图像分类

在此项目中，你将对 CIFAR-10 数据集中的图片进行分类。该数据集包含飞机、猫狗和其他物体。你需要预处理这些图片，然后用所有样本训练一个卷积神经网络。图片需要标准化（normalized），标签需要采用 one-hot 编码。你需要应用所学的知识构建卷积的、最大池化（max pooling）、丢弃（dropout）和完全连接（fully connected）的层。最后，你需要在样本图片上看到神经网络的预测结果。

获取数据

请运行以下单元，以下载 CIFAR-10 数据集（Python版）。



In [1]:

    
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
from urllib.request import urlretrieve
from os.path import isfile, isdir
from tqdm import tqdm
import problem_unittests as tests
import tarfile

cifar10_dataset_folder_path = 'cifar-10-batches-py'

# Use Floyd's cifar-10 dataset if present
floyd_cifar10_location = '/input/cifar-10/python.tar.gz'
if isfile(floyd_cifar10_location):
    tar_gz_path = floyd_cifar10_location
else:
    tar_gz_path = 'cifar-10-python.tar.gz'

class DLProgress(tqdm):
    last_block = 0

    def hook(self, block_num=1, block_size=1, total_size=None):
        self.total = total_size
        self.update((block_num - self.last_block) * block_size)
        self.last_block = block_num

if not isfile(tar_gz_path):
    with DLProgress(unit='B', unit_scale=True, miniters=1, desc='CIFAR-10 Dataset') as pbar:
        urlretrieve(
            'https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz',
            tar_gz_path,
            pbar.hook)

if not isdir(cifar10_dataset_folder_path):
    with tarfile.open(tar_gz_path) as tar:
        tar.extractall()
        tar.close()


tests.test_folder_path(cifar10_dataset_folder_path)









    



All files found!

探索数据

该数据集分成了几部分／批次（batches），以免你的机器在计算时内存不足。CIFAR-10 数据集包含 5 个部分，名称分别为 data_batch_1、data_batch_2，以此类推。每个部分都包含以下某个类别的标签和图片：

飞机
汽车
鸟类
猫
鹿
狗
青蛙
马
船只
卡车

了解数据集也是对数据进行预测的必经步骤。你可以通过更改 batch_id 和 sample_id 探索下面的代码单元。batch_id 是数据集一个部分的 ID（1 到 5）。sample_id 是该部分中图片和标签对（label pair）的 ID。

问问你自己：“可能的标签有哪些？”、“图片数据的值范围是多少？”、“标签是按顺序排列，还是随机排列的？”。思考类似的问题，有助于你预处理数据，并使预测结果更准确。



In [2]:

    
%matplotlib inline
%config InlineBackend.figure_format = 'retina'

import helper
import numpy as np

# Explore the dataset
batch_id = 1
sample_id = 50
helper.display_stats(cifar10_dataset_folder_path, batch_id, sample_id)









    



Stats of batch 1:
Samples: 10000
Label Counts: {0: 1005, 1: 974, 2: 1032, 3: 1016, 4: 999, 5: 937, 6: 1030, 7: 1001, 8: 1025, 9: 981}
First 20 Labels: [6, 9, 9, 4, 1, 1, 2, 7, 8, 3, 4, 7, 7, 2, 9, 9, 9, 3, 2, 6]

Example of Image 50:
Image - Min Value: 8 Max Value: 243
Image - Shape: (32, 32, 3)
Label - Label Id: 9 Name: truck

实现预处理函数

标准化

在下面的单元中，实现 normalize 函数，传入图片数据 x，并返回标准化 Numpy 数组。值应该在 0 到 1 的范围内（含 0 和 1）。返回对象应该和 x 的形状一样。



In [3]:

    
def normalize(x):
    """
    Normalize a list of sample image data in the range of 0 to 1
    : x: List of image data.  The image shape is (32, 32, 3)
    : return: Numpy array of normalize data
    """
    # TODO: Implement Function
    return (x / 255)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_normalize(normalize)









    



Tests Passed

One-hot 编码

和之前的代码单元一样，你将为预处理实现一个函数。这次，你将实现 one_hot_encode 函数。输入，也就是 x，是一个标签列表。实现该函数，以返回为 one_hot 编码的 Numpy 数组的标签列表。标签的可能值为 0 到 9。每次调用 one_hot_encode 时，对于每个值，one_hot 编码函数应该返回相同的编码。确保将编码映射保存到该函数外面。

提示：不要重复发明轮子。



In [4]:

    
import numpy as np
from sklearn import preprocessing

def one_hot_encode(x):
    """
    One hot encode a list of sample labels. Return a one-hot encoded vector for each label.
    : x: List of sample Labels
    : return: Numpy array of one-hot encoded labels
    """
    # TODO: Implement Function
    return np.eye(10)[x]


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_one_hot_encode(one_hot_encode)









    



Tests Passed

随机化数据

之前探索数据时，你已经了解到，样本的顺序是随机的。再随机化一次也不会有什么关系，但是对于这个数据集没有必要。

预处理所有数据并保存

运行下方的代码单元，将预处理所有 CIFAR-10 数据，并保存到文件中。下面的代码还使用了 10% 的训练数据，用来验证。



In [5]:

    
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
# Preprocess Training, Validation, and Testing Data
helper.preprocess_and_save_data(cifar10_dataset_folder_path, normalize, one_hot_encode)

检查点

这是你的第一个检查点。如果你什么时候决定再回到该记事本，或需要重新启动该记事本，你可以从这里开始。预处理的数据已保存到本地。



In [6]:

    
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import pickle
import problem_unittests as tests
import helper

# Load the Preprocessed Validation data
valid_features, valid_labels = pickle.load(open('preprocess_validation.p', mode='rb'))

构建网络

对于该神经网络，你需要将每层都构建为一个函数。你看到的大部分代码都位于函数外面。要更全面地测试你的代码，我们需要你将每层放入一个函数中。这样使我们能够提供更好的反馈，并使用我们的统一测试检测简单的错误，然后再提交项目。

注意：如果你觉得每周很难抽出足够的时间学习这门课程，我们为此项目提供了一个小捷径。对于接下来的几个问题，你可以使用 TensorFlow Layers 或 TensorFlow Layers (contrib) 程序包中的类来构建每个层级，但是“卷积和最大池化层级”部分的层级除外。TF Layers 和 Keras 及 TFLearn 层级类似，因此很容易学会。

但是，如果你想充分利用这门课程，请尝试自己解决所有问题，不使用 TF Layers 程序包中的任何类。你依然可以使用其他程序包中的类，这些类和你在 TF Layers 中的类名称是一样的！例如，你可以使用 TF Neural Network 版本的 conv2d 类 tf.nn.conv2d，而不是 TF Layers 版本的 conv2d 类 tf.layers.conv2d。

我们开始吧！

输入

神经网络需要读取图片数据、one-hot 编码标签和丢弃保留概率（dropout keep probability）。请实现以下函数：

实现 neural_net_image_input
- 返回 TF Placeholder
- 使用 image_shape 设置形状，部分大小设为 None
- 使用 TF Placeholder 中的 TensorFlow name 参数对 TensorFlow 占位符 "x" 命名
实现 neural_net_label_input
- 返回 TF Placeholder
- 使用 n_classes 设置形状，部分大小设为 None
- 使用 TF Placeholder 中的 TensorFlow name 参数对 TensorFlow 占位符 "y" 命名
实现 neural_net_keep_prob_input
- 返回 TF Placeholder，用于丢弃保留概率
- 使用 TF Placeholder 中的 TensorFlow name 参数对 TensorFlow 占位符 "keep_prob" 命名

这些名称将在项目结束时，用于加载保存的模型。

注意：TensorFlow 中的 None 表示形状可以是动态大小。



In [7]:

    
import tensorflow as tf

def neural_net_image_input(image_shape):
    """
    Return a Tensor for a batch of image input
    : image_shape: Shape of the images
    : return: Tensor for image input.
    """
    # TODO: Implement Function
    return tf.placeholder(tf.float32, shape = (None, *image_shape), name = "x")


def neural_net_label_input(n_classes):
    """
    Return a Tensor for a batch of label input
    : n_classes: Number of classes
    : return: Tensor for label input.
    """
    # TODO: Implement Function
    return tf.placeholder(tf.int8, shape = (None, n_classes), name = "y")


def neural_net_keep_prob_input():
    """
    Return a Tensor for keep probability
    : return: Tensor for keep probability.
    """
    # TODO: Implement Function
    return tf.placeholder(tf.float32, shape = None, name = "keep_prob")


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tf.reset_default_graph()
tests.test_nn_image_inputs(neural_net_image_input)
tests.test_nn_label_inputs(neural_net_label_input)
tests.test_nn_keep_prob_inputs(neural_net_keep_prob_input)









    



Image Input Tests Passed.
Label Input Tests Passed.
Keep Prob Tests Passed.

卷积和最大池化层

卷积层级适合处理图片。对于此代码单元，你应该实现函数 conv2d_maxpool 以便应用卷积然后进行最大池化：

使用 conv_ksize、conv_num_outputs 和 x_tensor 的形状创建权重（weight）和偏置（bias）。
使用权重和 conv_strides 对 x_tensor 应用卷积。
- 建议使用我们建议的间距（padding），当然也可以使用任何其他间距。
添加偏置
向卷积中添加非线性激活（nonlinear activation）
使用 pool_ksize 和 pool_strides 应用最大池化
- 建议使用我们建议的间距（padding），当然也可以使用任何其他间距。

注意：对于此层，请勿使用 TensorFlow Layers 或 TensorFlow Layers (contrib)，但是仍然可以使用 TensorFlow 的 Neural Network 包。对于所有其他层，你依然可以使用快捷方法。



In [8]:

    
def conv2d_maxpool(x_tensor, conv_num_outputs, conv_ksize, conv_strides, pool_ksize, pool_strides):
    """
    Apply convolution then max pooling to x_tensor
    :param x_tensor: TensorFlow Tensor
    :param conv_num_outputs: Number of outputs for the convolutional layer
    :param conv_ksize: kernal size 2-D Tuple for the convolutional layer
    :param conv_strides: Stride 2-D Tuple for convolution
    :param pool_ksize: kernal size 2-D Tuple for pool
    :param pool_strides: Stride 2-D Tuple for pool
    : return: A tensor that represents convolution and max pooling of x_tensor
    """
    # TODO: Implement Function
    input_chanel = int(x_tensor.shape[3])
    output_chanel = conv_num_outputs
    weight_shape = (*conv_ksize,input_chanel,output_chanel)  # * 
    weight = tf.Variable(tf.random_normal(weight_shape, stddev = 0.1))  #权重
    bias = tf.Variable(tf.zeros(output_chanel))  #设置偏置项
    l_active = tf.nn.conv2d(x_tensor, weight, (1, *conv_strides, 1), 'SAME')
    l_active = tf.nn.bias_add(l_active,bias)
    #active_layers = tf.nn.relu(tf.add(tf.matmul(features,label),bias))    #ReLu
    mx_layer = tf.nn.relu(l_active)
    
    return tf.nn.max_pool(mx_layer, (1, *pool_ksize, 1), (1, *pool_strides, 1), 'VALID')


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_con_pool(conv2d_maxpool)









    



Tests Passed

扁平化层

实现 flatten 函数，将 x_tensor 的维度从四维张量（4-D tensor）变成二维张量。输出应该是形状（部分大小（Batch Size），扁平化图片大小（Flattened Image Size））。快捷方法：对于此层，你可以使用 TensorFlow Layers 或 TensorFlow Layers (contrib) 包中的类。如果你想要更大挑战，可以仅使用其他 TensorFlow 程序包。



In [9]:

    
from functools import reduce
from operator import mul

def flatten(x_tensor):
    """
    Flatten x_tensor to (Batch Size, Flattened Image Size)
    : x_tensor: A tensor of size (Batch Size, ...), where ... are the image dimensions.
    : return: A tensor of size (Batch Size, Flattened Image Size).
    """
    # TODO: Implement Function
    _, *image_size = x_tensor.get_shape().as_list()
    #print(*image_size)
    return tf.reshape(x_tensor, (-1, reduce(mul, image_size)))


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_flatten(flatten)









    



Tests Passed

完全连接的层

实现 fully_conn 函数，以向 x_tensor 应用完全连接的层级，形状为（部分大小（Batch Size），num_outputs）。快捷方法：对于此层，你可以使用 TensorFlow Layers 或 TensorFlow Layers (contrib) 包中的类。如果你想要更大挑战，可以仅使用其他 TensorFlow 程序包。



In [10]:

    
def fully_conn(x_tensor, num_outputs):
    """
    Apply a fully connected layer to x_tensor using weight and bias
    : x_tensor: A 2-D tensor where the first dimension is batch size.
    : num_outputs: The number of output that the new tensor should be.
    : return: A 2-D tensor where the second dimension is num_outputs.
    """
    # TODO: Implement Function
    num_input = x_tensor.get_shape().as_list()[1]
    weight_shape = (num_input, num_outputs)
    #print(weight_shape)
    weight = tf.Variable(tf.truncated_normal(weight_shape, stddev = 0.1))
    
    bias = tf.Variable(tf.zeros(num_outputs))
     
    activation = tf.nn.bias_add(tf.matmul(x_tensor, weight), bias)
    return tf.nn.relu(activation)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_fully_conn(fully_conn)









    



Tests Passed

输出层

实现 output 函数，向 x_tensor 应用完全连接的层级，形状为（部分大小（Batch Size），num_outputs）。快捷方法：对于此层，你可以使用 TensorFlow Layers 或 TensorFlow Layers (contrib) 包中的类。如果你想要更大挑战，可以仅使用其他 TensorFlow 程序包。

注意：该层级不应应用 Activation、softmax 或交叉熵（cross entropy）。



In [11]:

    
def output(x_tensor, num_outputs):
    """
    Apply a output layer to x_tensor using weight and bias
    : x_tensor: A 2-D tensor where the first dimension is batch size.
    : num_outputs: The number of output that the new tensor should be.
    : return: A 2-D tensor where the second dimension is num_outputs.
    """
    # TODO: Implement Function
    num_input = x_tensor.get_shape().as_list()[1] #not 0
    weight_shape = (num_input, num_outputs) 
    weight = tf.Variable(tf.truncated_normal(weight_shape, stddev = 0.1))
    
    bias = tf.Variable(tf.zeros(num_outputs))
    
    return tf.nn.bias_add(tf.matmul(x_tensor,weight),bias)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_output(output)









