def __init__(self, env_name):
self.env_name = env_name
env = gym.make(self.env_name)
self.action_size = env.action_space.n
self.state_size = env.observation_space.shape
env.close()
self.EPISODES = 10000 # total episodes to train through all environments
self.episode = 0 # used to track the episodes total count of episodes played through all thread environments
self.max_average = 0 # when average score is above 0 model will be saved
self.lr = 0.00025
self.epochs = 10 # training epochs
self.shuffle=False
self.Training_batch = 1000
self.replay_count = 0
self.optimizer = optim.Adam
self.writer = SummaryWriter(comment="_"+self.env_name+"_"+self.optimizer.__name__+"_"+str(self.lr))
# Instantiate plot memory
self.scores_, self.episodes_, self.average_ = [], [], [] # used in matplotlib plots
# Instantiate games and plot memory
self.eps = np.finfo(np.float32).eps.item()
self.scores, self.episodes, self.average = [], [], []
self.save_path = 'Models'
if not os.path.exists(self.save_path): os.makedirs(self.save_path)
self.actor_path = os.path.join(self.save_path, f'{self.env_name}_Actor_{self.lr}_torch.h5')
self.critic_path = os.path.join(self.save_path, f'{self.env_name}_Critic_{self.lr}_torch.h5')
# Create ActorCritic network model
self.actor, self.critic = getModels(input_shape=self.state_size, action_space = self.action_size)
self.actor_optimizer = self.optimizer(params = self.actor.parameters(), lr=self.lr)
self.critic_optimizer = self.optimizer(params = self.critic.parameters(), lr=self.lr)
def solution5():
print("Start problem 5")
print("Start data preparation for problem 5")
dataPreparation_for_5()
print("Results: ")
model = getModels()
for i in range(1, 11):
for j in range(1, 4):
filename = 'problem 5 sample' + str(i) + '_period' + str(
j) + '.txt'
tweets, parameters = [], []
prediction = []
#f = open('problem 5 sample'+str(i)+'_period'+'j'+'.txt')
if os.path.isfile('problem 5 sample' + str(i) + '_period' +
str(j) + '.txt'):
print filename, 'problem 5 sample' + str(i) + '_period' + str(
j) + '.txt'
f = open('problem 5 sample' + str(i) + '_period' + str(j) +
'.txt')
line = f.readline()
while len(line):
p = line.split()
tweets.append(float(p[0]))
parameters.append([float(p[i]) for i in range(len(p))])
line = f.readline()
# del(tweets[0])
# next_hour_tweets = tweets
# parameters.pop()
#print parameters, tweets
# X=[1,1,1,1,1]
# y=[1]
# model=[]
# for i in range(3):
# clf=linear_model.LinearRegression()
# clf.fit(X,y)
# model.append(clf)
clf = model[j - 1]
prediction = clf.predict(parameters)
#print parameters, tweets
print "prediction of tweets in next hour of " + filename
print prediction[-1]
f.close()
def solution5():
print("Start problem 5")
print("Start data preparation for problem 5")
dataPreparation_for_5()
print("Results: ")
model=getModels()
for i in range(1,11):
for j in range(1,4):
filename='problem 5 sample'+str(i)+'_period'+str(j)+'.txt'
tweets,parameters=[],[]
prediction=[]
#f = open('problem 5 sample'+str(i)+'_period'+'j'+'.txt')
if os.path.isfile('problem 5 sample'+str(i)+'_period'+str(j)+'.txt'):
print filename,'problem 5 sample'+str(i)+'_period'+str(j)+'.txt'
f = open('problem 5 sample'+str(i)+'_period'+str(j)+'.txt')
line = f.readline()
while len(line):
p = line.split()
tweets.append(float(p[0]))
parameters.append([float(p[i]) for i in range(len(p))])
line = f.readline()
# del(tweets[0])
# next_hour_tweets = tweets
# parameters.pop()
#print parameters, tweets
# X=[1,1,1,1,1]
# y=[1]
# model=[]
# for i in range(3):
# clf=linear_model.LinearRegression()
# clf.fit(X,y)
# model.append(clf)
clf=model[j-1]
prediction=clf.predict(parameters)
#print parameters, tweets
print "prediction of tweets in next hour of "+filename
print prediction[-1]
f.close()
def fitSermivariogramModel(self, modelname, nlag=15, tsill=None,
trange=None, tnugget=0.0):
"""Fit a semivariogram model to the datas
Returns a model with the best parameters (sill, range, nugget) to
fit the datas (semivariance against distance)
Parameters
----------
modelname : String
The desired model name to use. One of:
Spherical, Exponential, Gaussian, Pentaspherical, Nugget
nlag : Integer
If not None, the semivariogram will be fitted using nlags
bins (empirical semivariogram), between distance[0, max], else
all the datas will be used (global semivariogram).
tsill : Float
Temporary initial sill, used by the fitted algorithm. If
None it will be estimated from the datas.
trange : Float
Temporary initial range, used by the fitted algorithm.
If None it will be estimated from the datas.
tnugget : Float
Temporary initial nugget, used by the fitted algorithm.
