def fit(self, epochs, batch_size=32, verbose=1, workers=10):
self.createModel()
with open('layoutannotations.json') as f:
annotations = json.load(f)
_, split = load()
with open('lstminput.json') as f:
lstminput = json.load(f)
training_generator = DataGeneratorEntity(annotations=annotations,
video_files=split['train'],
F=self.F,
batch_size=batch_size,
LSTM=self.LSTM,
lstminput=lstminput)
validation_generator = DataGeneratorEntity(annotations=annotations,
video_files=split['val'],
F=self.F,
batch_size=batch_size,
LSTM=self.LSTM,
lstminput=lstminput)
self.model.fit_generator(generator=training_generator,
validation_data=validation_generator,
epochs=epochs,
use_multiprocessing=True,
workers=workers,
verbose=verbose)
return
def fit(self, epochs, batch_size=32):
self.createModel()
with open('layoutannotations.json') as f:
annotations = json.load(f)
_, split = load()
with open('lstminput.json') as f:
lstminput = json.load(f)
tf.flags.DEFINE_integer("batch_size", 32, "Batch size during training")
tf.flags.DEFINE_integer("eval_batch_size", 8,
"Batch size during evaluation")
training_generator = DataGeneratorLayout(annotations=annotations,
video_files=split['train'],
F=self.F,
batch_size=8,
LSTM=self.LSTM,
lstminput=lstminput,
graph=self.graph)
validation_generator = DataGeneratorLayout(annotations=annotations,
video_files=split['val'],
F=self.F,
batch_size=8,
LSTM=self.LSTM,
lstminput=lstminput,
graph=self.graph)
self.model.fit_generator(generator=training_generator,
validation_data=validation_generator,
epochs=50,
use_multiprocessing=True,
workers=10)
'''
def main():
filename = 'data/inflammation-01.csv'
data = loaddata.load(filename)
print(filename)
print(data.mean(axis=1))
def f**k():
annotations, split = load()
layoutanno = dict()
i = 1
for annotation in annotations:
layoutanno.update({annotation['globalID']:annotation})
print("[{}] ".format(i)+annotation['globalID'])
i = i + 1
with open('layoutannotations.json','w') as fp:
json.dump(layoutanno,fp)
return
def fit(self, epochs, batch_size=32, verbose=1, workers=30):
self.createModel()
#self.model.load_weights('LayoutCheckpoints/weights-improvement-15-15603.58.hdf5')
#opt = Adam(lr=0.001, decay=0.5, amsgrad=False) #weight decay 0.0001
#self.model.compile(optimizer=opt, loss={'dense_4':self.loss2, 'activation_1':self.loss1}, loss_weights=[1, 1], metrics=[])
with open('layoutannotations.json') as f:
annotations = json.load(f)
_, split = load()
with open('lstminput.json') as f:
lstminput = json.load(f)
tf.flags.DEFINE_integer("batch_size", 32, "Batch size during training")
tf.flags.DEFINE_integer("eval_batch_size", 8, "Batch size during evaluation")
filepath="LayoutCheckpoints/weights-improvement-{epoch:02d}-{val_loss:.2f}.hdf5"
checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode='min')
training_generator = DataGeneratorLayout(annotations=annotations, video_files=split['train'], F=self.F, batch_size=batch_size, LSTM=self.LSTM, lstminput=lstminput, graph=self.graph)
validation_generator = DataGeneratorLayout(annotations=annotations, video_files=split['val'], F=self.F, batch_size=batch_size, LSTM=self.LSTM, lstminput=lstminput, graph=self.graph)
self.history = self.model.fit_generator(generator=training_generator, validation_data=validation_generator, epochs=epochs, use_multiprocessing=True, workers=workers, verbose=verbose, callbacks=[checkpoint])
self.model.save('LayoutComposerModel.h5')
'''
# -*- coding: utf-8 -*-
import numpy as np
import loaddata as ld
import gradientdescent as GD
import normalization as norm
X, Y, n = ld.load('data.txt')
X = norm.normalize(X, n)
X = X.reshape((n * 2))
tmp = []
for i in xrange(0, 2 * n, 2):#Оторвать мне руки
tmp.append(1)
tmp.append(X[i])
tmp.append(X[i + 1])
X = np.array(tmp).reshape(n, 3)
print X
alpha = 0.01;
iterations = 400;
theta = np.zeros((3, 1))#init fitting params
import numpy as np
import loaddata as ld
import cost
X, Y, m = ld.load('data.txt')
theta = np.zeros((2 + 1, 1))
print cost.costfunc(theta, X, Y, m)
import numpy as np
import loaddata as ld
import computeCost as cC
import gradientdescent as GD
import math
X, Y, m = ld.load('ex1data.txt')#load X's and Y's m - len of dataset
theta = np.zeros((2, 1))#init fitting params
#should be 32.07
print 'Cost:', cC.compCost(X, Y, theta)
#Some gradient descent settings
iterations = 1500;
alpha = 0.01;
theta,J_history = GD.GDescent(X, Y, theta, alpha, iterations)
print 'theta: ', theta
pvalue = 3.5
#predict = 1/(1+ math.exp( np.dot( np.array([1, pvalue]).reshape(1, 2), theta)))
predict = np.dot( np.array([1, pvalue]).reshape(1, 2), theta )