    



Tests Passed

创建卷积模型

实现函数 conv_net，创建卷积神经网络模型。该函数传入一批图片 x，并输出对数（logits）。使用你在上方创建的层创建此模型：

应用 1、2 或 3 个卷积和最大池化层（Convolution and Max Pool layers）
应用一个扁平层（Flatten Layer）
应用 1、2 或 3 个完全连接层（Fully Connected Layers）
应用一个输出层（Output Layer）
返回输出
使用 keep_prob 向模型中的一个或多个层应用 TensorFlow 的 Dropout



In [12]:

    
def conv_net(x, keep_prob):
    """
    Create a convolutional neural network model
    : x: Placeholder tensor that holds image data.
    : keep_prob: Placeholder tensor that hold dropout keep probability.
    : return: Tensor that represents logits
    """
    # TODO: Apply 1, 2, or 3 Convolution and Max Pool layers
    #    Play around with different number of outputs, kernel size and stride
    # Function Definition from Above:
    #    conv2d_maxpool(x_tensor, conv_num_outputs, conv_ksize, conv_strides, pool_ksize, pool_strides)
    x = conv2d_maxpool(x, 64, (3, 3), (1, 1), (2, 2), (2, 2))
    x = tf.nn.dropout(x, keep_prob)
    
    x = conv2d_maxpool(x, 128, (3, 3), (1, 1), (2, 2), (2, 2))
    x = tf.nn.dropout(x, keep_prob)
    # x has shape (batch, 8, 8, 128)
    x = conv2d_maxpool(x, 256, (3, 3), (1, 1), (2, 2), (2, 2))
    x = tf.nn.dropout(x, keep_prob)
    
    # TODO: Apply a Flatten Layer
    # Function Definition from Above:
    #   flatten(x_tensor)
    x = flatten(x)

    # TODO: Apply 1, 2, or 3 Fully Connected Layers
    #    Play around with different number of outputs
    # Function Definition from Above:
    #   fully_conn(x_tensor, num_outputs)
    x = fully_conn(x, 512)
    x = tf.nn.dropout(x, keep_prob)
    
    # TODO: Apply an Output Layer
    #    Set this to the number of classes
    # Function Definition from Above:
    #   output(x_tensor, num_outputs)
    
    
    # TODO: return output
    return output(x, 10)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

##############################
## Build the Neural Network ##
##############################

# Remove previous weights, bias, inputs, etc..
tf.reset_default_graph()

# Inputs
x = neural_net_image_input((32, 32, 3))
y = neural_net_label_input(10)
keep_prob = neural_net_keep_prob_input()

# Model
logits = conv_net(x, keep_prob)

# Name logits Tensor, so that is can be loaded from disk after training
logits = tf.identity(logits, name='logits')

# Loss and Optimizer
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y))
optimizer = tf.train.AdamOptimizer().minimize(cost)

# Accuracy
correct_pred = tf.equal(tf.argmax(logits, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32), name='accuracy')

tests.test_conv_net(conv_net)









    



Neural Network Built!

训练神经网络

单次优化

实现函数 train_neural_network 以进行单次优化（single optimization）。该优化应该使用 optimizer 优化 session，其中 feed_dict 具有以下参数：

x 表示图片输入
y 表示标签
keep_prob 表示丢弃的保留率

每个部分都会调用该函数，所以 tf.global_variables_initializer() 已经被调用。

注意：不需要返回任何内容。该函数只是用来优化神经网络。



In [13]:

    
def train_neural_network(session, optimizer, keep_probability, feature_batch, label_batch):
    """
    Optimize the session on a batch of images and labels
    : session: Current TensorFlow session
    : optimizer: TensorFlow optimizer function
    : keep_probability: keep probability
    : feature_batch: Batch of Numpy image data
    : label_batch: Batch of Numpy label data
    """
    # TODO: Implement Function
    session.run(optimizer, feed_dict={x: feature_batch, y: label_batch, keep_prob: keep_probability})


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_train_nn(train_neural_network)









    



Tests Passed

显示数据

实现函数 print_stats 以输出损失和验证准确率。使用全局变量 valid_features 和 valid_labels 计算验证准确率。使用保留率 1.0 计算损失和验证准确率（loss and validation accuracy）。



In [14]:

    
def print_stats(session, feature_batch, label_batch, cost, accuracy):
    """
    Print information about loss and validation accuracy
    : session: Current TensorFlow session
    : feature_batch: Batch of Numpy image data
    : label_batch: Batch of Numpy label data
    : cost: TensorFlow cost function
    : accuracy: TensorFlow accuracy function
    """
    # TODO: Implement Function
    global valid_features, valid_labels
    validation_accuracy = session.run(accuracy, feed_dict={x: valid_features, y: valid_labels, keep_prob: 1.0})
    loss = session.run( cost, feed_dict={x: feature_batch, y: label_batch, keep_prob: 1.0})
    
    prt = 'Loss: {:.4f} Accuracy: {:.4f}'
    print(prt.format(loss, validation_accuracy, prec=3))

超参数

调试以下超参数：

设置 epochs 表示神经网络停止学习或开始过拟合的迭代次数
设置 batch_size，表示机器内存允许的部分最大体积。大部分人设为以下常见内存大小：
- 64
- 128
- 256
- ...
设置 keep_probability 表示使用丢弃时保留节点的概率



In [15]:

    
# TODO: Tune Parameters
epochs = 200
batch_size = 128
keep_probability = 0.5

在单个 CIFAR-10 部分上训练

我们先用单个部分，而不是用所有的 CIFAR-10 批次训练神经网络。这样可以节省时间，并对模型进行迭代，以提高准确率。最终验证准确率达到 50% 或以上之后，在下一部分对所有数据运行模型。



In [16]:

    
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
print('Checking the Training on a Single Batch...')
with tf.Session() as sess:
    # Initializing the variables
    sess.run(tf.global_variables_initializer())
    
    # Training cycle
    for epoch in range(epochs):
        batch_i = 1
        for batch_features, batch_labels in helper.load_preprocess_training_batch(batch_i, batch_size):
            train_neural_network(sess, optimizer, keep_probability, batch_features, batch_labels)
        print('Epoch {:>2}, CIFAR-10 Batch {}:  '.format(epoch + 1, batch_i), end='')
        print_stats(sess, batch_features, batch_labels, cost, accuracy)









    



