dataset.py 文件源码

def load_data(self, file_name):
        with open(file_name, 'rb') as file:
            unpickler = pickle._Unpickler(file)
            unpickler.encoding = 'latin1'
            contents = unpickler.load()
            X, Y = np.asarray(contents['data'], dtype=np.float32), np.asarray(contents['labels'])
            one_hot = np.zeros((Y.size, Y.max() + 1))
            one_hot[np.arange(Y.size), Y] = 1
            return X, one_hot