def next_dose_old(self, batch_size = 50) :
def convert_to_one_hot( label ) :
label_one_hot = np.zeros((len(label),2))
for i in range(len(label)) :
label_one_hot[i,label[i]] = 1
return label_one_hot
start = self._index_in_datafile
if ( self._file_index == self.start_file_index ) and ( start == 0 ) :
self.batch_size = batch_size
while np.modf(float(self.nrows)/self.batch_size)[0] > 0.0 :
print 'Warning! Number of data per file/ dose size must be an integer.'
print 'number of data per file: %d' % self.nrows
print 'dose size: %d' % self.batch_size
self.batch_size = int(input('Input new dose size: '))
print 'dose size : %d' % self.batch_size
print 'number of data: %d' % self._ndata
self.shuffle_index_dose_old = np.arange(0,self.batch_size,1)
self._index_in_datafile += self.batch_size
if self._index_in_datafile > self.nrows :
self._file_index += 1
start = 0
self._index_in_datafile = self.batch_size
assert self.batch_size <= self.nrows
if self._file_index > self.end_file_index :
# Number of training epochs completed
self._epochs_completed += 1
self._file_index = self.start_file_index
# Reinitialize conunter
start = 0
self._index_in_datafile = self.batch_size
end = self._index_in_datafile
# Read in small dosage of data
data = np.genfromtxt(self.full_file_path%(self._file_index) ,dtype=int,
skip_header=start, skip_footer=self.nrows-end)
self._images = data[:,:-1].astype('int')
labels = data[:,-1:].astype('int')
if self.convert_to_one_hot :
self._labels = convert_to_one_hot(labels)
# Shufle data
random.shuffle(self.shuffle_index_dose_old)
self._images = self._images[self.shuffle_index_dose_old]
self._labels = self._labels[self.shuffle_index_dose_old]
return self._images, self._labels
评论列表
文章目录
