def __init__(self, para=Parameters.gtnc(), debug_mode=False, device='cpu'):
# Initialize Parameters
Programclass.Program.__init__(self, device=device, dtype=para['dtype'])
MLclass.MachineLearning.__init__(self, para, debug_mode)
self.debug_mode = debug_mode
self.initialize_parameters_gtnc()
self.name_md5_generate()
self.inner_product = dict()
self.data_mapped = dict()
self.right_label = dict()
self.accuracy = dict()
self.test_info = dict()
self.is_all_gtn_trained = False
if not self.debug_mode:
self.load_accuracy()
# -*- encoding: utf-8 -*-
from library import MPSMLclass
from library import Parameters
import matplotlib.pyplot as plt
import numpy as np
import json
import copy
para = Parameters.gtnc()
# get the relation between cutting dimention and testing accuracy
batch_bond = []
batch_acc = []
saved_para = dict()
for i in range(1, 9):
bond = 2**i # get the cutting bond
para['virtual_bond_limitation'] = bond
print('The current cutting bond is %d' % bond)
A = MPSMLclass.GTNC(para=para,
device='cpu') # change device='cuda' to use GPU
A.training_gtn() # if the GTN are not trained
acc = A.calculate_accuracy('test')
print('Bond {0} acc is {1}'.format(bond, acc))
batch_bond.append(bond)
# para[''] = acc
batch_acc.append(acc)
counterpart_para = copy.deepcopy(para)
for item in counterpart_para.keys():
counterpart_para[item] = str(counterpart_para[item])
counterpart_para['Accuracy'] = acc
saved_para["bond" + str(bond)] = counterpart_para