Checking the Training on a Single Batch...
Epoch  1, CIFAR-10 Batch 1:  Loss: 2.2983 Accuracy: 0.1052
Epoch  2, CIFAR-10 Batch 1:  Loss: 2.3019 Accuracy: 0.1010
Epoch  3, CIFAR-10 Batch 1:  Loss: 2.2950 Accuracy: 0.1258
Epoch  4, CIFAR-10 Batch 1:  Loss: 2.2787 Accuracy: 0.1510
Epoch  5, CIFAR-10 Batch 1:  Loss: 2.2601 Accuracy: 0.1592
Epoch  6, CIFAR-10 Batch 1:  Loss: 2.2481 Accuracy: 0.1556
Epoch  7, CIFAR-10 Batch 1:  Loss: 2.2328 Accuracy: 0.1614
Epoch  8, CIFAR-10 Batch 1:  Loss: 2.1974 Accuracy: 0.1950
Epoch  9, CIFAR-10 Batch 1:  Loss: 2.1336 Accuracy: 0.2274
Epoch 10, CIFAR-10 Batch 1:  Loss: 2.0589 Accuracy: 0.2510
Epoch 11, CIFAR-10 Batch 1:  Loss: 1.9907 Accuracy: 0.3046
Epoch 12, CIFAR-10 Batch 1:  Loss: 2.0030 Accuracy: 0.3036
Epoch 13, CIFAR-10 Batch 1:  Loss: 1.9698 Accuracy: 0.3454
Epoch 14, CIFAR-10 Batch 1:  Loss: 1.9381 Accuracy: 0.3614
Epoch 15, CIFAR-10 Batch 1:  Loss: 1.9321 Accuracy: 0.3576
Epoch 16, CIFAR-10 Batch 1:  Loss: 1.9110 Accuracy: 0.3924
Epoch 17, CIFAR-10 Batch 1:  Loss: 1.8847 Accuracy: 0.3914
Epoch 18, CIFAR-10 Batch 1:  Loss: 1.8655 Accuracy: 0.4000
Epoch 19, CIFAR-10 Batch 1:  Loss: 1.7937 Accuracy: 0.4150
Epoch 20, CIFAR-10 Batch 1:  Loss: 1.8148 Accuracy: 0.4054
Epoch 21, CIFAR-10 Batch 1:  Loss: 1.8180 Accuracy: 0.4088
Epoch 22, CIFAR-10 Batch 1:  Loss: 1.7589 Accuracy: 0.4220
Epoch 23, CIFAR-10 Batch 1:  Loss: 1.7052 Accuracy: 0.4270
Epoch 24, CIFAR-10 Batch 1:  Loss: 1.6943 Accuracy: 0.4394
Epoch 25, CIFAR-10 Batch 1:  Loss: 1.7342 Accuracy: 0.4196
Epoch 26, CIFAR-10 Batch 1:  Loss: 1.6755 Accuracy: 0.4378
Epoch 27, CIFAR-10 Batch 1:  Loss: 1.7083 Accuracy: 0.4226
Epoch 28, CIFAR-10 Batch 1:  Loss: 1.6642 Accuracy: 0.4306
Epoch 29, CIFAR-10 Batch 1:  Loss: 1.6127 Accuracy: 0.4432
Epoch 30, CIFAR-10 Batch 1:  Loss: 1.6448 Accuracy: 0.4346
Epoch 31, CIFAR-10 Batch 1:  Loss: 1.7516 Accuracy: 0.4092
Epoch 32, CIFAR-10 Batch 1:  Loss: 1.8120 Accuracy: 0.3852
Epoch 33, CIFAR-10 Batch 1:  Loss: 1.6668 Accuracy: 0.4274
Epoch 34, CIFAR-10 Batch 1:  Loss: 1.6681 Accuracy: 0.4178
Epoch 35, CIFAR-10 Batch 1:  Loss: 1.4836 Accuracy: 0.4710
Epoch 36, CIFAR-10 Batch 1:  Loss: 1.5404 Accuracy: 0.4686
Epoch 37, CIFAR-10 Batch 1:  Loss: 1.5020 Accuracy: 0.4682
Epoch 38, CIFAR-10 Batch 1:  Loss: 1.6325 Accuracy: 0.4268
Epoch 39, CIFAR-10 Batch 1:  Loss: 1.5342 Accuracy: 0.4504
Epoch 40, CIFAR-10 Batch 1:  Loss: 1.4532 Accuracy: 0.4696
Epoch 41, CIFAR-10 Batch 1:  Loss: 1.4693 Accuracy: 0.4648
Epoch 42, CIFAR-10 Batch 1:  Loss: 1.4926 Accuracy: 0.4594
Epoch 43, CIFAR-10 Batch 1:  Loss: 1.4819 Accuracy: 0.4456
Epoch 44, CIFAR-10 Batch 1:  Loss: 1.3472 Accuracy: 0.4768
Epoch 45, CIFAR-10 Batch 1:  Loss: 1.4583 Accuracy: 0.4568
Epoch 46, CIFAR-10 Batch 1:  Loss: 1.3864 Accuracy: 0.4916
Epoch 47, CIFAR-10 Batch 1:  Loss: 1.4450 Accuracy: 0.4776
Epoch 48, CIFAR-10 Batch 1:  Loss: 1.4338 Accuracy: 0.4794
Epoch 49, CIFAR-10 Batch 1:  Loss: 1.4163 Accuracy: 0.4852
Epoch 50, CIFAR-10 Batch 1:  Loss: 1.5015 Accuracy: 0.4518
Epoch 51, CIFAR-10 Batch 1:  Loss: 1.2733 Accuracy: 0.4982
Epoch 52, CIFAR-10 Batch 1:  Loss: 1.2747 Accuracy: 0.5010
Epoch 53, CIFAR-10 Batch 1:  Loss: 1.2742 Accuracy: 0.5102
Epoch 54, CIFAR-10 Batch 1:  Loss: 1.1779 Accuracy: 0.5264
Epoch 55, CIFAR-10 Batch 1:  Loss: 1.2011 Accuracy: 0.5122
Epoch 56, CIFAR-10 Batch 1:  Loss: 1.1687 Accuracy: 0.5296
Epoch 57, CIFAR-10 Batch 1:  Loss: 1.2115 Accuracy: 0.5148
Epoch 58, CIFAR-10 Batch 1:  Loss: 1.1978 Accuracy: 0.5310
Epoch 59, CIFAR-10 Batch 1:  Loss: 1.0567 Accuracy: 0.5438
Epoch 60, CIFAR-10 Batch 1:  Loss: 1.0710 Accuracy: 0.5416
Epoch 61, CIFAR-10 Batch 1:  Loss: 1.3034 Accuracy: 0.5246
Epoch 62, CIFAR-10 Batch 1:  Loss: 1.2312 Accuracy: 0.5232
Epoch 63, CIFAR-10 Batch 1:  Loss: 0.8969 Accuracy: 0.5768
Epoch 64, CIFAR-10 Batch 1:  Loss: 1.1836 Accuracy: 0.5162
Epoch 65, CIFAR-10 Batch 1:  Loss: 0.9282 Accuracy: 0.5576
Epoch 66, CIFAR-10 Batch 1:  Loss: 0.9694 Accuracy: 0.5752
Epoch 67, CIFAR-10 Batch 1:  Loss: 0.9271 Accuracy: 0.5564
Epoch 68, CIFAR-10 Batch 1:  Loss: 1.0297 Accuracy: 0.5514
Epoch 69, CIFAR-10 Batch 1:  Loss: 0.9722 Accuracy: 0.5600
Epoch 70, CIFAR-10 Batch 1:  Loss: 0.8621 Accuracy: 0.5584
Epoch 71, CIFAR-10 Batch 1:  Loss: 0.7735 Accuracy: 0.5928
Epoch 72, CIFAR-10 Batch 1:  Loss: 0.7675 Accuracy: 0.5842
Epoch 73, CIFAR-10 Batch 1:  Loss: 0.7573 Accuracy: 0.5896
Epoch 74, CIFAR-10 Batch 1:  Loss: 0.8836 Accuracy: 0.5774
Epoch 75, CIFAR-10 Batch 1:  Loss: 0.7967 Accuracy: 0.5888
Epoch 76, CIFAR-10 Batch 1:  Loss: 0.8046 Accuracy: 0.5856
Epoch 77, CIFAR-10 Batch 1:  Loss: 0.8143 Accuracy: 0.5876
Epoch 78, CIFAR-10 Batch 1:  Loss: 0.6007 Accuracy: 0.6284
Epoch 79, CIFAR-10 Batch 1:  Loss: 0.7138 Accuracy: 0.6140
Epoch 80, CIFAR-10 Batch 1:  Loss: 0.7523 Accuracy: 0.5996
Epoch 81, CIFAR-10 Batch 1:  Loss: 0.5827 Accuracy: 0.6090
Epoch 82, CIFAR-10 Batch 1:  Loss: 0.5288 Accuracy: 0.6216
Epoch 83, CIFAR-10 Batch 1:  Loss: 0.6867 Accuracy: 0.5948
Epoch 84, CIFAR-10 Batch 1:  Loss: 0.5612 Accuracy: 0.6026
Epoch 85, CIFAR-10 Batch 1:  Loss: 0.5291 Accuracy: 0.6164
Epoch 86, CIFAR-10 Batch 1:  Loss: 0.5034 Accuracy: 0.6268
Epoch 87, CIFAR-10 Batch 1:  Loss: 0.4220 Accuracy: 0.6418
Epoch 88, CIFAR-10 Batch 1:  Loss: 0.4642 Accuracy: 0.6306
Epoch 89, CIFAR-10 Batch 1:  Loss: 0.4608 Accuracy: 0.6152
Epoch 90, CIFAR-10 Batch 1:  Loss: 0.4452 Accuracy: 0.6202
Epoch 91, CIFAR-10 Batch 1:  Loss: 0.4382 Accuracy: 0.6230
Epoch 92, CIFAR-10 Batch 1:  Loss: 0.4110 Accuracy: 0.6286
Epoch 93, CIFAR-10 Batch 1:  Loss: 0.3510 Accuracy: 0.6536
Epoch 94, CIFAR-10 Batch 1:  Loss: 0.5068 Accuracy: 0.6160
Epoch 95, CIFAR-10 Batch 1:  Loss: 0.3317 Accuracy: 0.6320
Epoch 96, CIFAR-10 Batch 1:  Loss: 0.3361 Accuracy: 0.6398
Epoch 97, CIFAR-10 Batch 1:  Loss: 0.3171 Accuracy: 0.6428
Epoch 98, CIFAR-10 Batch 1:  Loss: 0.3584 Accuracy: 0.6360
Epoch 99, CIFAR-10 Batch 1:  Loss: 0.3089 Accuracy: 0.6298
Epoch 100, CIFAR-10 Batch 1:  Loss: 0.2543 Accuracy: 0.6514
Epoch 101, CIFAR-10 Batch 1:  Loss: 0.3009 Accuracy: 0.6462
Epoch 102, CIFAR-10 Batch 1:  Loss: 0.2655 Accuracy: 0.6338
Epoch 103, CIFAR-10 Batch 1:  Loss: 0.3240 Accuracy: 0.6238
Epoch 104, CIFAR-10 Batch 1:  Loss: 0.2384 Accuracy: 0.6342
Epoch 105, CIFAR-10 Batch 1:  Loss: 0.2554 Accuracy: 0.6394
Epoch 106, CIFAR-10 Batch 1:  Loss: 0.2115 Accuracy: 0.6658
Epoch 107, CIFAR-10 Batch 1:  Loss: 0.1919 Accuracy: 0.6430
Epoch 108, CIFAR-10 Batch 1:  Loss: 0.1703 Accuracy: 0.6618
Epoch 109, CIFAR-10 Batch 1:  Loss: 0.1701 Accuracy: 0.6584
Epoch 110, CIFAR-10 Batch 1:  Loss: 0.2421 Accuracy: 0.6316
Epoch 111, CIFAR-10 Batch 1:  Loss: 0.1909 Accuracy: 0.6492
Epoch 112, CIFAR-10 Batch 1:  Loss: 0.1853 Accuracy: 0.6550
Epoch 113, CIFAR-10 Batch 1:  Loss: 0.1865 Accuracy: 0.6508
Epoch 114, CIFAR-10 Batch 1:  Loss: 0.1740 Accuracy: 0.6566
Epoch 115, CIFAR-10 Batch 1:  Loss: 0.1679 Accuracy: 0.6458
Epoch 116, CIFAR-10 Batch 1:  Loss: 0.1648 Accuracy: 0.6430
Epoch 117, CIFAR-10 Batch 1:  Loss: 0.1855 Accuracy: 0.6450
Epoch 118, CIFAR-10 Batch 1:  Loss: 0.1670 Accuracy: 0.6528
Epoch 119, CIFAR-10 Batch 1:  Loss: 0.1661 Accuracy: 0.6412
Epoch 120, CIFAR-10 Batch 1:  Loss: 0.1051 Accuracy: 0.6566
Epoch 121, CIFAR-10 Batch 1:  Loss: 0.0815 Accuracy: 0.6694
Epoch 122, CIFAR-10 Batch 1:  Loss: 0.0835 Accuracy: 0.6744
Epoch 123, CIFAR-10 Batch 1:  Loss: 0.0970 Accuracy: 0.6672
Epoch 124, CIFAR-10 Batch 1:  Loss: 0.0846 Accuracy: 0.6546
Epoch 125, CIFAR-10 Batch 1:  Loss: 0.0916 Accuracy: 0.6660
Epoch 126, CIFAR-10 Batch 1:  Loss: 0.0668 Accuracy: 0.6756
Epoch 127, CIFAR-10 Batch 1:  Loss: 0.0661 Accuracy: 0.6726
Epoch 128, CIFAR-10 Batch 1:  Loss: 0.0693 Accuracy: 0.6738
Epoch 129, CIFAR-10 Batch 1:  Loss: 0.0782 Accuracy: 0.6628
Epoch 130, CIFAR-10 Batch 1:  Loss: 0.0725 Accuracy: 0.6670
Epoch 131, CIFAR-10 Batch 1:  Loss: 0.0957 Accuracy: 0.6616
Epoch 132, CIFAR-10 Batch 1:  Loss: 0.0959 Accuracy: 0.6536
Epoch 133, CIFAR-10 Batch 1:  Loss: 0.0861 Accuracy: 0.6546
Epoch 134, CIFAR-10 Batch 1:  Loss: 0.0524 Accuracy: 0.6670
Epoch 135, CIFAR-10 Batch 1:  Loss: 0.0457 Accuracy: 0.6680
Epoch 136, CIFAR-10 Batch 1:  Loss: 0.0494 Accuracy: 0.6748
Epoch 137, CIFAR-10 Batch 1:  Loss: 0.0328 Accuracy: 0.6736
Epoch 138, CIFAR-10 Batch 1:  Loss: 0.0367 Accuracy: 0.6800
Epoch 139, CIFAR-10 Batch 1:  Loss: 0.0377 Accuracy: 0.6724
Epoch 140, CIFAR-10 Batch 1:  Loss: 0.0390 Accuracy: 0.6774
Epoch 141, CIFAR-10 Batch 1:  Loss: 0.0469 Accuracy: 0.6624
Epoch 142, CIFAR-10 Batch 1:  Loss: 0.0420 Accuracy: 0.6754
Epoch 143, CIFAR-10 Batch 1:  Loss: 0.0444 Accuracy: 0.6684
Epoch 144, CIFAR-10 Batch 1:  Loss: 0.0311 Accuracy: 0.6678
Epoch 145, CIFAR-10 Batch 1:  Loss: 0.0244 Accuracy: 0.6870
Epoch 146, CIFAR-10 Batch 1:  Loss: 0.0256 Accuracy: 0.6916
Epoch 147, CIFAR-10 Batch 1:  Loss: 0.0382 Accuracy: 0.6788
Epoch 148, CIFAR-10 Batch 1:  Loss: 0.0364 Accuracy: 0.6724
Epoch 149, CIFAR-10 Batch 1:  Loss: 0.0314 Accuracy: 0.6722
Epoch 150, CIFAR-10 Batch 1:  Loss: 0.0238 Accuracy: 0.6872
Epoch 151, CIFAR-10 Batch 1:  Loss: 0.0274 Accuracy: 0.6614
Epoch 152, CIFAR-10 Batch 1:  Loss: 0.0197 Accuracy: 0.6724
Epoch 153, CIFAR-10 Batch 1:  Loss: 0.0162 Accuracy: 0.6890
Epoch 154, CIFAR-10 Batch 1:  Loss: 0.0305 Accuracy: 0.6794
Epoch 155, CIFAR-10 Batch 1:  Loss: 0.0187 Accuracy: 0.6912
Epoch 156, CIFAR-10 Batch 1:  Loss: 0.0214 Accuracy: 0.6702
Epoch 157, CIFAR-10 Batch 1:  Loss: 0.0166 Accuracy: 0.6830
Epoch 158, CIFAR-10 Batch 1:  Loss: 0.0180 Accuracy: 0.6758
Epoch 159, CIFAR-10 Batch 1:  Loss: 0.0201 Accuracy: 0.6864
Epoch 160, CIFAR-10 Batch 1:  Loss: 0.0141 Accuracy: 0.6844
Epoch 161, CIFAR-10 Batch 1:  Loss: 0.0152 Accuracy: 0.6800
Epoch 162, CIFAR-10 Batch 1:  Loss: 0.0171 Accuracy: 0.6790
Epoch 163, CIFAR-10 Batch 1:  Loss: 0.0108 Accuracy: 0.6830
Epoch 164, CIFAR-10 Batch 1:  Loss: 0.0092 Accuracy: 0.6878
Epoch 165, CIFAR-10 Batch 1:  Loss: 0.0300 Accuracy: 0.6662
Epoch 166, CIFAR-10 Batch 1:  Loss: 0.0240 Accuracy: 0.6798
Epoch 167, CIFAR-10 Batch 1:  Loss: 0.0102 Accuracy: 0.6732
Epoch 168, CIFAR-10 Batch 1:  Loss: 0.0129 Accuracy: 0.6688
Epoch 169, CIFAR-10 Batch 1:  Loss: 0.0084 Accuracy: 0.6810
Epoch 170, CIFAR-10 Batch 1:  Loss: 0.0082 Accuracy: 0.6846
Epoch 171, CIFAR-10 Batch 1:  Loss: 0.0064 Accuracy: 0.6870
Epoch 172, CIFAR-10 Batch 1:  Loss: 0.0122 Accuracy: 0.6616
Epoch 173, CIFAR-10 Batch 1:  Loss: 0.0074 Accuracy: 0.6762
Epoch 174, CIFAR-10 Batch 1:  Loss: 0.0214 Accuracy: 0.6730
Epoch 175, CIFAR-10 Batch 1:  Loss: 0.0078 Accuracy: 0.6768
Epoch 176, CIFAR-10 Batch 1:  Loss: 0.0061 Accuracy: 0.6896
Epoch 177, CIFAR-10 Batch 1:  Loss: 0.0217 Accuracy: 0.6620
Epoch 178, CIFAR-10 Batch 1:  Loss: 0.0059 Accuracy: 0.6882
Epoch 179, CIFAR-10 Batch 1:  Loss: 0.0048 Accuracy: 0.6914
Epoch 180, CIFAR-10 Batch 1:  Loss: 0.0081 Accuracy: 0.6790
Epoch 181, CIFAR-10 Batch 1:  Loss: 0.0058 Accuracy: 0.6862
Epoch 182, CIFAR-10 Batch 1:  Loss: 0.0070 Accuracy: 0.6808
Epoch 183, CIFAR-10 Batch 1:  Loss: 0.0163 Accuracy: 0.6722
Epoch 184, CIFAR-10 Batch 1:  Loss: 0.0093 Accuracy: 0.6856
Epoch 185, CIFAR-10 Batch 1:  Loss: 0.0096 Accuracy: 0.6850
Epoch 186, CIFAR-10 Batch 1:  Loss: 0.0100 Accuracy: 0.6802
Epoch 187, CIFAR-10 Batch 1:  Loss: 0.0064 Accuracy: 0.6758
Epoch 188, CIFAR-10 Batch 1:  Loss: 0.0148 Accuracy: 0.6520
Epoch 189, CIFAR-10 Batch 1:  Loss: 0.0065 Accuracy: 0.6796
Epoch 190, CIFAR-10 Batch 1:  Loss: 0.0063 Accuracy: 0.6796
Epoch 191, CIFAR-10 Batch 1:  Loss: 0.0041 Accuracy: 0.6740
Epoch 192, CIFAR-10 Batch 1:  Loss: 0.0039 Accuracy: 0.6714
Epoch 193, CIFAR-10 Batch 1:  Loss: 0.0027 Accuracy: 0.6938
Epoch 194, CIFAR-10 Batch 1:  Loss: 0.0022 Accuracy: 0.6938
Epoch 195, CIFAR-10 Batch 1:  Loss: 0.0029 Accuracy: 0.6778
Epoch 196, CIFAR-10 Batch 1:  Loss: 0.0093 Accuracy: 0.6772
Epoch 197, CIFAR-10 Batch 1:  Loss: 0.0029 Accuracy: 0.6734
Epoch 198, CIFAR-10 Batch 1:  Loss: 0.0023 Accuracy: 0.6838
Epoch 199, CIFAR-10 Batch 1:  Loss: 0.0034 Accuracy: 0.6754
Epoch 200, CIFAR-10 Batch 1:  Loss: 0.0028 Accuracy: 0.6836

完全训练模型

现在，单个 CIFAR-10 部分的准确率已经不错了，试试所有五个部分吧。



In [18]:

    
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
save_model_path = './image_classification'

print('Training...')
with tf.Session() as sess:
    # Initializing the variables
    sess.run(tf.global_variables_initializer())
    