Returns
----------
a model instance with sill, range and nugget computed to best fit
the datas. Highly depending from the initial inputs, due to the used
algorithm (scipy.optimize.leastsqr)
"""
# the input datas
dist = self.infos.dist
svar = self.infos.svar
# lag?
if nlag is None:
nlag = len(dist)
if nlag < 1:
raise ValueError, "nlag must be >=1"
# Initials parameters estimations
# TODO : find best solutions
if tsill is None:
tsill = 9/10*np.max(svar)
if trange is None:
trange = 0.5*np.max(dist)
# Sort by distance
sortind = np.argsort(dist)
sortdist = dist[sortind]
sortsvar = svar[sortind]
# Select by bins
index = sortdist.searchsorted(np.linspace(0, dist.max(), nlag+1))
dist = [sortdist[index[i-1]:index[i]].mean() for i in range(1, len(index))]
svar = [sortsvar[index[i-1]:index[i]].mean() for i in range(1, len(index))]
# Retrieve the model class from his name
model = models.getModels(sill=tsill, range=trange, nugget=tnugget)[modelname]
# these are our inital guesses for the sill and range
params = (model.sill, model.range, model.nugget)
# perform the least squares
lstsqResult = optimize.leastsq(model.residual, params, args=(dist, svar), full_output=0)
if(lstsqResult != 1):
p = lstsqResult[0]
model.sill, model.range, model.nugget = p
if model.range > 0:
# work out the square deviation too
squareDeviates = model.residual(p, dist, svar)**2
# we divide by number of points minus the degrees of freedom
denom = 1 / float(len(svar)-len(p))
model.variance = denom*sum(squareDeviates)
pprint([["Model Type", model.type], ["Sill", model.sill],
["Range", model.range], ["Nugget", model.nugget]])
return model
else:
pprint([["Error", "Computed range <=0 :("],])
return None
else:
pprint([["Error", "Bad fitting computation :("],])
return None
"""
helps with stuff
"""
from sys import argv
from os import system
from models import getModels
from peewee import SqliteDatabase
if "--create" in argv:
print "[+] Creating Tables..."
db = SqliteDatabase("devClone.db")
Rant, Action = getModels(db)
db.create_tables([Rant, Action])
elif "--recreate" in argv:
print "[+] Recreating..."
system("rm devClone.db")
system("python tool.py --create")
padding='post',
value=0)
decoder_out = pad_sequences(decoder_out,
maxlen=MAX_SEQ_LEN,
padding="post",
value=0)
print("decoder input shape", decoder_inp.shape, "decoder output shape",
decoder_out.shape)
images = result.image_name.apply(
lambda x: 'flickr30k_images/flickr30k_images/' + x).values
SAMPLE_SIZE = len(images)
print('SAMPLE_SIZE', SAMPLE_SIZE)
model, captionGenerator = getModels(embedding_matrix,
MAX_SEQ_LEN,
LATENT_DIM=100)
#custom acc because we dont want to consider padding
def acc(y_true, y_pred):
# both are of shape ( _, Ty, VOCAB_SIZE )
targ = K.argmax(y_true, axis=-1)
pred = K.argmax(y_pred, axis=-1)
correct = K.cast(K.equal(targ, pred),
dtype='float32') #cast bool tensor to float
# 0 is padding, don't include those- mask is tensor representing non-pad value
mask = K.cast(K.greater(targ, 0),
dtype='float32') #cast bool-tensor to float
n_correct = K.sum(mask * correct) #
def fitSermivariogramModel(self,
modelname,
nlag=15,
tsill=None,
trange=None,
tnugget=0.0):
"""Fit a semivariogram model to the datas
Returns a model with the best parameters (sill, range, nugget) to
fit the datas (semivariance against distance)
Parameters
----------
modelname : String
The desired model name to use. One of:
Spherical, Exponential, Gaussian, Pentaspherical, Nugget
nlag : Integer
If not None, the semivariogram will be fitted using nlags
bins (empirical semivariogram), between distance[0, max], else
all the datas will be used (global semivariogram).
tsill : Float
Temporary initial sill, used by the fitted algorithm. If
None it will be estimated from the datas.
trange : Float
Temporary initial range, used by the fitted algorithm.
If None it will be estimated from the datas.
tnugget : Float
Temporary initial nugget, used by the fitted algorithm.
Returns
----------
a model instance with sill, range and nugget computed to best fit
the datas. Highly depending from the initial inputs, due to the used
algorithm (scipy.optimize.leastsqr)
"""
# the input datas
dist = self.infos.dist
svar = self.infos.svar
# lag?
if nlag is None:
nlag = len(dist)
if nlag < 1:
raise ValueError, "nlag must be >=1"
# Initials parameters estimations
# TODO : find best solutions
if tsill is None:
tsill = 9 / 10 * np.max(svar)
if trange is None:
trange = 0.5 * np.max(dist)
# Sort by distance
sortind = np.argsort(dist)
sortdist = dist[sortind]
sortsvar = svar[sortind]
# Select by bins
index = sortdist.searchsorted(np.linspace(0, dist.max(), nlag + 1))
dist = [
sortdist[index[i - 1]:index[i]].mean()
for i in range(1, len(index))
]
svar = [
sortsvar[index[i - 1]:index[i]].mean()
for i in range(1, len(index))
]
# Retrieve the model class from his name
model = models.getModels(sill=tsill, range=trange,
nugget=tnugget)[modelname]
# these are our inital guesses for the sill and range
params = (model.sill, model.range, model.nugget)
# perform the least squares
lstsqResult = optimize.leastsq(model.residual,
params,
args=(dist, svar),
full_output=0)
if (lstsqResult != 1):
p = lstsqResult[0]
model.sill, model.range, model.nugget = p
if model.range > 0:
# work out the square deviation too
squareDeviates = model.residual(p, dist, svar)**2
# we divide by number of points minus the degrees of freedom
denom = 1 / float(len(svar) - len(p))
model.variance = denom * sum(squareDeviates)
pprint([["Model Type", model.type], ["Sill", model.sill],
["Range", model.range], ["Nugget", model.nugget]])
return model
else:
pprint([
["Error", "Computed range <=0 :("],
])
return None
else:
pprint([
["Error", "Bad fitting computation :("],
])
return None