os.system("ls -l N%dx%dx%d_L%d_U%d_Mu%s_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu%s_T//g -e s/.HSF.stream//g > dtau2.dat" %(n_x,n_x,n_x,L,U2,Mu,n_x,n_x,n_x,L,U2,Mu))
dtau2 = np.genfromtxt("dtau2.dat")
dtau = np.hstack((dtau1,dtau2))
os.remove("dtau1.dat")
os.remove("dtau2.dat")
# Array of shuffled file's file number
filenumber = np.arange(1+File_index_offset,len(dtau)+1,1)
if len(filenumber) > Max_nfile :
filenumber = filenumber[:Max_nfile]
#if not(use_single_U) and np.mod(len(filenumber),2) == 1:
# print 'Attention! When using data set with two Us for training, make sure that there are EVEN number of files. Exiting...'
# sys.exit()
# Provide file information to the data_reader module.
HSF = data_reader.insert_file_info(filename,filenumber, load_test_data_only=load_test_data_only)
# Load and catogorize data into either training data, test data, validation data, or
# all of them. If validation data is needed, set include_validation_data to (T)
# in the insert_file_info() module above.
HSF = HSF.categorize_data(make_spin_down_negative=True)
elif not(perform_classification_with_label) :
# Get temperature and save them to a file.
os.system("ls -l N%dx%dx%d_L%d_U%d_Mu%s_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu%s_T//g -e s/.HSF.stream//g > dtau.dat" %(n_x,n_x,n_x,L,U,Mu,n_x,n_x,n_x,L,U,Mu))
# Load temperature into a list of string
dtau = np.genfromtxt("dtau.dat",dtype='str')
os.remove("dtau.dat")
# The number of lines to skip at the beginning of the file if not all of the data is
# to be loaded.
sh = SKIPHEADER_tmp #ndata_per_temp - classification_data_per_temp
print 'Process: classification with label.'
else :
print 'Process: classification without label.'
print 'Using %d data per temperature.' % classification_data_per_temp
# xxxxx Don't change the following variable. xxxxx
# When perform_classification is set to True, only test data will be loaded.
perform_classification = T
if perform_classification_with_label == True :
# Get temperature and save them to a file.
os.system("ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau.dat" %(n_x,n_x,n_x,L,U,n_x,n_x,n_x,L,U))
dtau = np.genfromtxt("dtau.dat")
# Array of shuffled file's file number
filenumber = np.arange(1,len(dtau)+1,1)
# Provide file information to the data_reader module.
HSF = data_reader.insert_file_info(filename,filenumber, performing_classification=perform_classification)
# Load and catogorize data into either training data, test data, validation data, or
# all of them. If validation data is needed, set include_validation_data to (T)
# in the insert_file_info() module above.
HSF = HSF.categorize_data()
else :
# Get temperature and save them to a file.
os.system("ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau.dat" %(n_x,n_x,n_x,L,U,n_x,n_x,n_x,L,U))
# Load temperature into a list of string
dtau = np.genfromtxt("dtau.dat",dtype='str')
# The number of lines to skip at the beginning of the file if not all of the data is
# to be loaded.
sh = ndata_per_temp - classification_data_per_temp
while ( ndata_per_temp - sh - classification_data_per_temp ) < 0 :
print 'Sum of classification data per temperature and the number of lines skip at the beginning of the file must be equal to number of data per temnperature.'
print 'Process: classification without label.'