    # Training cycle
    for epoch in range(epochs):
        # Loop over all batches
        n_batches = 5
        for batch_i in range(1, n_batches + 1):
            for batch_features, batch_labels in helper.load_preprocess_training_batch(batch_i, batch_size):
                train_neural_network(sess, optimizer, keep_probability, batch_features, batch_labels)
            print('Epoch {:>2}, CIFAR-10 Batch {}:  '.format(epoch + 1, batch_i), end='')
            print_stats(sess, batch_features, batch_labels, cost, accuracy)
            
    # Save Model
    saver = tf.train.Saver()
    save_path = saver.save(sess, save_model_path)









    



Training...
Epoch  1, CIFAR-10 Batch 1:  Loss: 2.2990 Accuracy: 0.1242
Epoch  1, CIFAR-10 Batch 2:  Loss: 2.2989 Accuracy: 0.0984
Epoch  1, CIFAR-10 Batch 3:  Loss: 2.2903 Accuracy: 0.1134
Epoch  1, CIFAR-10 Batch 4:  Loss: 2.2772 Accuracy: 0.1192
Epoch  1, CIFAR-10 Batch 5:  Loss: 2.2432 Accuracy: 0.1762
Epoch  2, CIFAR-10 Batch 1:  Loss: 2.1821 Accuracy: 0.2398
Epoch  2, CIFAR-10 Batch 2:  Loss: 2.0755 Accuracy: 0.2804
Epoch  2, CIFAR-10 Batch 3:  Loss: 1.9589 Accuracy: 0.2936
Epoch  2, CIFAR-10 Batch 4:  Loss: 1.9475 Accuracy: 0.3336
Epoch  2, CIFAR-10 Batch 5:  Loss: 1.9736 Accuracy: 0.3324
Epoch  3, CIFAR-10 Batch 1:  Loss: 2.0157 Accuracy: 0.3342
Epoch  3, CIFAR-10 Batch 2:  Loss: 1.8046 Accuracy: 0.3512
Epoch  3, CIFAR-10 Batch 3:  Loss: 1.7284 Accuracy: 0.3564
Epoch  3, CIFAR-10 Batch 4:  Loss: 1.7664 Accuracy: 0.3980
Epoch  3, CIFAR-10 Batch 5:  Loss: 1.7433 Accuracy: 0.3826
Epoch  4, CIFAR-10 Batch 1:  Loss: 1.8219 Accuracy: 0.4198
Epoch  4, CIFAR-10 Batch 2:  Loss: 1.6349 Accuracy: 0.4172
Epoch  4, CIFAR-10 Batch 3:  Loss: 1.5965 Accuracy: 0.4204
Epoch  4, CIFAR-10 Batch 4:  Loss: 1.6699 Accuracy: 0.4382
Epoch  4, CIFAR-10 Batch 5:  Loss: 1.6740 Accuracy: 0.4078
Epoch  5, CIFAR-10 Batch 1:  Loss: 1.7249 Accuracy: 0.4488
Epoch  5, CIFAR-10 Batch 2:  Loss: 1.4835 Accuracy: 0.4432
Epoch  5, CIFAR-10 Batch 3:  Loss: 1.4859 Accuracy: 0.4434
Epoch  5, CIFAR-10 Batch 4:  Loss: 1.6320 Accuracy: 0.4500
Epoch  5, CIFAR-10 Batch 5:  Loss: 1.5858 Accuracy: 0.4222
Epoch  6, CIFAR-10 Batch 1:  Loss: 1.6674 Accuracy: 0.4660
Epoch  6, CIFAR-10 Batch 2:  Loss: 1.5061 Accuracy: 0.4328
Epoch  6, CIFAR-10 Batch 3:  Loss: 1.4265 Accuracy: 0.4772
Epoch  6, CIFAR-10 Batch 4:  Loss: 1.5376 Accuracy: 0.4732
Epoch  6, CIFAR-10 Batch 5:  Loss: 1.4722 Accuracy: 0.4830
Epoch  7, CIFAR-10 Batch 1:  Loss: 1.5917 Accuracy: 0.4764
Epoch  7, CIFAR-10 Batch 2:  Loss: 1.4341 Accuracy: 0.4738
Epoch  7, CIFAR-10 Batch 3:  Loss: 1.3711 Accuracy: 0.4870
Epoch  7, CIFAR-10 Batch 4:  Loss: 1.4752 Accuracy: 0.4874
Epoch  7, CIFAR-10 Batch 5:  Loss: 1.4793 Accuracy: 0.4506
Epoch  8, CIFAR-10 Batch 1:  Loss: 1.5289 Accuracy: 0.5098
Epoch  8, CIFAR-10 Batch 2:  Loss: 1.4208 Accuracy: 0.5026
Epoch  8, CIFAR-10 Batch 3:  Loss: 1.3289 Accuracy: 0.5000
Epoch  8, CIFAR-10 Batch 4:  Loss: 1.4045 Accuracy: 0.5242
Epoch  8, CIFAR-10 Batch 5:  Loss: 1.4499 Accuracy: 0.4628
Epoch  9, CIFAR-10 Batch 1:  Loss: 1.4232 Accuracy: 0.5146
Epoch  9, CIFAR-10 Batch 2:  Loss: 1.3775 Accuracy: 0.5368
Epoch  9, CIFAR-10 Batch 3:  Loss: 1.3581 Accuracy: 0.5020
Epoch  9, CIFAR-10 Batch 4:  Loss: 1.3345 Accuracy: 0.5460
Epoch  9, CIFAR-10 Batch 5:  Loss: 1.3752 Accuracy: 0.5066
Epoch 10, CIFAR-10 Batch 1:  Loss: 1.3309 Accuracy: 0.5494
Epoch 10, CIFAR-10 Batch 2:  Loss: 1.3696 Accuracy: 0.5344
Epoch 10, CIFAR-10 Batch 3:  Loss: 1.2076 Accuracy: 0.5432
Epoch 10, CIFAR-10 Batch 4:  Loss: 1.2601 Accuracy: 0.5446
Epoch 10, CIFAR-10 Batch 5:  Loss: 1.3157 Accuracy: 0.5092
Epoch 11, CIFAR-10 Batch 1:  Loss: 1.2238 Accuracy: 0.5454
Epoch 11, CIFAR-10 Batch 2:  Loss: 1.4382 Accuracy: 0.5516
Epoch 11, CIFAR-10 Batch 3:  Loss: 1.2283 Accuracy: 0.5364
Epoch 11, CIFAR-10 Batch 4:  Loss: 1.1790 Accuracy: 0.5680
Epoch 11, CIFAR-10 Batch 5:  Loss: 1.2405 Accuracy: 0.5320
Epoch 12, CIFAR-10 Batch 1:  Loss: 1.1388 Accuracy: 0.5820
Epoch 12, CIFAR-10 Batch 2:  Loss: 1.3224 Accuracy: 0.5634
Epoch 12, CIFAR-10 Batch 3:  Loss: 1.1503 Accuracy: 0.5740
Epoch 12, CIFAR-10 Batch 4:  Loss: 1.1259 Accuracy: 0.5864
Epoch 12, CIFAR-10 Batch 5:  Loss: 1.2610 Accuracy: 0.5458
Epoch 13, CIFAR-10 Batch 1:  Loss: 1.1514 Accuracy: 0.5936
Epoch 13, CIFAR-10 Batch 2:  Loss: 1.2045 Accuracy: 0.6064
Epoch 13, CIFAR-10 Batch 3:  Loss: 1.1842 Accuracy: 0.5526
Epoch 13, CIFAR-10 Batch 4:  Loss: 1.0902 Accuracy: 0.5976
Epoch 13, CIFAR-10 Batch 5:  Loss: 1.2523 Accuracy: 0.5404
Epoch 14, CIFAR-10 Batch 1:  Loss: 1.0472 Accuracy: 0.6164
Epoch 14, CIFAR-10 Batch 2:  Loss: 1.1703 Accuracy: 0.6236
Epoch 14, CIFAR-10 Batch 3:  Loss: 0.9812 Accuracy: 0.6112
Epoch 14, CIFAR-10 Batch 4:  Loss: 1.0638 Accuracy: 0.5832
Epoch 14, CIFAR-10 Batch 5:  Loss: 1.1103 Accuracy: 0.5834
Epoch 15, CIFAR-10 Batch 1:  Loss: 1.0332 Accuracy: 0.6114
Epoch 15, CIFAR-10 Batch 2:  Loss: 1.1341 Accuracy: 0.6414
Epoch 15, CIFAR-10 Batch 3:  Loss: 1.0505 Accuracy: 0.5838
Epoch 15, CIFAR-10 Batch 4:  Loss: 0.9442 Accuracy: 0.6464
Epoch 15, CIFAR-10 Batch 5:  Loss: 1.1141 Accuracy: 0.5910
Epoch 16, CIFAR-10 Batch 1:  Loss: 1.0301 Accuracy: 0.6044
Epoch 16, CIFAR-10 Batch 2:  Loss: 1.1813 Accuracy: 0.6204
Epoch 16, CIFAR-10 Batch 3:  Loss: 1.0379 Accuracy: 0.5800
Epoch 16, CIFAR-10 Batch 4:  Loss: 0.9038 Accuracy: 0.6382
Epoch 16, CIFAR-10 Batch 5:  Loss: 1.1423 Accuracy: 0.6112
Epoch 17, CIFAR-10 Batch 1:  Loss: 0.9734 Accuracy: 0.6424
Epoch 17, CIFAR-10 Batch 2:  Loss: 1.0955 Accuracy: 0.6250
Epoch 17, CIFAR-10 Batch 3:  Loss: 0.9875 Accuracy: 0.5926
Epoch 17, CIFAR-10 Batch 4:  Loss: 0.8259 Accuracy: 0.6594
Epoch 17, CIFAR-10 Batch 5:  Loss: 1.0233 Accuracy: 0.6412
Epoch 18, CIFAR-10 Batch 1:  Loss: 0.8530 Accuracy: 0.6746
Epoch 18, CIFAR-10 Batch 2:  Loss: 1.0345 Accuracy: 0.6548
Epoch 18, CIFAR-10 Batch 3:  Loss: 0.8824 Accuracy: 0.6372
Epoch 18, CIFAR-10 Batch 4:  Loss: 0.8165 Accuracy: 0.6660
Epoch 18, CIFAR-10 Batch 5:  Loss: 0.9477 Accuracy: 0.6742
Epoch 19, CIFAR-10 Batch 1:  Loss: 0.8394 Accuracy: 0.6670
Epoch 19, CIFAR-10 Batch 2:  Loss: 1.0045 Accuracy: 0.6578
Epoch 19, CIFAR-10 Batch 3:  Loss: 0.7900 Accuracy: 0.6604
Epoch 19, CIFAR-10 Batch 4:  Loss: 0.7782 Accuracy: 0.6710
Epoch 19, CIFAR-10 Batch 5:  Loss: 0.9517 Accuracy: 0.6470
Epoch 20, CIFAR-10 Batch 1:  Loss: 0.7991 Accuracy: 0.6802
Epoch 20, CIFAR-10 Batch 2:  Loss: 1.0704 Accuracy: 0.6352
Epoch 20, CIFAR-10 Batch 3:  Loss: 0.8044 Accuracy: 0.6554
Epoch 20, CIFAR-10 Batch 4:  Loss: 0.8120 Accuracy: 0.6718
Epoch 20, CIFAR-10 Batch 5:  Loss: 0.9089 Accuracy: 0.6626
Epoch 21, CIFAR-10 Batch 1:  Loss: 0.7858 Accuracy: 0.6850
Epoch 21, CIFAR-10 Batch 2:  Loss: 1.0655 Accuracy: 0.6426
Epoch 21, CIFAR-10 Batch 3:  Loss: 0.7813 Accuracy: 0.6518
Epoch 21, CIFAR-10 Batch 4:  Loss: 0.8061 Accuracy: 0.6624
Epoch 21, CIFAR-10 Batch 5:  Loss: 0.8049 Accuracy: 0.6740
Epoch 22, CIFAR-10 Batch 1:  Loss: 0.7169 Accuracy: 0.6954
Epoch 22, CIFAR-10 Batch 2:  Loss: 0.9672 Accuracy: 0.6764
Epoch 22, CIFAR-10 Batch 3:  Loss: 0.7923 Accuracy: 0.6416
Epoch 22, CIFAR-10 Batch 4:  Loss: 0.7833 Accuracy: 0.6720
Epoch 22, CIFAR-10 Batch 5:  Loss: 0.8444 Accuracy: 0.6804
Epoch 23, CIFAR-10 Batch 1:  Loss: 0.7841 Accuracy: 0.6918
Epoch 23, CIFAR-10 Batch 2:  Loss: 0.8397 Accuracy: 0.7058
Epoch 23, CIFAR-10 Batch 3:  Loss: 0.7241 Accuracy: 0.6434
Epoch 23, CIFAR-10 Batch 4:  Loss: 0.7098 Accuracy: 0.6898
Epoch 23, CIFAR-10 Batch 5:  Loss: 0.8419 Accuracy: 0.6642
Epoch 24, CIFAR-10 Batch 1:  Loss: 0.7428 Accuracy: 0.6838
Epoch 24, CIFAR-10 Batch 2:  Loss: 0.7919 Accuracy: 0.6878
Epoch 24, CIFAR-10 Batch 3:  Loss: 0.6420 Accuracy: 0.6858
Epoch 24, CIFAR-10 Batch 4:  Loss: 0.6830 Accuracy: 0.6962
Epoch 24, CIFAR-10 Batch 5:  Loss: 0.8711 Accuracy: 0.6720
Epoch 25, CIFAR-10 Batch 1:  Loss: 0.6751 Accuracy: 0.7030
Epoch 25, CIFAR-10 Batch 2:  Loss: 0.7409 Accuracy: 0.6998
Epoch 25, CIFAR-10 Batch 3:  Loss: 0.5714 Accuracy: 0.6954
