learning_rate = 0.1
# 훈련 스텝
step_num = 2000
'''
네트워크 옵션
'''
image_size = 28
class_num = 10
'''
데이터 로드 (훈련용 / 평가용)
'''
# 데이터 로드
path = '../data/MNIST/'
(train_label,
train_img) = mnist_data.read_data(path + 'train-labels-idx1-ubyte.gz',
path + 'train-images-idx3-ubyte.gz')
(val_label, val_img) = mnist_data.read_data(path + 't10k-labels-idx1-ubyte.gz',
path + 't10k-images-idx3-ubyte.gz')
# 10개의 이미지와 레이블 plotting
for i in range(10):
plt.subplot(1, 10, i + 1)
plt.imshow(train_img[i], cmap='Greys_r')
plt.axis('off')
plt.show()
print('label: %s' % (train_label[0:10]))
# 타입 변환
train_label = train_label.astype(np.int64)
train_img = train_img.astype(np.float32) / 255
val_label = val_label.astype(np.int64)
import numpy as np
import os
import urllib.request
import gzip
import struct
import tqdm
from collections import namedtuple
import mxnet as mx
from mnist_data import read_data
path='http://yann.lecun.com/exdb/mnist/'
(train_lbl, train_img) = read_data(path+'train-labels-idx1-ubyte.gz', path+'train-images-idx3-ubyte.gz')
(val_lbl, val_img) = read_data(path+'t10k-labels-idx1-ubyte.gz', path+'t10k-images-idx3-ubyte.gz')
def create_network():
data = mx.sym.Variable('data')
h = mx.sym.Convolution(data, kernel=(3, 3), pad=(1, 1), num_filter=32, name = "conv1")
h = mx.sym.Activation(h, name='relu1', act_type="relu")
h = mx.sym.Convolution(data, kernel=(3, 3), pad=(1, 1), num_filter=128, name = "conv2")
h = mx.sym.Activation(h, name='relu2', act_type="relu")
h = mx.sym.Convolution(data, kernel=(3, 3), pad=(1, 1), num_filter=256, name = "conv3")
h = mx.sym.Activation(h, name='relu3', act_type="relu")
h = mx.sym.Flatten(h)
h = mx.sym.FullyConnected(h, name='fc3', num_hidden = 512)
h = mx.sym.Activation(h, name='relu4', act_type="relu")
h = mx.sym.FullyConnected(h, name='fc4', num_hidden=10)
return mx.sym.softmax(h)
mlp = create_network()
mod = mx.mod.Module(symbol=mlp, context=mx.cpu(), label_names=None)
mod.bind(data_shapes=[('data', (1, 1, 28, 28))], for_training=False)