print 'Using %d data per temperature.' % classification_data_per_temp
# xxxxx Don't change the following variable. xxxxx
# When perform_classification is set to True, only test data will be loaded.
perform_classification = T
if perform_classification_with_label == True:
# Get temperature and save them to a file.
os.system(
"ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau.dat"
% (n_x, n_x, n_x, L, U, n_x, n_x, n_x, L, U))
dtau = np.genfromtxt("dtau.dat")
# Array of shuffled file's file number
filenumber = np.arange(1, len(dtau) + 1, 1)
# Provide file information to the data_reader module.
HSF = data_reader.insert_file_info(
filename, filenumber, performing_classification=perform_classification)
# Load and catogorize data into either training data, test data, validation data, or
# all of them. If validation data is needed, set include_validation_data to (T)
# in the insert_file_info() module above.
HSF = HSF.categorize_data()
else:
# Get temperature and save them to a file.
os.system(
"ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau.dat"
% (n_x, n_x, n_x, L, U, n_x, n_x, n_x, L, U))
# Load temperature into a list of string
dtau = np.genfromtxt("dtau.dat", dtype='str')
# The number of lines to skip at the beginning of the file if not all of the data is
# to be loaded.
sh = ndata_per_temp - classification_data_per_temp
def __init__(self, use_single_U = True, U = 5, U1 = 5, U2 = 20, name_output_file_by_date_first = True):
self.U = U
self.U1 = U1
self.U2 = U2
self.use_single_U = use_single_U
self.name_output_file_by_date_first = name_output_file_by_date_first
if use_single_U :
print 'Training using U = %d' % U
else :
print 'Training using U = %d and U = %d.' % (U1,U2)
# System size
# number of spin in each of the cube dimension
self.n_x = 4
n_x = self.n_x
# number of imaginary time dimension
self.L = 200
L = self.L
# Volume of tesseract
self.V4d = L*(n_x)**3
# Maximum number of data file to be used for training and testing.
Max_nfile = 100
# Offset to the file index (to load)
File_index_offset = 0
if use_single_U :
# Input labelled and shuffled filename for training and performaing classification
# with labels.
filename = './N%dx%dx%d_L%d_U%d_Mu0_T_shuffled' % (n_x,n_x,n_x,L,U) + '_%.2d.dat'
# Input raw filename for performing classification without labels.
rawdata_filename = './N%dx%dx%d_L%d_U%d_Mu0_T' % (n_x,n_x,n_x,L,U) + '%s.HSF.stream'
else :
# Input labelled and shuffled filename for training and performaing classification
# with labels.
if U1 < U2 :
filename = './N%dx%dx%d_L%d_U%d+U%d_Mu0_T_shuffled' % (n_x,n_x,n_x,L,U1,U2) + '_%.2d.dat'
else :
filename = './N%dx%dx%d_L%d_U%d+U%d_Mu0_T_shuffled' % (n_x,n_x,n_x,L,U2,U1) + '_%.2d.dat'
# Load data
if use_single_U :
# Get temperature and save them to a file.
os.system("ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau.dat" %(n_x,n_x,n_x,L,U,n_x,n_x,n_x,L,U))
self.dtau = np.genfromtxt("dtau.dat")
os.remove("dtau.dat")
# Array of shuffled file's file number
else :
# Get temperature and save them to a file.
os.system("ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau1.dat" %(n_x,n_x,n_x,L,U1,n_x,n_x,n_x,L,U1))
dtau1 = np.genfromtxt("dtau1.dat")
# Get temperature and save them to a file.
os.system("ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau2.dat" %(n_x,n_x,n_x,L,U2,n_x,n_x,n_x,L,U2))
dtau2 = np.genfromtxt("dtau2.dat")
self.dtau = np.hstack((dtau1,dtau2))
os.remove("dtau1.dat")
os.remove("dtau2.dat")
# Array of shuffled file's file number
filenumber = np.arange(1+File_index_offset,len(self.dtau)+1,1)
if len(filenumber) > Max_nfile :
filenumber = filenumber[:Max_nfile]