Epoch 25, CIFAR-10 Batch 4:  Loss: 0.5789 Accuracy: 0.7102
Epoch 25, CIFAR-10 Batch 5:  Loss: 0.6612 Accuracy: 0.7132
Epoch 26, CIFAR-10 Batch 1:  Loss: 0.6890 Accuracy: 0.6960
Epoch 26, CIFAR-10 Batch 2:  Loss: 0.7506 Accuracy: 0.7046
Epoch 26, CIFAR-10 Batch 3:  Loss: 0.7047 Accuracy: 0.6746
Epoch 26, CIFAR-10 Batch 4:  Loss: 0.7435 Accuracy: 0.6876
Epoch 26, CIFAR-10 Batch 5:  Loss: 0.7175 Accuracy: 0.6966
Epoch 27, CIFAR-10 Batch 1:  Loss: 0.6820 Accuracy: 0.7110
Epoch 27, CIFAR-10 Batch 2:  Loss: 0.7786 Accuracy: 0.6828
Epoch 27, CIFAR-10 Batch 3:  Loss: 0.5677 Accuracy: 0.6962
Epoch 27, CIFAR-10 Batch 4:  Loss: 0.5545 Accuracy: 0.7230
Epoch 27, CIFAR-10 Batch 5:  Loss: 0.7726 Accuracy: 0.7030
Epoch 28, CIFAR-10 Batch 1:  Loss: 0.7231 Accuracy: 0.7082
Epoch 28, CIFAR-10 Batch 2:  Loss: 0.7387 Accuracy: 0.6892
Epoch 28, CIFAR-10 Batch 3:  Loss: 0.5264 Accuracy: 0.6916
Epoch 28, CIFAR-10 Batch 4:  Loss: 0.5427 Accuracy: 0.7216
Epoch 28, CIFAR-10 Batch 5:  Loss: 0.6609 Accuracy: 0.7050
Epoch 29, CIFAR-10 Batch 1:  Loss: 0.6431 Accuracy: 0.7190
Epoch 29, CIFAR-10 Batch 2:  Loss: 0.7013 Accuracy: 0.7066
Epoch 29, CIFAR-10 Batch 3:  Loss: 0.4793 Accuracy: 0.7118
Epoch 29, CIFAR-10 Batch 4:  Loss: 0.6576 Accuracy: 0.6996
Epoch 29, CIFAR-10 Batch 5:  Loss: 0.7519 Accuracy: 0.6874
Epoch 30, CIFAR-10 Batch 1:  Loss: 0.6193 Accuracy: 0.7420
Epoch 30, CIFAR-10 Batch 2:  Loss: 0.6538 Accuracy: 0.7138
Epoch 30, CIFAR-10 Batch 3:  Loss: 0.4199 Accuracy: 0.7212
Epoch 30, CIFAR-10 Batch 4:  Loss: 0.6472 Accuracy: 0.7076
Epoch 30, CIFAR-10 Batch 5:  Loss: 0.6747 Accuracy: 0.6896
Epoch 31, CIFAR-10 Batch 1:  Loss: 0.5582 Accuracy: 0.7334
Epoch 31, CIFAR-10 Batch 2:  Loss: 0.6682 Accuracy: 0.6976
Epoch 31, CIFAR-10 Batch 3:  Loss: 0.5242 Accuracy: 0.7138
Epoch 31, CIFAR-10 Batch 4:  Loss: 0.5034 Accuracy: 0.7270
Epoch 31, CIFAR-10 Batch 5:  Loss: 0.5787 Accuracy: 0.7188
Epoch 32, CIFAR-10 Batch 1:  Loss: 0.5440 Accuracy: 0.7352
Epoch 32, CIFAR-10 Batch 2:  Loss: 0.6663 Accuracy: 0.6984
Epoch 32, CIFAR-10 Batch 3:  Loss: 0.4486 Accuracy: 0.7248
Epoch 32, CIFAR-10 Batch 4:  Loss: 0.5507 Accuracy: 0.7260
Epoch 32, CIFAR-10 Batch 5:  Loss: 0.5716 Accuracy: 0.7218
Epoch 33, CIFAR-10 Batch 1:  Loss: 0.4697 Accuracy: 0.7300
Epoch 33, CIFAR-10 Batch 2:  Loss: 0.5284 Accuracy: 0.7210
Epoch 33, CIFAR-10 Batch 3:  Loss: 0.5077 Accuracy: 0.7036
Epoch 33, CIFAR-10 Batch 4:  Loss: 0.5529 Accuracy: 0.7162
Epoch 33, CIFAR-10 Batch 5:  Loss: 0.5547 Accuracy: 0.7324
Epoch 34, CIFAR-10 Batch 1:  Loss: 0.4196 Accuracy: 0.7514
Epoch 34, CIFAR-10 Batch 2:  Loss: 0.5479 Accuracy: 0.7204
Epoch 34, CIFAR-10 Batch 3:  Loss: 0.4248 Accuracy: 0.7364
Epoch 34, CIFAR-10 Batch 4:  Loss: 0.4809 Accuracy: 0.7402
Epoch 34, CIFAR-10 Batch 5:  Loss: 0.5629 Accuracy: 0.7230
Epoch 35, CIFAR-10 Batch 1:  Loss: 0.5213 Accuracy: 0.7484
Epoch 35, CIFAR-10 Batch 2:  Loss: 0.7084 Accuracy: 0.6712
Epoch 35, CIFAR-10 Batch 3:  Loss: 0.3471 Accuracy: 0.7396
Epoch 35, CIFAR-10 Batch 4:  Loss: 0.4446 Accuracy: 0.7326
Epoch 35, CIFAR-10 Batch 5:  Loss: 0.4609 Accuracy: 0.7348
Epoch 36, CIFAR-10 Batch 1:  Loss: 0.4922 Accuracy: 0.7350
Epoch 36, CIFAR-10 Batch 2:  Loss: 0.6410 Accuracy: 0.7082
Epoch 36, CIFAR-10 Batch 3:  Loss: 0.3765 Accuracy: 0.7278
Epoch 36, CIFAR-10 Batch 4:  Loss: 0.4533 Accuracy: 0.7218
Epoch 36, CIFAR-10 Batch 5:  Loss: 0.4645 Accuracy: 0.7250
Epoch 37, CIFAR-10 Batch 1:  Loss: 0.4627 Accuracy: 0.7498
Epoch 37, CIFAR-10 Batch 2:  Loss: 0.5271 Accuracy: 0.7338
Epoch 37, CIFAR-10 Batch 3:  Loss: 0.3728 Accuracy: 0.7374
Epoch 37, CIFAR-10 Batch 4:  Loss: 0.4523 Accuracy: 0.7340
Epoch 37, CIFAR-10 Batch 5:  Loss: 0.4630 Accuracy: 0.7520
Epoch 38, CIFAR-10 Batch 1:  Loss: 0.5189 Accuracy: 0.7442
Epoch 38, CIFAR-10 Batch 2:  Loss: 0.5561 Accuracy: 0.7230
Epoch 38, CIFAR-10 Batch 3:  Loss: 0.2987 Accuracy: 0.7396
Epoch 38, CIFAR-10 Batch 4:  Loss: 0.4000 Accuracy: 0.7484
Epoch 38, CIFAR-10 Batch 5:  Loss: 0.4616 Accuracy: 0.7384
Epoch 39, CIFAR-10 Batch 1:  Loss: 0.4557 Accuracy: 0.7560
Epoch 39, CIFAR-10 Batch 2:  Loss: 0.5801 Accuracy: 0.7286
Epoch 39, CIFAR-10 Batch 3:  Loss: 0.3125 Accuracy: 0.7412
Epoch 39, CIFAR-10 Batch 4:  Loss: 0.4242 Accuracy: 0.7276
Epoch 39, CIFAR-10 Batch 5:  Loss: 0.4778 Accuracy: 0.7144
Epoch 40, CIFAR-10 Batch 1:  Loss: 0.4739 Accuracy: 0.7380
Epoch 40, CIFAR-10 Batch 2:  Loss: 0.4849 Accuracy: 0.7390
Epoch 40, CIFAR-10 Batch 3:  Loss: 0.3120 Accuracy: 0.7432
Epoch 40, CIFAR-10 Batch 4:  Loss: 0.4082 Accuracy: 0.7340
Epoch 40, CIFAR-10 Batch 5:  Loss: 0.5077 Accuracy: 0.7244
Epoch 41, CIFAR-10 Batch 1:  Loss: 0.4157 Accuracy: 0.7522
Epoch 41, CIFAR-10 Batch 2:  Loss: 0.4557 Accuracy: 0.7448
Epoch 41, CIFAR-10 Batch 3:  Loss: 0.3861 Accuracy: 0.7230
Epoch 41, CIFAR-10 Batch 4:  Loss: 0.3625 Accuracy: 0.7510
Epoch 41, CIFAR-10 Batch 5:  Loss: 0.4289 Accuracy: 0.7454
Epoch 42, CIFAR-10 Batch 1:  Loss: 0.4259 Accuracy: 0.7476
Epoch 42, CIFAR-10 Batch 2:  Loss: 0.4318 Accuracy: 0.7506
Epoch 42, CIFAR-10 Batch 3:  Loss: 0.3366 Accuracy: 0.7372
Epoch 42, CIFAR-10 Batch 4:  Loss: 0.4306 Accuracy: 0.7356
Epoch 42, CIFAR-10 Batch 5:  Loss: 0.3897 Accuracy: 0.7408
Epoch 43, CIFAR-10 Batch 1:  Loss: 0.3964 Accuracy: 0.7608
Epoch 43, CIFAR-10 Batch 2:  Loss: 0.4090 Accuracy: 0.7450
Epoch 43, CIFAR-10 Batch 3:  Loss: 0.2613 Accuracy: 0.7630
Epoch 43, CIFAR-10 Batch 4:  Loss: 0.4524 Accuracy: 0.7304
Epoch 43, CIFAR-10 Batch 5:  Loss: 0.3883 Accuracy: 0.7356
Epoch 44, CIFAR-10 Batch 1:  Loss: 0.3773 Accuracy: 0.7708
Epoch 44, CIFAR-10 Batch 2:  Loss: 0.4500 Accuracy: 0.7426
Epoch 44, CIFAR-10 Batch 3:  Loss: 0.2481 Accuracy: 0.7578
Epoch 44, CIFAR-10 Batch 4:  Loss: 0.3320 Accuracy: 0.7530
Epoch 44, CIFAR-10 Batch 5:  Loss: 0.4072 Accuracy: 0.7472
Epoch 45, CIFAR-10 Batch 1:  Loss: 0.3763 Accuracy: 0.7614
Epoch 45, CIFAR-10 Batch 2:  Loss: 0.4845 Accuracy: 0.7414
Epoch 45, CIFAR-10 Batch 3:  Loss: 0.2826 Accuracy: 0.7508
Epoch 45, CIFAR-10 Batch 4:  Loss: 0.3281 Accuracy: 0.7436
Epoch 45, CIFAR-10 Batch 5:  Loss: 0.3487 Accuracy: 0.7534
Epoch 46, CIFAR-10 Batch 1:  Loss: 0.4031 Accuracy: 0.7546
Epoch 46, CIFAR-10 Batch 2:  Loss: 0.3633 Accuracy: 0.7678
Epoch 46, CIFAR-10 Batch 3:  Loss: 0.2409 Accuracy: 0.7632
Epoch 46, CIFAR-10 Batch 4:  Loss: 0.3666 Accuracy: 0.7542
Epoch 46, CIFAR-10 Batch 5:  Loss: 0.3941 Accuracy: 0.7300
Epoch 47, CIFAR-10 Batch 1:  Loss: 0.3484 Accuracy: 0.7542
Epoch 47, CIFAR-10 Batch 2:  Loss: 0.4104 Accuracy: 0.7450
Epoch 47, CIFAR-10 Batch 3:  Loss: 0.2377 Accuracy: 0.7626
Epoch 47, CIFAR-10 Batch 4:  Loss: 0.3378 Accuracy: 0.7440
Epoch 47, CIFAR-10 Batch 5:  Loss: 0.3510 Accuracy: 0.7476
Epoch 48, CIFAR-10 Batch 1:  Loss: 0.4207 Accuracy: 0.7376
Epoch 48, CIFAR-10 Batch 2:  Loss: 0.4389 Accuracy: 0.7316
Epoch 48, CIFAR-10 Batch 3:  Loss: 0.3238 Accuracy: 0.7228
Epoch 48, CIFAR-10 Batch 4:  Loss: 0.3445 Accuracy: 0.7444
Epoch 48, CIFAR-10 Batch 5:  Loss: 0.3402 Accuracy: 0.7372
Epoch 49, CIFAR-10 Batch 1:  Loss: 0.3516 Accuracy: 0.7594
Epoch 49, CIFAR-10 Batch 2:  Loss: 0.3818 Accuracy: 0.7540
Epoch 49, CIFAR-10 Batch 3:  Loss: 0.2065 Accuracy: 0.7682
Epoch 49, CIFAR-10 Batch 4:  Loss: 0.3164 Accuracy: 0.7456
Epoch 49, CIFAR-10 Batch 5:  Loss: 0.2689 Accuracy: 0.7660
Epoch 50, CIFAR-10 Batch 1:  Loss: 0.3109 Accuracy: 0.7694
Epoch 50, CIFAR-10 Batch 2:  Loss: 0.4700 Accuracy: 0.7272
Epoch 50, CIFAR-10 Batch 3:  Loss: 0.2431 Accuracy: 0.7674
Epoch 50, CIFAR-10 Batch 4:  Loss: 0.2287 Accuracy: 0.7774
Epoch 50, CIFAR-10 Batch 5:  Loss: 0.3236 Accuracy: 0.7576
Epoch 51, CIFAR-10 Batch 1:  Loss: 0.3493 Accuracy: 0.7778
Epoch 51, CIFAR-10 Batch 2:  Loss: 0.3794 Accuracy: 0.7658
Epoch 51, CIFAR-10 Batch 3:  Loss: 0.2293 Accuracy: 0.7552
Epoch 51, CIFAR-10 Batch 4:  Loss: 0.2817 Accuracy: 0.7592
Epoch 51, CIFAR-10 Batch 5:  Loss: 0.3450 Accuracy: 0.7548
Epoch 52, CIFAR-10 Batch 1:  Loss: 0.3077 Accuracy: 0.7844
Epoch 52, CIFAR-10 Batch 2:  Loss: 0.3874 Accuracy: 0.7544
