tester.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import pandas as pd
import random
import h5py

import torch.optim as optim
from torch.utils.data import DataLoader, Dataset
import numpy as np
import editdistance as ed
from siacnn_models_gpu import *

#setting d1, d2
d1 = 1
d2 = 2

dataset2 = data_reader1(15, d1, d2, '~/seq-n20-ED15-2.txt', 0, 50000)

df2 = pd.DataFrame(data=dataset2)

test_a, test_b, test_t, test_y = aby_sep(df2)

test_t = torch.tensor(test_t)
test_y = torch.tensor(test_y)

#setting of k, m, batchsize
num_b = 20
m_dim = 40
batch_size = 20000
th_x = 0.5
##################### Model uploading #################

cnnk = torch.load('~/trained_model/cnn2_'+str(num_b)+'k_'+str(m_dim)+'m_('+str(d1)+'-'+str(d2)+')s_10xs.pt')
siacnn = torch.load('~/trained_model/siacnn2_'+str(num_b)+'k_'+str(m_dim)+'m_('+str(d1)+'-'+str(d2)+')s_10xs.pt')

#######################################TESTING############################################################
print('Results n = 20:')

acc_t = acc_test_batch(test_a, test_b, test_y, batch_size, siacnn, th_x, m_dim, num_b, device)

print('acc_'+str(d1)+'-'+str(d2)+'eds_'+str(num_b)+'k_'+str(m_dim)+'m_bl')

print('acc_test:', acc_t)

print('###################BreakDown Acc###############################')


####
print('bd_acc_('+str(d1)+','+str(d2)+')-eds_'+str(num_b)+'k_'+str(m_dim)+'m'+'_test')
eds = ed_sp(df2)
print('edits number')
ed_num_test = []
for i in sorted(eds.keys()):
    print('ed = ', i, ': ', len(eds[i]))
    ed_num_test.append([i, len(eds[i])])

res = breakdown_acc(eds, d1, d2, siacnn, batch_size, th_x, m_dim, num_b, device)

#hash code generating

hash_a, hash_b = matrixes_maker(test_a, test_b, batch_size)

f = h5py.File('hashcode_'+str(num_b)+'k_'+str(m_dim)+'m_('+str(d1)+','+str(d2)+')s.hdf5', 'w')

f.create_dataset('hash_a', data=np.array(hash_a.to('cpu')))
f.create_dataset('hash_b', data=np.array(hash_b.to('cpu')))

f.close()