# Provide file information to the data_reader module.
HSF = data_reader.insert_file_info(filename,filenumber, load_test_data_only=False)
# Load and catogorize data into either training data, test data, validation data, or
# all of them. If validation data is needed, set include_validation_data to (T)
# in the insert_file_info() module above.
self.HSF = HSF.categorize_data()
import sys import numpy as np import data_reader import time filename = '/home/kelvin/Desktop/Theano test/HSF_N4x4x4_L200_U9_Mu0_UniformTGrid/N4x4x4_L200_U9_Mu0_T_shuffled_%.2d.HSF.stream' filename_weight_bias = "./wb.ckpt" filename_measure = "./HSF_measure.dat" filenumber = np.arange(1, 41, 1) HSF = data_reader.insert_file_info(filename, filenumber) HSF = HSF.categorize_data() i_training = 6400 n_feature_map1 = 32 n_feature_map2 = 32 n_feature_map3 = 8 n_feature_map4 = 8 n_spin = 4 n_time_dimension = 200 n_output_neuron = 2 n_fully_connected_neuron = 30 filter_d = 2 filter_h = filter_d filter_w = filter_d import tensorflow as tf # If you are not using an InteractiveSession, then you should build # the entire computation graph before starting a session and # launching the graph. sess = tf.InteractiveSession()
dtau2 = np.genfromtxt("dtau2.dat")
dtau = np.hstack((dtau1, dtau2))
os.remove("dtau1.dat")
os.remove("dtau2.dat")
# Array of shuffled file's file number
filenumber = np.arange(1 + File_index_offset, len(dtau) + 1, 1)
if len(filenumber) > Max_nfile:
filenumber = filenumber[:Max_nfile]
#if not(use_single_U) and np.mod(len(filenumber),2) == 1:
# print 'Attention! When using data set with two Us for training, make sure that there are EVEN number of files. Exiting...'
# sys.exit()
# Provide file information to the data_reader module.
HSF = data_reader.insert_file_info(filename,
filenumber,
load_test_data_only=load_test_data_only)
# Load and catogorize data into either training data, test data, validation data, or
# all of them. If validation data is needed, set include_validation_data to (T)
# in the insert_file_info() module above.
HSF = HSF.categorize_data(make_spin_down_negative=True)
elif not (perform_classification_with_label):
# Get temperature and save them to a file.
os.system(
"ls -l N%dx%dx%d_L%d_U%d_Mu%s_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu%s_T//g -e s/.HSF.stream//g > dtau.dat"
% (n_x, n_x, n_x, L, U, Mu, n_x, n_x, n_x, L, U, Mu))
# Load temperature into a list of string
dtau = np.genfromtxt("dtau.dat", dtype='str')
os.remove("dtau.dat")
# The number of lines to skip at the beginning of the file if not all of the data is
import sys import numpy as np import data_reader import time filename = '/home/kelvin/Desktop/Theano test/HSF_N4x4x4_L200_U9_Mu0_UniformTGrid/N4x4x4_L200_U9_Mu0_T_shuffled_%.2d.HSF.stream' filename_weight_bias = "./wb.ckpt" filename_measure = "./HSF_measure.dat" filenumber = np.arange(1,41,1) HSF = data_reader.insert_file_info( filename, filenumber ) HSF = HSF.categorize_data() i_training = 6400 n_feature_map1 = 32 n_feature_map2 = 32 n_feature_map3 = 8 n_feature_map4 = 8 n_spin = 4 n_time_dimension = 200 n_output_neuron = 2 n_fully_connected_neuron = 30 filter_d = 2 filter_h = filter_d filter_w = filter_d import tensorflow as tf # If you are not using an InteractiveSession, then you should build # the entire computation graph before starting a session and # launching the graph. sess = tf.InteractiveSession()
"If you set your logs directory manually make sure"
"to use /logs/ when running on ClusterOne cloud.")
flags.DEFINE_string(
"log_dir", get_logs_path(root=PATH_TO_LOCAL_LOGS),
"Path to dataset. It is recommended to use get_data_path()"
"to define your data directory.so that you can switch "
"from local to clusterone without changing your code."
"If you set the data directory manually makue sure to use"
"/data/ as root path when running on ClusterOne cloud.")