Epoch 52, CIFAR-10 Batch 3:  Loss: 0.2020 Accuracy: 0.7696
Epoch 52, CIFAR-10 Batch 4:  Loss: 0.2275 Accuracy: 0.7654
Epoch 52, CIFAR-10 Batch 5:  Loss: 0.2890 Accuracy: 0.7626
Epoch 53, CIFAR-10 Batch 1:  Loss: 0.2955 Accuracy: 0.7866
Epoch 53, CIFAR-10 Batch 2:  Loss: 0.4292 Accuracy: 0.7440
Epoch 53, CIFAR-10 Batch 3:  Loss: 0.1879 Accuracy: 0.7640
Epoch 53, CIFAR-10 Batch 4:  Loss: 0.2703 Accuracy: 0.7720
Epoch 53, CIFAR-10 Batch 5:  Loss: 0.3677 Accuracy: 0.7470
Epoch 54, CIFAR-10 Batch 1:  Loss: 0.2827 Accuracy: 0.7698
Epoch 54, CIFAR-10 Batch 2:  Loss: 0.2753 Accuracy: 0.7724
Epoch 54, CIFAR-10 Batch 3:  Loss: 0.1960 Accuracy: 0.7746
Epoch 54, CIFAR-10 Batch 4:  Loss: 0.2482 Accuracy: 0.7790
Epoch 54, CIFAR-10 Batch 5:  Loss: 0.2988 Accuracy: 0.7722
Epoch 55, CIFAR-10 Batch 1:  Loss: 0.2741 Accuracy: 0.7806
Epoch 55, CIFAR-10 Batch 2:  Loss: 0.2970 Accuracy: 0.7642
Epoch 55, CIFAR-10 Batch 3:  Loss: 0.2392 Accuracy: 0.7692
Epoch 55, CIFAR-10 Batch 4:  Loss: 0.1716 Accuracy: 0.7882
Epoch 55, CIFAR-10 Batch 5:  Loss: 0.2274 Accuracy: 0.7718
Epoch 56, CIFAR-10 Batch 1:  Loss: 0.2951 Accuracy: 0.7694
Epoch 56, CIFAR-10 Batch 2:  Loss: 0.3403 Accuracy: 0.7468
Epoch 56, CIFAR-10 Batch 3:  Loss: 0.1810 Accuracy: 0.7836
Epoch 56, CIFAR-10 Batch 4:  Loss: 0.2145 Accuracy: 0.7850
Epoch 56, CIFAR-10 Batch 5:  Loss: 0.3042 Accuracy: 0.7594
Epoch 57, CIFAR-10 Batch 1:  Loss: 0.2884 Accuracy: 0.7658
Epoch 57, CIFAR-10 Batch 2:  Loss: 0.2674 Accuracy: 0.7804
Epoch 57, CIFAR-10 Batch 3:  Loss: 0.1921 Accuracy: 0.7710
Epoch 57, CIFAR-10 Batch 4:  Loss: 0.2170 Accuracy: 0.7668
Epoch 57, CIFAR-10 Batch 5:  Loss: 0.2518 Accuracy: 0.7892
Epoch 58, CIFAR-10 Batch 1:  Loss: 0.2383 Accuracy: 0.7792
Epoch 58, CIFAR-10 Batch 2:  Loss: 0.2831 Accuracy: 0.7554
Epoch 58, CIFAR-10 Batch 3:  Loss: 0.2315 Accuracy: 0.7632
Epoch 58, CIFAR-10 Batch 4:  Loss: 0.2486 Accuracy: 0.7730
Epoch 58, CIFAR-10 Batch 5:  Loss: 0.2755 Accuracy: 0.7628
Epoch 59, CIFAR-10 Batch 1:  Loss: 0.2333 Accuracy: 0.7958
Epoch 59, CIFAR-10 Batch 2:  Loss: 0.2962 Accuracy: 0.7600
Epoch 59, CIFAR-10 Batch 3:  Loss: 0.1798 Accuracy: 0.7806
Epoch 59, CIFAR-10 Batch 4:  Loss: 0.2057 Accuracy: 0.7830
Epoch 59, CIFAR-10 Batch 5:  Loss: 0.3137 Accuracy: 0.7486
Epoch 60, CIFAR-10 Batch 1:  Loss: 0.2201 Accuracy: 0.7898
Epoch 60, CIFAR-10 Batch 2:  Loss: 0.2238 Accuracy: 0.7750
Epoch 60, CIFAR-10 Batch 3:  Loss: 0.1981 Accuracy: 0.7818
Epoch 60, CIFAR-10 Batch 4:  Loss: 0.2792 Accuracy: 0.7504
Epoch 60, CIFAR-10 Batch 5:  Loss: 0.2755 Accuracy: 0.7642
Epoch 61, CIFAR-10 Batch 1:  Loss: 0.2574 Accuracy: 0.7784
Epoch 61, CIFAR-10 Batch 2:  Loss: 0.2490 Accuracy: 0.7576
Epoch 61, CIFAR-10 Batch 3:  Loss: 0.1884 Accuracy: 0.7766
Epoch 61, CIFAR-10 Batch 4:  Loss: 0.1939 Accuracy: 0.7762
Epoch 61, CIFAR-10 Batch 5:  Loss: 0.2356 Accuracy: 0.7808
Epoch 62, CIFAR-10 Batch 1:  Loss: 0.2343 Accuracy: 0.7926
Epoch 62, CIFAR-10 Batch 2:  Loss: 0.3334 Accuracy: 0.7350
Epoch 62, CIFAR-10 Batch 3:  Loss: 0.1817 Accuracy: 0.7824
Epoch 62, CIFAR-10 Batch 4:  Loss: 0.2095 Accuracy: 0.7612
Epoch 62, CIFAR-10 Batch 5:  Loss: 0.2359 Accuracy: 0.7862
Epoch 63, CIFAR-10 Batch 1:  Loss: 0.2654 Accuracy: 0.7874
Epoch 63, CIFAR-10 Batch 2:  Loss: 0.3100 Accuracy: 0.7570
Epoch 63, CIFAR-10 Batch 3:  Loss: 0.2000 Accuracy: 0.7798
Epoch 63, CIFAR-10 Batch 4:  Loss: 0.2135 Accuracy: 0.7792
Epoch 63, CIFAR-10 Batch 5:  Loss: 0.3026 Accuracy: 0.7548
Epoch 64, CIFAR-10 Batch 1:  Loss: 0.2528 Accuracy: 0.7826
Epoch 64, CIFAR-10 Batch 2:  Loss: 0.2637 Accuracy: 0.7676
Epoch 64, CIFAR-10 Batch 3:  Loss: 0.2186 Accuracy: 0.7760
Epoch 64, CIFAR-10 Batch 4:  Loss: 0.1635 Accuracy: 0.7900
Epoch 64, CIFAR-10 Batch 5:  Loss: 0.2607 Accuracy: 0.7766
Epoch 65, CIFAR-10 Batch 1:  Loss: 0.2484 Accuracy: 0.7794
Epoch 65, CIFAR-10 Batch 2:  Loss: 0.1882 Accuracy: 0.7842
Epoch 65, CIFAR-10 Batch 3:  Loss: 0.1645 Accuracy: 0.7824
Epoch 65, CIFAR-10 Batch 4:  Loss: 0.1705 Accuracy: 0.7884
Epoch 65, CIFAR-10 Batch 5:  Loss: 0.2601 Accuracy: 0.7752
Epoch 66, CIFAR-10 Batch 1:  Loss: 0.2260 Accuracy: 0.7830
Epoch 66, CIFAR-10 Batch 2:  Loss: 0.2603 Accuracy: 0.7730
Epoch 66, CIFAR-10 Batch 3:  Loss: 0.1915 Accuracy: 0.7780
Epoch 66, CIFAR-10 Batch 4:  Loss: 0.3277 Accuracy: 0.7510
Epoch 66, CIFAR-10 Batch 5:  Loss: 0.2697 Accuracy: 0.7752
Epoch 67, CIFAR-10 Batch 1:  Loss: 0.2551 Accuracy: 0.7844
Epoch 67, CIFAR-10 Batch 2:  Loss: 0.2440 Accuracy: 0.7720
Epoch 67, CIFAR-10 Batch 3:  Loss: 0.1791 Accuracy: 0.7800
Epoch 67, CIFAR-10 Batch 4:  Loss: 0.1722 Accuracy: 0.7726
Epoch 67, CIFAR-10 Batch 5:  Loss: 0.2186 Accuracy: 0.7840
Epoch 68, CIFAR-10 Batch 1:  Loss: 0.2175 Accuracy: 0.7900
Epoch 68, CIFAR-10 Batch 2:  Loss: 0.2495 Accuracy: 0.7510
Epoch 68, CIFAR-10 Batch 3:  Loss: 0.1426 Accuracy: 0.7798
Epoch 68, CIFAR-10 Batch 4:  Loss: 0.2684 Accuracy: 0.7616
Epoch 68, CIFAR-10 Batch 5:  Loss: 0.1852 Accuracy: 0.7928
Epoch 69, CIFAR-10 Batch 1:  Loss: 0.2183 Accuracy: 0.7860
Epoch 69, CIFAR-10 Batch 2:  Loss: 0.3440 Accuracy: 0.7412
Epoch 69, CIFAR-10 Batch 3:  Loss: 0.1636 Accuracy: 0.7816
Epoch 69, CIFAR-10 Batch 4:  Loss: 0.1744 Accuracy: 0.7842
Epoch 69, CIFAR-10 Batch 5:  Loss: 0.2064 Accuracy: 0.7742
Epoch 70, CIFAR-10 Batch 1:  Loss: 0.2010 Accuracy: 0.7926
Epoch 70, CIFAR-10 Batch 2:  Loss: 0.2566 Accuracy: 0.7720
Epoch 70, CIFAR-10 Batch 3:  Loss: 0.1592 Accuracy: 0.7802
Epoch 70, CIFAR-10 Batch 4:  Loss: 0.1877 Accuracy: 0.7862
Epoch 70, CIFAR-10 Batch 5:  Loss: 0.1922 Accuracy: 0.7680
Epoch 71, CIFAR-10 Batch 1:  Loss: 0.1795 Accuracy: 0.7926
Epoch 71, CIFAR-10 Batch 2:  Loss: 0.2239 Accuracy: 0.7724
Epoch 71, CIFAR-10 Batch 3:  Loss: 0.1505 Accuracy: 0.7886
Epoch 71, CIFAR-10 Batch 4:  Loss: 0.2080 Accuracy: 0.7698
Epoch 71, CIFAR-10 Batch 5:  Loss: 0.1783 Accuracy: 0.7746
Epoch 72, CIFAR-10 Batch 1:  Loss: 0.2619 Accuracy: 0.7806
Epoch 72, CIFAR-10 Batch 2:  Loss: 0.1514 Accuracy: 0.7862
Epoch 72, CIFAR-10 Batch 3:  Loss: 0.1488 Accuracy: 0.7874
Epoch 72, CIFAR-10 Batch 4:  Loss: 0.2200 Accuracy: 0.7806
Epoch 72, CIFAR-10 Batch 5:  Loss: 0.1892 Accuracy: 0.7672
Epoch 73, CIFAR-10 Batch 1:  Loss: 0.2338 Accuracy: 0.7884
Epoch 73, CIFAR-10 Batch 2:  Loss: 0.1864 Accuracy: 0.7856
Epoch 73, CIFAR-10 Batch 3:  Loss: 0.1493 Accuracy: 0.7812
Epoch 73, CIFAR-10 Batch 4:  Loss: 0.1876 Accuracy: 0.7778
Epoch 73, CIFAR-10 Batch 5:  Loss: 0.1603 Accuracy: 0.7786
Epoch 74, CIFAR-10 Batch 1:  Loss: 0.1754 Accuracy: 0.8002
Epoch 74, CIFAR-10 Batch 2:  Loss: 0.3595 Accuracy: 0.7656
Epoch 74, CIFAR-10 Batch 3:  Loss: 0.1570 Accuracy: 0.7756
Epoch 74, CIFAR-10 Batch 4:  Loss: 0.2290 Accuracy: 0.7806
Epoch 74, CIFAR-10 Batch 5:  Loss: 0.2107 Accuracy: 0.7598
Epoch 75, CIFAR-10 Batch 1:  Loss: 0.1622 Accuracy: 0.8036
Epoch 75, CIFAR-10 Batch 2:  Loss: 0.2064 Accuracy: 0.7762
Epoch 75, CIFAR-10 Batch 3:  Loss: 0.1033 Accuracy: 0.7940
Epoch 75, CIFAR-10 Batch 4:  Loss: 0.1807 Accuracy: 0.7758
Epoch 75, CIFAR-10 Batch 5:  Loss: 0.1870 Accuracy: 0.7864
Epoch 76, CIFAR-10 Batch 1:  Loss: 0.1635 Accuracy: 0.7968
Epoch 76, CIFAR-10 Batch 2:  Loss: 0.2108 Accuracy: 0.7980
Epoch 76, CIFAR-10 Batch 3:  Loss: 0.1323 Accuracy: 0.7894
Epoch 76, CIFAR-10 Batch 4:  Loss: 0.2962 Accuracy: 0.7454
Epoch 76, CIFAR-10 Batch 5:  Loss: 0.1669 Accuracy: 0.7772
Epoch 77, CIFAR-10 Batch 1:  Loss: 0.1335 Accuracy: 0.7950
Epoch 77, CIFAR-10 Batch 2:  Loss: 0.2360 Accuracy: 0.7780
Epoch 77, CIFAR-10 Batch 3:  Loss: 0.1434 Accuracy: 0.7880
Epoch 77, CIFAR-10 Batch 4:  Loss: 0.1853 Accuracy: 0.7884
Epoch 77, CIFAR-10 Batch 5:  Loss: 0.1815 Accuracy: 0.7968
Epoch 78, CIFAR-10 Batch 1:  Loss: 0.1258 Accuracy: 0.7974
Epoch 78, CIFAR-10 Batch 2:  Loss: 0.2947 Accuracy: 0.7640
Epoch 78, CIFAR-10 Batch 3:  Loss: 0.1014 Accuracy: 0.7970
Epoch 78, CIFAR-10 Batch 4:  Loss: 0.1411 Accuracy: 0.7940
Epoch 78, CIFAR-10 Batch 5:  Loss: 0.1531 Accuracy: 0.7888
Epoch 79, CIFAR-10 Batch 1:  Loss: 0.1449 Accuracy: 0.8040