import numpy as np
import loaddata as ld
import computeCost as cC
import gradientdescent as GD
import math
X, Y, m = ld.load('ex1data.txt') #load X's and Y's m - len of dataset
theta = np.zeros((2, 1)) #init fitting params
#should be 32.07
print 'Cost:', cC.compCost(X, Y, theta)
#Some gradient descent settings
iterations = 1500
alpha = 0.01
theta, J_history = GD.GDescent(X, Y, theta, alpha, iterations)
print 'theta: ', theta
pvalue = 3.5
#predict = 1/(1+ math.exp( np.dot( np.array([1, pvalue]).reshape(1, 2), theta)))
predict = np.dot(np.array([1, pvalue]).reshape(1, 2), theta)
print 'predict for ', pvalue, predict
from scipy.io import savemat
import loaddata
import scipy
import numpy as np
import torch
import os
torch.manual_seed(0)
dataset_name = 'ppi'
for noise_level in [0, 1.0, 2.0, 3.0]:
a1, f1, a2, f2, ground_truth = loaddata.load(dataset_name,
noise_level=noise_level)
print(f1, f2)
feature_size = f1.size(1)
ns = [f1.size(0), f2.size(0)]
# edge_list_1 = get_edge_list(a1, f1.size(0))
# edge_list_2 = get_edge_list(a2, f2.size(0))
# features = [f1, f2]
# edges = [a1, a2]
f = open('ppi_combined_edges.txt', 'w')
print(ns)
print(a1.size())
print(ground_truth)
n = ns[0]
print(n)
t1 = (a1[0] < a1[1]).nonzero()
t2 = (a2[0] < a2[1]).nonzero()
a1 = a1[:, t1]
a2 = a2[:, t2]
g = ground_truth
import numpy as np
import json
from lstm import lstm
from loaddata import load
from keras.models import *
import cv2
annotations, _ = load()
F = 75
'''n_words=6414
n_tags=42
LSTM = lstm(hidden=64)
LSTM.load_weights('entity_lstm.h5')'''
from preprocess import preProcessData
from keras.models import Model, Input
from keras.layers import *
from keras_contrib.layers import CRF
from keras.utils import plot_model
from keras.models import load_model
import json
import numpy as np
from keras.callbacks import ModelCheckpoint
#from livelossplot import PlotLossesKeras
BATCH_SIZE = 512 # Number of examples used in each iteration
EPOCHS = 20 # Number of passes through entire dataset
MAX_LEN = 75 # Max length of review (in words)
EMBEDDING = 100 # Dimension of word embedding vector
# -*- coding: utf-8 -*-
import numpy as np
import loaddata as ld
import gradientdescent as GD
import normalization as norm
X, Y, n = ld.load('data.txt')
X = norm.normalize(X, n)
X = X.reshape((n * 2))
tmp = []
for i in xrange(0, 2 * n, 2): #Оторвать мне руки
tmp.append(1)
tmp.append(X[i])
tmp.append(X[i + 1])
X = np.array(tmp).reshape(n, 3)
print X
alpha = 0.01
iterations = 400
theta = np.zeros((3, 1)) #init fitting params
def main():
filename=sys.argv[1] # a new change!!
data = loaddata.load(filename)
print filename
print data.mean(axis=1)
args.__dict__[t] = arg_dict[t]
except:
print('Error in loading config and use default setting instead')
print(args)
if args.setup == 1:
args.net = GCNNet
elif args.setup == 2:
args.net = GATNet
elif args.setup == 3 or args.setup == 4:
args.net = LGCN
dataset_name = args.dataset
noise_level = args.noise
if dataset_name in ['douban']:
a1, f1, a2, f2, ground_truth, prior = load(dataset_name,
noise_level=noise_level)
feature_size = f1.shape[1]
ns = [a1.shape[0], a2.shape[0]]
edge_1 = torch.LongTensor(np.array(a1.nonzero()))
edge_2 = torch.LongTensor(np.array(a2.nonzero()))
ground_truth = torch.tensor(np.array(
ground_truth, dtype=int)) - 1 # Original index start from 1
features = [
torch.FloatTensor(f1.todense()),
torch.FloatTensor(f2.todense())
]
edges = [edge_1, edge_2]
prior = torch.FloatTensor(prior)
prior_rate = args.prior_rate
elif dataset_name in ['ppi', 'arena']:
a1, f1, a2, f2, ground_truth = load(dataset_name, noise_level=noise_level)
if args.pospath:
modelname += '_pos'
pre_textname += '_pos'
pre_embedname += '_pos'
if args.embed_size != 100:
modelname += f'_v{args.embed_size}'
pre_textname += f'_v{args.embed_size}'
pre_embedname += f'_v{args.embed_size}'
################################
# train or evaluation
################################
if args.train:
print('[{0:.15s}] Train'.format('STATE'))
# 1. load data
data, _, infos = load(reviewpath, productpath, infopath)
# 2. preprocessing train set
text = GP.fit(data.content.tolist(),
wordfix_path=args.wordpath,
posfix_path=args.pospath)
# save preprocessed text
with open(f'{args.savedir}/{pre_textname}.pickle', 'wb') as f:
pickle.dump(text, f)
# 2.1 product description
description = infos.description.str.replace('\n', ' ').tolist()
description = list(map(GP.stopword, description))
description = list(map(GP.kkma.nouns, description))
def main():
filename = sys.argv[1]
data = loaddata.load(filename)
print(filename)
print(data.mean(axis=1))