def __init__(self):
self.u = Util()
gpu.board_id_to_use = 1
print 'USE GPU' + str(gpu.board_id_to_use)
gpu.expensive_check_probability = 0
b = batch_creator()
path_train = '/home/tim/development/train.csv'
path_test = '/home/tim/development/test_X.csv'
batch_size = 100
set_sizes = [1.00,0.00,0.00]
data = b.create_batches([path_train, path_test], [0, -1], set_sizes,batch_size, standardize = False)
self.data = gpu.garray(data[0][0]/255.)
self.v_original = None
#self.w = gpu.garray(self.u.create_sparse_weight(784, 800))
self.w = gpu.garray(np.random.randn(784,800))*0.1
self.bias_h = gpu.zeros((1,800))
self.bias_v = gpu.zeros((1,784))
self.w_updt = gpu.zeros((784, 800))
self.bias_h_updt = gpu.zeros((1,800))
self.bias_v_updt = gpu.zeros((1,784))
self.h = gpu.zeros((100,800))
self.v = gpu.zeros((100,784))
self.time_interval = 0
import gnumpy as gpu
import numpy as np
from util import Util
'''Logistic regression to benchmark different momentum procedures.
'''
gpu.board_id_to_use = 0
print 'USE GPU' + str(gpu.board_id_to_use)
gpu.expensive_check_probability = 0
rng = np.random.RandomState(1234)
rng = np.random.RandomState(1234)
u = Util()
b = batch_creator()
path_train = '/home/tim/development/train.csv'
result_dir = '/home/tim/development/results/'
batch_size = 128
set_sizes = [0.80,0.20]
data = b.create_batches([path_train], [0], set_sizes,batch_size, standardize = False)
X_test = np.float32(np.load('/home/tim/development/test_X.npy'))/255.
X = gpu.garray(data[0][0])/255.
t = gpu.garray(data[0][2])
y = gpu.garray(data[0][1])
sys.path.append('/home/tim/cudamat/')
import gnumpy as gpu
import numpy as np
from util import Util
'''Logistic regression to benchmark different momentum procedures.
'''
gpu.board_id_to_use = 0
print 'USE GPU' + str(gpu.board_id_to_use)
gpu.expensive_check_probability = 0
rng = np.random.RandomState(1234)
rng = np.random.RandomState(1234)
u = Util()
b = batch_creator()
path_train = '/home/tim/development/train.csv'
result_dir = '/home/tim/development/results/'
batch_size = 128
set_sizes = [0.80, 0.20]
data = b.create_batches([path_train], [0],
set_sizes,
batch_size,
standardize=False)
X_test = np.float32(np.load('/home/tim/development/test_X.npy')) / 255.
X = gpu.garray(data[0][0]) / 255.
t = gpu.garray(data[0][2])
y = gpu.garray(data[0][1])