Epoch 79, CIFAR-10 Batch 2:  Loss: 0.2323 Accuracy: 0.7800
Epoch 79, CIFAR-10 Batch 3:  Loss: 0.1313 Accuracy: 0.7974
Epoch 79, CIFAR-10 Batch 4:  Loss: 0.1834 Accuracy: 0.7778
Epoch 79, CIFAR-10 Batch 5:  Loss: 0.1888 Accuracy: 0.7722
Epoch 80, CIFAR-10 Batch 1:  Loss: 0.1487 Accuracy: 0.7974
Epoch 80, CIFAR-10 Batch 2:  Loss: 0.1959 Accuracy: 0.7980
Epoch 80, CIFAR-10 Batch 3:  Loss: 0.0917 Accuracy: 0.7958
Epoch 80, CIFAR-10 Batch 4:  Loss: 0.1517 Accuracy: 0.7900
Epoch 80, CIFAR-10 Batch 5:  Loss: 0.1888 Accuracy: 0.7684
Epoch 81, CIFAR-10 Batch 1:  Loss: 0.1392 Accuracy: 0.7970
Epoch 81, CIFAR-10 Batch 2:  Loss: 0.2158 Accuracy: 0.7686
Epoch 81, CIFAR-10 Batch 3:  Loss: 0.0954 Accuracy: 0.7908
Epoch 81, CIFAR-10 Batch 4:  Loss: 0.1366 Accuracy: 0.7808
Epoch 81, CIFAR-10 Batch 5:  Loss: 0.1450 Accuracy: 0.7726
Epoch 82, CIFAR-10 Batch 1:  Loss: 0.1794 Accuracy: 0.7808
Epoch 82, CIFAR-10 Batch 2:  Loss: 0.2211 Accuracy: 0.7808
Epoch 82, CIFAR-10 Batch 3:  Loss: 0.1187 Accuracy: 0.7898
Epoch 82, CIFAR-10 Batch 4:  Loss: 0.1217 Accuracy: 0.7854
Epoch 82, CIFAR-10 Batch 5:  Loss: 0.1527 Accuracy: 0.7974
Epoch 83, CIFAR-10 Batch 1:  Loss: 0.1929 Accuracy: 0.7808
Epoch 83, CIFAR-10 Batch 2:  Loss: 0.2924 Accuracy: 0.7734
Epoch 83, CIFAR-10 Batch 3:  Loss: 0.1117 Accuracy: 0.7836
Epoch 83, CIFAR-10 Batch 4:  Loss: 0.1438 Accuracy: 0.7788
Epoch 83, CIFAR-10 Batch 5:  Loss: 0.1404 Accuracy: 0.7820
Epoch 84, CIFAR-10 Batch 1:  Loss: 0.1202 Accuracy: 0.7968
Epoch 84, CIFAR-10 Batch 2:  Loss: 0.1920 Accuracy: 0.7740
Epoch 84, CIFAR-10 Batch 3:  Loss: 0.1085 Accuracy: 0.7820
Epoch 84, CIFAR-10 Batch 4:  Loss: 0.2698 Accuracy: 0.7524
Epoch 84, CIFAR-10 Batch 5:  Loss: 0.1950 Accuracy: 0.7684
Epoch 85, CIFAR-10 Batch 1:  Loss: 0.1272 Accuracy: 0.7976
Epoch 85, CIFAR-10 Batch 2:  Loss: 0.1620 Accuracy: 0.7878
Epoch 85, CIFAR-10 Batch 3:  Loss: 0.1265 Accuracy: 0.7946
Epoch 85, CIFAR-10 Batch 4:  Loss: 0.1759 Accuracy: 0.7804
Epoch 85, CIFAR-10 Batch 5:  Loss: 0.1856 Accuracy: 0.7720
Epoch 86, CIFAR-10 Batch 1:  Loss: 0.1135 Accuracy: 0.7922
Epoch 86, CIFAR-10 Batch 2:  Loss: 0.1171 Accuracy: 0.7912
Epoch 86, CIFAR-10 Batch 3:  Loss: 0.1472 Accuracy: 0.7764
Epoch 86, CIFAR-10 Batch 4:  Loss: 0.1904 Accuracy: 0.7844
Epoch 86, CIFAR-10 Batch 5:  Loss: 0.1731 Accuracy: 0.7856
Epoch 87, CIFAR-10 Batch 1:  Loss: 0.1536 Accuracy: 0.7910
Epoch 87, CIFAR-10 Batch 2:  Loss: 0.3383 Accuracy: 0.7642
Epoch 87, CIFAR-10 Batch 3:  Loss: 0.1288 Accuracy: 0.7812
Epoch 87, CIFAR-10 Batch 4:  Loss: 0.1544 Accuracy: 0.7934
Epoch 87, CIFAR-10 Batch 5:  Loss: 0.1477 Accuracy: 0.7800
Epoch 88, CIFAR-10 Batch 1:  Loss: 0.1224 Accuracy: 0.7976
Epoch 88, CIFAR-10 Batch 2:  Loss: 0.1642 Accuracy: 0.7954
Epoch 88, CIFAR-10 Batch 3:  Loss: 0.1710 Accuracy: 0.7674
Epoch 88, CIFAR-10 Batch 4:  Loss: 0.1429 Accuracy: 0.7988
Epoch 88, CIFAR-10 Batch 5:  Loss: 0.1591 Accuracy: 0.7764
Epoch 89, CIFAR-10 Batch 1:  Loss: 0.1344 Accuracy: 0.7978
Epoch 89, CIFAR-10 Batch 2:  Loss: 0.1679 Accuracy: 0.7954
Epoch 89, CIFAR-10 Batch 3:  Loss: 0.1350 Accuracy: 0.7762
Epoch 89, CIFAR-10 Batch 4:  Loss: 0.1534 Accuracy: 0.7892
Epoch 89, CIFAR-10 Batch 5:  Loss: 0.1498 Accuracy: 0.7732
Epoch 90, CIFAR-10 Batch 1:  Loss: 0.0992 Accuracy: 0.8042
Epoch 90, CIFAR-10 Batch 2:  Loss: 0.1749 Accuracy: 0.7946
Epoch 90, CIFAR-10 Batch 3:  Loss: 0.1035 Accuracy: 0.7924
Epoch 90, CIFAR-10 Batch 4:  Loss: 0.1504 Accuracy: 0.7994
Epoch 90, CIFAR-10 Batch 5:  Loss: 0.1795 Accuracy: 0.7638
Epoch 91, CIFAR-10 Batch 1:  Loss: 0.1520 Accuracy: 0.7862
Epoch 91, CIFAR-10 Batch 2:  Loss: 0.2888 Accuracy: 0.7520
Epoch 91, CIFAR-10 Batch 3:  Loss: 0.1588 Accuracy: 0.7558
Epoch 91, CIFAR-10 Batch 4:  Loss: 0.1664 Accuracy: 0.7760
Epoch 91, CIFAR-10 Batch 5:  Loss: 0.1491 Accuracy: 0.7670
Epoch 92, CIFAR-10 Batch 1:  Loss: 0.0892 Accuracy: 0.8040
Epoch 92, CIFAR-10 Batch 2:  Loss: 0.2005 Accuracy: 0.7770
Epoch 92, CIFAR-10 Batch 3:  Loss: 0.1064 Accuracy: 0.7818
Epoch 92, CIFAR-10 Batch 4:  Loss: 0.1695 Accuracy: 0.7834
Epoch 92, CIFAR-10 Batch 5:  Loss: 0.1903 Accuracy: 0.7612
Epoch 93, CIFAR-10 Batch 1:  Loss: 0.1117 Accuracy: 0.8054
Epoch 93, CIFAR-10 Batch 2:  Loss: 0.1942 Accuracy: 0.8026
Epoch 93, CIFAR-10 Batch 3:  Loss: 0.1192 Accuracy: 0.7930
Epoch 93, CIFAR-10 Batch 4:  Loss: 0.0932 Accuracy: 0.8006
Epoch 93, CIFAR-10 Batch 5:  Loss: 0.1382 Accuracy: 0.7816
Epoch 94, CIFAR-10 Batch 1:  Loss: 0.0974 Accuracy: 0.8028
Epoch 94, CIFAR-10 Batch 2:  Loss: 0.2104 Accuracy: 0.7888
Epoch 94, CIFAR-10 Batch 3:  Loss: 0.0862 Accuracy: 0.7926
Epoch 94, CIFAR-10 Batch 4:  Loss: 0.1150 Accuracy: 0.7972
Epoch 94, CIFAR-10 Batch 5:  Loss: 0.1109 Accuracy: 0.7950
Epoch 95, CIFAR-10 Batch 1:  Loss: 0.0732 Accuracy: 0.8030
Epoch 95, CIFAR-10 Batch 2:  Loss: 0.1463 Accuracy: 0.7982
Epoch 95, CIFAR-10 Batch 3:  Loss: 0.1428 Accuracy: 0.7850
Epoch 95, CIFAR-10 Batch 4:  Loss: 0.1353 Accuracy: 0.7928
Epoch 95, CIFAR-10 Batch 5:  Loss: 0.1090 Accuracy: 0.7924
Epoch 96, CIFAR-10 Batch 1:  Loss: 0.0720 Accuracy: 0.8048
Epoch 96, CIFAR-10 Batch 2:  Loss: 0.1102 Accuracy: 0.7990
Epoch 96, CIFAR-10 Batch 3:  Loss: 0.0865 Accuracy: 0.8028
Epoch 96, CIFAR-10 Batch 4:  Loss: 0.1282 Accuracy: 0.7920
Epoch 96, CIFAR-10 Batch 5:  Loss: 0.1164 Accuracy: 0.7912
Epoch 97, CIFAR-10 Batch 1:  Loss: 0.1014 Accuracy: 0.7912
Epoch 97, CIFAR-10 Batch 2:  Loss: 0.1291 Accuracy: 0.7958
Epoch 97, CIFAR-10 Batch 3:  Loss: 0.0814 Accuracy: 0.8014
Epoch 97, CIFAR-10 Batch 4:  Loss: 0.1019 Accuracy: 0.7980
Epoch 97, CIFAR-10 Batch 5:  Loss: 0.1268 Accuracy: 0.7920
Epoch 98, CIFAR-10 Batch 1:  Loss: 0.0837 Accuracy: 0.7988
Epoch 98, CIFAR-10 Batch 2:  Loss: 0.1363 Accuracy: 0.7896
Epoch 98, CIFAR-10 Batch 3:  Loss: 0.0744 Accuracy: 0.8028
Epoch 98, CIFAR-10 Batch 4:  Loss: 0.1629 Accuracy: 0.7948
Epoch 98, CIFAR-10 Batch 5:  Loss: 0.1063 Accuracy: 0.7902
Epoch 99, CIFAR-10 Batch 1:  Loss: 0.1060 Accuracy: 0.7936
Epoch 99, CIFAR-10 Batch 2:  Loss: 0.1079 Accuracy: 0.8010
Epoch 99, CIFAR-10 Batch 3:  Loss: 0.0980 Accuracy: 0.7948
Epoch 99, CIFAR-10 Batch 4:  Loss: 0.0954 Accuracy: 0.7986
Epoch 99, CIFAR-10 Batch 5:  Loss: 0.1054 Accuracy: 0.7884
Epoch 100, CIFAR-10 Batch 1:  Loss: 0.1235 Accuracy: 0.7900
Epoch 100, CIFAR-10 Batch 2:  Loss: 0.2820 Accuracy: 0.7626
Epoch 100, CIFAR-10 Batch 3:  Loss: 0.0753 Accuracy: 0.7966
Epoch 100, CIFAR-10 Batch 4:  Loss: 0.1661 Accuracy: 0.7916
Epoch 100, CIFAR-10 Batch 5:  Loss: 0.0861 Accuracy: 0.7882
Epoch 101, CIFAR-10 Batch 1:  Loss: 0.0907 Accuracy: 0.8092
Epoch 101, CIFAR-10 Batch 2:  Loss: 0.1732 Accuracy: 0.7762
Epoch 101, CIFAR-10 Batch 3:  Loss: 0.0798 Accuracy: 0.7942
Epoch 101, CIFAR-10 Batch 4:  Loss: 0.1809 Accuracy: 0.7762
Epoch 101, CIFAR-10 Batch 5:  Loss: 0.2050 Accuracy: 0.7522
Epoch 102, CIFAR-10 Batch 1:  Loss: 0.0810 Accuracy: 0.8052
Epoch 102, CIFAR-10 Batch 2:  Loss: 0.1305 Accuracy: 0.7838
Epoch 102, CIFAR-10 Batch 3:  Loss: 0.0809 Accuracy: 0.8042
Epoch 102, CIFAR-10 Batch 4:  Loss: 0.1608 Accuracy: 0.7862
Epoch 102, CIFAR-10 Batch 5:  Loss: 0.1086 Accuracy: 0.7842
Epoch 103, CIFAR-10 Batch 1:  Loss: 0.1055 Accuracy: 0.7996
Epoch 103, CIFAR-10 Batch 2:  Loss: 0.1220 Accuracy: 0.7960
Epoch 103, CIFAR-10 Batch 3:  Loss: 0.1136 Accuracy: 0.7860
Epoch 103, CIFAR-10 Batch 4:  Loss: 0.1569 Accuracy: 0.7962
Epoch 103, CIFAR-10 Batch 5:  Loss: 0.1045 Accuracy: 0.7888
Epoch 104, CIFAR-10 Batch 1:  Loss: 0.1448 Accuracy: 0.7954
Epoch 104, CIFAR-10 Batch 2:  Loss: 0.1258 Accuracy: 0.7974
Epoch 104, CIFAR-10 Batch 3:  Loss: 0.0888 Accuracy: 0.7890
Epoch 104, CIFAR-10 Batch 4:  Loss: 0.1362 Accuracy: 0.7998
Epoch 104, CIFAR-10 Batch 5:  Loss: 0.0801 Accuracy: 0.7922
Epoch 105, CIFAR-10 Batch 1:  Loss: 0.0985 Accuracy: 0.8030
Epoch 105, CIFAR-10 Batch 2:  Loss: 0.1245 Accuracy: 0.7868
Epoch 105, CIFAR-10 Batch 3:  Loss: 0.1200 Accuracy: 0.7858
Epoch 105, CIFAR-10 Batch 4:  Loss: 0.1354 Accuracy: 0.7940
Epoch 105, CIFAR-10 Batch 5:  Loss: 0.0956 Accuracy: 0.7796