FLAGS = flags.FLAGS
n_x = 10
Ising = data_reader.insert_file_info(FLAGS.data_dir + "2D%d" % (n_x) +
"_p_%.1d.txt",
np.arange(1, 3),
load_test_data_only=False)
Ising = Ising.categorize_data()
n_output_neuron = 2
filter_d = 2
filter_h = 2
n_feature_map1 = 16
n_feature_map2 = 8
n_fully_connected_neuron = 16
batch_size = 50
n_train_data = 50000
def __init__(self,
use_single_U=True,
U=5,
U1=5,
U2=20,
name_output_file_by_date_first=True):
self.U = U
self.U1 = U1
self.U2 = U2
self.use_single_U = use_single_U
self.name_output_file_by_date_first = name_output_file_by_date_first
if use_single_U:
print 'Training using U = %d' % U
else:
print 'Training using U = %d and U = %d.' % (U1, U2)
# System size
# number of spin in each of the cube dimension
self.n_x = 4
n_x = self.n_x
# number of imaginary time dimension
self.L = 200
L = self.L
# Volume of tesseract
self.V4d = L * (n_x)**3
# Maximum number of data file to be used for training and testing.
Max_nfile = 100
# Offset to the file index (to load)
File_index_offset = 0
if use_single_U:
# Input labelled and shuffled filename for training and performaing classification
# with labels.
filename = './N%dx%dx%d_L%d_U%d_Mu0_T_shuffled' % (
n_x, n_x, n_x, L, U) + '_%.2d.dat'
# Input raw filename for performing classification without labels.
rawdata_filename = './N%dx%dx%d_L%d_U%d_Mu0_T' % (
n_x, n_x, n_x, L, U) + '%s.HSF.stream'
else:
# Input labelled and shuffled filename for training and performaing classification
# with labels.
if U1 < U2:
filename = './N%dx%dx%d_L%d_U%d+U%d_Mu0_T_shuffled' % (
n_x, n_x, n_x, L, U1, U2) + '_%.2d.dat'
else:
filename = './N%dx%dx%d_L%d_U%d+U%d_Mu0_T_shuffled' % (
n_x, n_x, n_x, L, U2, U1) + '_%.2d.dat'
# Load data
if use_single_U:
# Get temperature and save them to a file.
os.system(
"ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau.dat"
% (n_x, n_x, n_x, L, U, n_x, n_x, n_x, L, U))
self.dtau = np.genfromtxt("dtau.dat")
os.remove("dtau.dat")
# Array of shuffled file's file number
else:
# Get temperature and save them to a file.
os.system(
"ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau1.dat"
% (n_x, n_x, n_x, L, U1, n_x, n_x, n_x, L, U1))
dtau1 = np.genfromtxt("dtau1.dat")
# Get temperature and save them to a file.
os.system(
"ls -l N%dx%dx%d_L%d_U%d_Mu0_T*.HSF.stream | awk '{print $9}' | sed -e s/N%dx%dx%d_L%d_U%d_Mu0_T//g -e s/.HSF.stream//g > dtau2.dat"
% (n_x, n_x, n_x, L, U2, n_x, n_x, n_x, L, U2))
dtau2 = np.genfromtxt("dtau2.dat")
self.dtau = np.hstack((dtau1, dtau2))
os.remove("dtau1.dat")
os.remove("dtau2.dat")
# Array of shuffled file's file number
filenumber = np.arange(1 + File_index_offset, len(self.dtau) + 1, 1)
if len(filenumber) > Max_nfile:
filenumber = filenumber[:Max_nfile]
# Provide file information to the data_reader module.
HSF = data_reader.insert_file_info(filename,
filenumber,
load_test_data_only=False)
# Load and catogorize data into either training data, test data, validation data, or
# all of them. If validation data is needed, set include_validation_data to (T)
# in the insert_file_info() module above.
self.HSF = HSF.categorize_data()