Пример #1
0
def ret80000():
	path = '/media/haboric/Ubuntu Data/yizhou/arctic-captions/json/real'

	sp = []
	#for i in range(1,9):
	for i in range(1,8):
		with open(path+'/train'+str(i*10000)+'.nd', 'rb') as f:
			sp.append(csr_matrix((lo(f),lo(f),lo(f))))

	#with open(path+'/train82783.nd', 'rb') as f:
	#	sp.append(csr_matrix((lo(f),lo(f),lo(f))))

	print 'before reduce'

	sv = reduce(conc, sp)
	del sp
	return sv
Пример #2
0
def ret80000():
    path = '/media/haboric/Ubuntu Data/yizhou/arctic-captions/json/real'

    sp = []
    #for i in range(1,9):
    for i in range(1, 8):
        with open(path + '/train' + str(i * 10000) + '.nd', 'rb') as f:
            sp.append(csr_matrix((lo(f), lo(f), lo(f))))

    #with open(path+'/train82783.nd', 'rb') as f:
    #	sp.append(csr_matrix((lo(f),lo(f),lo(f))))

    print 'before reduce'

    sv = reduce(conc, sp)
    del sp
    return sv
Пример #3
0

def conc(matrix1, matrix2):
    new_data = numpy.concatenate((matrix1.data, matrix2.data))
    new_indices = numpy.concatenate((matrix1.indices, matrix2.indices))
    new_ind_ptr = matrix2.indptr + len(matrix1.data)
    new_ind_ptr = new_ind_ptr[1:]
    new_ind_ptr = numpy.concatenate((matrix1.indptr, new_ind_ptr))
    return csr_matrix((new_data, new_indices, new_ind_ptr))


path = 'real'

sp = []
for i in range(1, 9):
    with open(path + '/train' + str(i * 10000) + '.nd', 'rb') as f:
        sp.append(csr_matrix((lo(f), lo(f), lo(f))))

with open(path + '/train82783.nd', 'rb') as f:
    sp.append(csr_matrix((lo(f), lo(f), lo(f))))

#print 'before reduce'

sv = reduce(conc, sp)
del sp

#print 'before dump'

with open(path + '/coco_align.train.feat.nd', 'wb') as f:
    pkl.dump(sv, f, protocol=pkl.HIGHEST_PROTOCOL)
Пример #4
0

def conc(matrix1, matrix2):
    new_data = numpy.concatenate((matrix1.data, matrix2.data))
    new_indices = numpy.concatenate((matrix1.indices, matrix2.indices))
    new_ind_ptr = matrix2.indptr + len(matrix1.data)
    new_ind_ptr = new_ind_ptr[1:]
    new_ind_ptr = numpy.concatenate((matrix1.indptr, new_ind_ptr))
    return csr_matrix((new_data, new_indices, new_ind_ptr))

path = 'real'

sp = []
for i in range(1,9):
	with open(path+'/train'+str(i*10000)+'.nd', 'rb') as f:
		sp.append(csr_matrix((lo(f),lo(f),lo(f))))

with open(path+'/train82783.nd', 'rb') as f:
	sp.append(csr_matrix((lo(f),lo(f),lo(f))))

#print 'before reduce'

sv = reduce(conc, sp)
del sp

#print 'before dump'

with open(path+'/coco_align.train.feat.nd', 'wb') as f:
	pkl.dump(sv,f,protocol=pkl.HIGHEST_PROTOCOL)