, w = 0, -1, None
for idx in range(max_iterations):
# e-step
w, log_prob = mm.e_step(data, means, phis, density=utils.log_poisson)
# m-step
means, phis = m_step(data, k, w)
# check for convergence
diff = abs(log_prob - prev_log_prob)
if diff < threshold:
break
else:
prev_log_prob = log_prob
if verbose:
# check how it's going
utils.plot_1d_data_responsibilities(data, w, means)
print 'log prob: {:.5f}\tchange in log prob: {:.5f}'.format(log_prob, diff)
return means, phis, log_prob, w
def load_cookies(driver):
cookies = pickle.load(open('cookies.txt', "rb"))
driver.delete_all_cookies()
# have to be on a page before you can add any cookies, any page - does not matter which
driver.get("https://amazon.com")
for cookie in cookies:
driver.add_cookie(cookie)
print('cookies loaded successfully')
if __name__ == '__main__':
dr = initialize()
action = actions()
#load_cookies(dr)
dr.get('https://instagram.com')
time.sleep(50)
follow_spree(dr)
#unfollow_spree(dr)
#save_cookies(dr)
dr.quit()
import pyxel
import random
import initialization
from ballclass import Ball
from playerclass import Player
from missileclass import Missile
pyxel.init(256, 256)
initialization.initialize()
counter = 0
player = Player()
balls = []
state = 'GAMEOVER'
hiscore = 0
missiles = []
def update():
global state
if state == 'GAMEOVER':
if pyxel.btn(pyxel.KEY_ENTER):
start_game()
state = 'PLAYING'
else:
update_playing()
def start_game():
global counter, balls, player
counter = 0
verbose = False
for o, a in opts:
if o == "-v":
verbose = True
elif o in ("-h", "--help"):
alphaBotUtility.usage()
sys.exit()
elif o in ("-k", "--handleOK"):
handleOK = True
elif o in ("-d", "--debug"):
debug = True
elif o in ("-r", "--report"):
report.reportDirectory = str(a)
elif o in ("-e", "--explorer"):
report = str(a)
reportExplorer.analyseReport(report)
sys.exit()
elif o in ("-l", "--exploreLatest"):
reportExplorer.analyseReport(None)
sys.exit()
else:
assert False, "unhandled option"
captchaDir = "./captcha"
if not os.path.exists(captchaDir):
os.makedirs(captchaDir)
initialization.initialize(handleOK=handleOK)
bot.run(handleOK=handleOK, debug=debug)
def do_POST(self):
"""
definiert den Umgang mit POST Requests
Liest den Body aus - gibt in zum konvertieren weiter
"""
global dataset_info, net, experiment_id, max_display, dataset, _neighbors
if self.path == "/nodes":
print("post /nodes")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
#self.send_header('Access-Control-Allow-Origin', self.headers['origin'])
self.end_headers()
# get body from request
content_len = int(self.headers['Content-Length'])
body = self.rfile.read(content_len)
# convert body to list
data = json.loads(str(body).decode('utf-8')) # python 2
#data = json.loads(str(body, encoding='utf-8')) # python 3
# print(data)
# Katjas code goes here
reset(experiment_id, dataset=dataset)
_neighbors = {'positive': [], 'negative': []}
net, dataset_info = initialize(dataset=dataset, experiment_id=experiment_id, batch_size=100,
lr=1e-4)
experiment_id = dataset_info['experiment_id']
# introduce scale factor
limits = (-15, 15)
pmax, pmin = dataset_info['position'].max(), dataset_info['position'].min()
dataset_info['scale_func'] = lambda x: np.divide((limits[1]-limits[0]) * (x.copy() - pmin), pmax-pmin) + limits[0]
dataset_info['inverse_scale_func'] = lambda x: np.divide((x.copy()-limits[0]) * (pmax-pmin), limits[1]-limits[0]) + pmin
if N is None:
N = len(dataset_info['name'])
nodes = make_nodes(position=dataset_info['scale_func'](dataset_info['position'][:N]),
name=dataset_info['name'][:N],
label=dataset_info['label'][:N],
index=True)
categories = dataset_info['categories']
data = {'nodes': nodes, 'categories': categories}
# make json
data = json.dumps(data).encode()
self.wfile.write(data) #body zurueckschicken
if self.path == "/trainSvm":
print("post /trainsvm")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
self.end_headers()
# get body from request
content_len = int(self.headers['Content-Length'])
body = self.rfile.read(content_len)
# convert body to list
data = json.loads(str(body).decode('utf-8')) # python 2
#data = json.loads(str(body, encoding='utf-8')) # python 3
print(data)
# Katjas code goes here
# p, n = katja_function(data.p, data.n)
# make json
# data = json.dumps({p: p, n: n}).encode()
self.wfile.write(data) #body zurueckschicken
if self.path == "/stopSvm":
print("post /stopSvm")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
self.end_headers()
# get body from request
#content_len = int(self.headers['Content-Length'])
#body = self.rfile.read(content_len)
# convert body to list
#data = json.loads(str(body).decode('utf-8')) # python 2
#data = json.loads(str(body, encoding='utf-8')) # python 3
#print(data)
# Katjas code goes here
# p, n = katja_function(data.p, data.n)
# make json
#data = json.dumps({p: p, n: n}).encode()
self.wfile.write("stopped Svm") #body zurueckschicken
if self.path == "/updateLabels":
print("post /updateLabels")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
self.end_headers()
# get body from request
#content_len = int(self.headers['Content-Length'])
#body = self.rfile.read(content_len)
# convert body to list
#data = json.loads(str(body).decode('utf-8')) # python 2
#data = json.loads(str(body, encoding='utf-8')) # python 3
#print(data)
# Katjas code goes here
# katja_function(data.p, data.n)
# make json
#data = json.dumps({}).encode()
self.wfile.write(data) #body zurueckschicken
if self.path == "/getGroupNeighbours":
print("post /getGroupNeighbours")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
self.end_headers()
# get body from request
content_len = int(self.headers['Content-Length'])
body = self.rfile.read(content_len)
# convert body to list
data = json.loads(str(body).decode('utf-8')) # python 2
# data = json.loads(str(body, encoding='utf-8')) # python 3
print(data)
print(len(data['positives']))
try:
_neighbors['negative'].extend(data['negatives'])
except KeyError:
pass
_neighbors['positive'], scores = select_neighbors(dataset_info['feature'][:N],
data['positives'], _neighbors['negative'],
k=2000, neighbor_fn=svm_k_nearest_neighbors, test=False)
print('New neighbors {}'.format(_neighbors['positive']))
# Katjas code goes here
# katja_function(data.p, data.n)
# make json
data['neighbours'] = {neigh: score for neigh, score in zip(_neighbors['positive'], scores)}
print(len(data['positives']))
data['group'] = list(data['positives'])
data = json.dumps(data).encode()
self.wfile.write(data) #body zurueckschicken
if self.path == "/startUpdateEmbedding":
print("post /startUpdateEmbedding")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
self.end_headers()
# get body from request
content_len = int(self.headers['Content-Length'])
body = self.rfile.read(content_len)
# convert body to list
body = json.loads(str(body).decode('utf-8')) # python 2
# data = json.loads(str(body, encoding='utf-8')) # python 3
#print(body)
#print(self.socket_id)
self.socket_id = body['socketId']
id = body['socketId']
data = read_nodes(body['nodes'])
self.wfile.write('update_embedding started for ' + str(self.socket_id)) # body zurueckschicken
# Katjas code goes here
new_position = np.stack([data['x'], data['y']], axis=1)
new_position = dataset_info['inverse_scale_func'](new_position)
old_position = dataset_info['position']
idx_modified = get_modified(old_position[:N], new_position, tol=dataset_info['inverse_scale_func'](np.array([3,])))
if len(idx_modified) == 0 or len(idx_modified) == len(dataset_info['name']): # TODO: fix recall bug
print('Modified {} samples. - Invalid for training.')
return 0
# idx_old_neighbors = get_neighborhood(old_position[:N], idx_modified)
# idx_new_neighbors = get_neighborhood(new_position[:N], idx_modified)
idx_old_neighbors, _ = mutual_k_nearest_neighbors(old_position[:N], idx_modified, k=50)
idx_new_neighbors = get_neighborhood(new_position[:N], idx_modified)
print('Train MapNet using {} positives and {} negatives.'.format(len(idx_modified) + len(idx_new_neighbors),
len(_neighbors['negative'])))
# net, dataset_info = initialize(dataset=dataset, experiment_id=experiment_id, batch_size=100,
# lr=1e-4)
# experiment_id = dataset_info['experiment_id']
# # introduce scale factor
# limits = (-15, 15)
# pmax, pmin = dataset_info['position'].max(), dataset_info['position'].min()
# dataset_info['scale_func'] = lambda x: np.divide((limits[1] - limits[0]) * (x.copy() - pmin), pmax - pmin) + \
# limits[0]
# dataset_info['inverse_scale_func'] = lambda x: np.divide((x.copy() - limits[0]) * (pmax - pmin),
# limits[1] - limits[0]) + pmin
# if N is None:
# N = len(dataset_info['name'])
new_position = train(net, dataset_info['feature'][:N], dataset_info['name'].values.astype(str)[:N],
old_position[:N], new_position,
idx_modified, idx_old_neighbors, idx_new_neighbors,
_neighbors['negative'],
categories=dataset_info['categories'], label=dataset_info['label'][:N],
lr=1e-3, experiment_id=experiment_id, socket_id=self.socket_id,
scale_func=dataset_info['scale_func'])
dataset_info['position'] = new_position
_neighbors = {'positive': [], 'negative': []}
return 0
# TODO was ist wenn das mehrfach gestartet wird
# self.inter = SetInterval(0.6, update_embedding_handler, id)
# self.inter.socket_id = id
# self.inter.start()
# t = threading.Timer(5, self.inter.cancel)
# t.start()
# make json
# data = json.dumps({}).encode()
if self.path == "/stopUpdateEmbedding":
print("post /stopUpdateEmbedding")
### POST Request Header ###
self.send_response(200)
self.send_header('Content-type', 'application/json')
self.end_headers()
return
verbose = False
for o, a in opts:
if o == "-v":
verbose = True
elif o in ("-h", "--help"):
alphaBotUtility.usage()
sys.exit()
elif o in ("-k", "--handleOK"):
handleOK = True
elif o in ("-d", "--debug"):
debug = True
elif o in ("-r", "--report"):
report.reportDirectory = str(a)
elif o in ("-e", "--explorer"):
report = str(a)
reportExplorer.analyseReport(report)
sys.exit()
elif o in ("-l", "--exploreLatest"):
reportExplorer.analyseReport(None)
sys.exit()
else:
assert False, "unhandled option"
captchaDir = "./captcha"
if not os.path.exists(captchaDir):
os.makedirs(captchaDir)
initialization.initialize(handleOK=handleOK)
bot.run(handleOK=handleOK, debug=debug)