Epoch 106, CIFAR-10 Batch 1:  Loss: 0.1035 Accuracy: 0.8052
Epoch 106, CIFAR-10 Batch 2:  Loss: 0.0942 Accuracy: 0.8114
Epoch 106, CIFAR-10 Batch 3:  Loss: 0.0724 Accuracy: 0.8052
Epoch 106, CIFAR-10 Batch 4:  Loss: 0.0935 Accuracy: 0.8068
Epoch 106, CIFAR-10 Batch 5:  Loss: 0.1139 Accuracy: 0.7810
Epoch 107, CIFAR-10 Batch 1:  Loss: 0.1066 Accuracy: 0.8042
Epoch 107, CIFAR-10 Batch 2:  Loss: 0.2231 Accuracy: 0.7796
Epoch 107, CIFAR-10 Batch 3:  Loss: 0.0746 Accuracy: 0.8014
Epoch 107, CIFAR-10 Batch 4:  Loss: 0.0821 Accuracy: 0.7928
Epoch 107, CIFAR-10 Batch 5:  Loss: 0.1028 Accuracy: 0.7836
Epoch 108, CIFAR-10 Batch 1:  Loss: 0.0706 Accuracy: 0.8090
Epoch 108, CIFAR-10 Batch 2:  Loss: 0.1134 Accuracy: 0.8004
Epoch 108, CIFAR-10 Batch 3:  Loss: 0.1207 Accuracy: 0.7826
Epoch 108, CIFAR-10 Batch 4:  Loss: 0.0979 Accuracy: 0.8036
Epoch 108, CIFAR-10 Batch 5:  Loss: 0.0770 Accuracy: 0.7872
Epoch 109, CIFAR-10 Batch 1:  Loss: 0.0838 Accuracy: 0.7982
Epoch 109, CIFAR-10 Batch 2:  Loss: 0.0967 Accuracy: 0.7966
Epoch 109, CIFAR-10 Batch 3:  Loss: 0.1000 Accuracy: 0.7956
Epoch 109, CIFAR-10 Batch 4:  Loss: 0.1234 Accuracy: 0.7898
Epoch 109, CIFAR-10 Batch 5:  Loss: 0.0972 Accuracy: 0.7880
Epoch 110, CIFAR-10 Batch 1:  Loss: 0.0765 Accuracy: 0.8034
Epoch 110, CIFAR-10 Batch 2:  Loss: 0.1914 Accuracy: 0.7880
Epoch 110, CIFAR-10 Batch 3:  Loss: 0.0967 Accuracy: 0.7914
Epoch 110, CIFAR-10 Batch 4:  Loss: 0.0892 Accuracy: 0.8016
Epoch 110, CIFAR-10 Batch 5:  Loss: 0.1224 Accuracy: 0.7656
Epoch 111, CIFAR-10 Batch 1:  Loss: 0.0853 Accuracy: 0.8048
Epoch 111, CIFAR-10 Batch 2:  Loss: 0.1268 Accuracy: 0.7990
Epoch 111, CIFAR-10 Batch 3:  Loss: 0.0800 Accuracy: 0.7932
Epoch 111, CIFAR-10 Batch 4:  Loss: 0.1123 Accuracy: 0.7942
Epoch 111, CIFAR-10 Batch 5:  Loss: 0.0782 Accuracy: 0.7964
Epoch 112, CIFAR-10 Batch 1:  Loss: 0.0619 Accuracy: 0.7990
Epoch 112, CIFAR-10 Batch 2:  Loss: 0.0931 Accuracy: 0.7870
Epoch 112, CIFAR-10 Batch 3:  Loss: 0.0805 Accuracy: 0.7992
Epoch 112, CIFAR-10 Batch 4:  Loss: 0.0866 Accuracy: 0.8088
Epoch 112, CIFAR-10 Batch 5:  Loss: 0.0831 Accuracy: 0.8022
Epoch 113, CIFAR-10 Batch 1:  Loss: 0.0708 Accuracy: 0.8118
Epoch 113, CIFAR-10 Batch 2:  Loss: 0.1265 Accuracy: 0.7814
Epoch 113, CIFAR-10 Batch 3:  Loss: 0.0571 Accuracy: 0.8064
Epoch 113, CIFAR-10 Batch 4:  Loss: 0.0709 Accuracy: 0.8070
Epoch 113, CIFAR-10 Batch 5:  Loss: 0.0980 Accuracy: 0.7852
Epoch 114, CIFAR-10 Batch 1:  Loss: 0.1238 Accuracy: 0.8064
Epoch 114, CIFAR-10 Batch 2:  Loss: 0.1130 Accuracy: 0.8064
Epoch 114, CIFAR-10 Batch 3:  Loss: 0.0705 Accuracy: 0.8012
Epoch 114, CIFAR-10 Batch 4:  Loss: 0.1281 Accuracy: 0.8006
Epoch 114, CIFAR-10 Batch 5:  Loss: 0.1003 Accuracy: 0.7940
Epoch 115, CIFAR-10 Batch 1:  Loss: 0.2878 Accuracy: 0.7700
Epoch 115, CIFAR-10 Batch 2:  Loss: 0.1255 Accuracy: 0.8024
Epoch 115, CIFAR-10 Batch 3:  Loss: 0.0637 Accuracy: 0.8006
Epoch 115, CIFAR-10 Batch 4:  Loss: 0.0744 Accuracy: 0.8072
Epoch 115, CIFAR-10 Batch 5:  Loss: 0.1021 Accuracy: 0.7868
Epoch 116, CIFAR-10 Batch 1:  Loss: 0.1043 Accuracy: 0.7908
Epoch 116, CIFAR-10 Batch 2:  Loss: 0.1340 Accuracy: 0.7984
Epoch 116, CIFAR-10 Batch 3:  Loss: 0.0839 Accuracy: 0.7972
Epoch 116, CIFAR-10 Batch 4:  Loss: 0.0698 Accuracy: 0.8160
Epoch 116, CIFAR-10 Batch 5:  Loss: 0.0765 Accuracy: 0.7948
Epoch 117, CIFAR-10 Batch 1:  Loss: 0.0484 Accuracy: 0.7932
Epoch 117, CIFAR-10 Batch 2:  Loss: 0.0914 Accuracy: 0.8020
Epoch 117, CIFAR-10 Batch 3:  Loss: 0.0499 Accuracy: 0.8050
Epoch 117, CIFAR-10 Batch 4:  Loss: 0.1053 Accuracy: 0.7968
Epoch 117, CIFAR-10 Batch 5:  Loss: 0.0557 Accuracy: 0.7914
Epoch 118, CIFAR-10 Batch 1:  Loss: 0.0575 Accuracy: 0.7968
Epoch 118, CIFAR-10 Batch 2:  Loss: 0.1275 Accuracy: 0.7982
Epoch 118, CIFAR-10 Batch 3:  Loss: 0.0436 Accuracy: 0.8094
Epoch 118, CIFAR-10 Batch 4:  Loss: 0.0990 Accuracy: 0.7894
Epoch 118, CIFAR-10 Batch 5:  Loss: 0.0818 Accuracy: 0.7902
Epoch 119, CIFAR-10 Batch 1:  Loss: 0.1306 Accuracy: 0.7936
Epoch 119, CIFAR-10 Batch 2:  Loss: 0.0768 Accuracy: 0.8054
Epoch 119, CIFAR-10 Batch 3:  Loss: 0.0710 Accuracy: 0.7994
Epoch 119, CIFAR-10 Batch 4:  Loss: 0.0903 Accuracy: 0.8016
Epoch 119, CIFAR-10 Batch 5:  Loss: 0.0859 Accuracy: 0.7890
Epoch 120, CIFAR-10 Batch 1:  Loss: 0.0761 Accuracy: 0.7934
Epoch 120, CIFAR-10 Batch 2:  Loss: 0.0918 Accuracy: 0.7994
Epoch 120, CIFAR-10 Batch 3:  Loss: 0.0624 Accuracy: 0.8030
Epoch 120, CIFAR-10 Batch 4:  Loss: 0.0677 Accuracy: 0.8062
Epoch 120, CIFAR-10 Batch 5:  Loss: 0.0746 Accuracy: 0.7896
Epoch 121, CIFAR-10 Batch 1:  Loss: 0.0528 Accuracy: 0.8068
Epoch 121, CIFAR-10 Batch 2:  Loss: 0.0982 Accuracy: 0.7932
Epoch 121, CIFAR-10 Batch 3:  Loss: 0.0483 Accuracy: 0.8030
Epoch 121, CIFAR-10 Batch 4:  Loss: 0.0859 Accuracy: 0.8108
Epoch 121, CIFAR-10 Batch 5:  Loss: 0.0608 Accuracy: 0.8086
Epoch 122, CIFAR-10 Batch 1:  Loss: 0.0497 Accuracy: 0.8060
Epoch 122, CIFAR-10 Batch 2:  Loss: 0.1312 Accuracy: 0.7942
Epoch 122, CIFAR-10 Batch 3:  Loss: 0.0721 Accuracy: 0.7994
Epoch 122, CIFAR-10 Batch 4:  Loss: 0.0655 Accuracy: 0.8074
Epoch 122, CIFAR-10 Batch 5:  Loss: 0.0647 Accuracy: 0.7952
Epoch 123, CIFAR-10 Batch 1:  Loss: 0.0636 Accuracy: 0.8078
Epoch 123, CIFAR-10 Batch 2:  Loss: 0.0940 Accuracy: 0.7914
Epoch 123, CIFAR-10 Batch 3:  Loss: 0.0539 Accuracy: 0.8104
Epoch 123, CIFAR-10 Batch 4:  Loss: 0.0842 Accuracy: 0.7950
Epoch 123, CIFAR-10 Batch 5:  Loss: 0.0725 Accuracy: 0.8054
Epoch 124, CIFAR-10 Batch 1:  Loss: 0.0639 Accuracy: 0.8040
Epoch 124, CIFAR-10 Batch 2:  Loss: 0.0823 Accuracy: 0.7966
Epoch 124, CIFAR-10 Batch 3:  Loss: 0.0572 Accuracy: 0.8064
Epoch 124, CIFAR-10 Batch 4:  Loss: 0.0935 Accuracy: 0.8010
Epoch 124, CIFAR-10 Batch 5:  Loss: 0.1143 Accuracy: 0.7756
Epoch 125, CIFAR-10 Batch 1:  Loss: 0.0741 Accuracy: 0.8042
Epoch 125, CIFAR-10 Batch 2:  Loss: 0.0998 Accuracy: 0.7998
Epoch 125, CIFAR-10 Batch 3:  Loss: 0.0774 Accuracy: 0.7944
Epoch 125, CIFAR-10 Batch 4:  Loss: 0.0793 Accuracy: 0.8146
Epoch 125, CIFAR-10 Batch 5:  Loss: 0.0812 Accuracy: 0.7932
Epoch 126, CIFAR-10 Batch 1:  Loss: 0.0457 Accuracy: 0.8168
Epoch 126, CIFAR-10 Batch 2:  Loss: 0.0890 Accuracy: 0.8078
Epoch 126, CIFAR-10 Batch 3:  Loss: 0.0453 Accuracy: 0.8024
Epoch 126, CIFAR-10 Batch 4:  Loss: 0.1018 Accuracy: 0.8046
Epoch 126, CIFAR-10 Batch 5:  Loss: 0.0660 Accuracy: 0.7948
Epoch 127, CIFAR-10 Batch 1:  Loss: 0.0463 Accuracy: 0.8184
Epoch 127, CIFAR-10 Batch 2:  Loss: 0.1228 Accuracy: 0.7766
Epoch 127, CIFAR-10 Batch 3:  Loss: 0.0758 Accuracy: 0.7976
Epoch 127, CIFAR-10 Batch 4:  Loss: 0.0815 Accuracy: 0.8132
Epoch 127, CIFAR-10 Batch 5:  Loss: 0.0633 Accuracy: 0.7960
Epoch 128, CIFAR-10 Batch 1:  Loss: 0.0562 Accuracy: 0.8028
Epoch 128, CIFAR-10 Batch 2:  Loss: 0.1388 Accuracy: 0.8016
Epoch 128, CIFAR-10 Batch 3:  Loss: 0.0515 Accuracy: 0.8068
Epoch 128, CIFAR-10 Batch 4:  Loss: 0.0778 Accuracy: 0.8140
Epoch 128, CIFAR-10 Batch 5:  Loss: 0.0619 Accuracy: 0.8032
Epoch 129, CIFAR-10 Batch 1:  Loss: 0.0422 Accuracy: 0.8078
Epoch 129, CIFAR-10 Batch 2:  Loss: 0.0970 Accuracy: 0.7990
Epoch 129, CIFAR-10 Batch 3:  Loss: 0.0580 Accuracy: 0.8086
Epoch 129, CIFAR-10 Batch 4:  Loss: 0.1259 Accuracy: 0.8012
Epoch 129, CIFAR-10 Batch 5:  Loss: 0.0527 Accuracy: 0.8068
Epoch 130, CIFAR-10 Batch 1:  Loss: 0.0447 Accuracy: 0.8138
Epoch 130, CIFAR-10 Batch 2:  Loss: 0.0716 Accuracy: 0.8032
Epoch 130, CIFAR-10 Batch 3:  Loss: 0.0424 Accuracy: 0.8092
Epoch 130, CIFAR-10 Batch 4:  Loss: 0.0809 Accuracy: 0.8114
Epoch 130, CIFAR-10 Batch 5:  Loss: 0.0515 Accuracy: 0.7984
Epoch 131, CIFAR-10 Batch 1:  Loss: 0.0368 Accuracy: 0.8068
Epoch 131, CIFAR-10 Batch 2:  Loss: 0.0560 Accuracy: 0.8088
Epoch 131, CIFAR-10 Batch 3:  Loss: 0.0432 Accuracy: 0.8112
Epoch 131, CIFAR-10 Batch 4:  Loss: 0.0620 Accuracy: 0.8098
Epoch 131, CIFAR-10 Batch 5:  Loss: 0.0594 Accuracy: 0.7976
Epoch 132, CIFAR-10 Batch 1:  Loss: 0.0460 Accuracy: 0.8084
Epoch 132, CIFAR-10 Batch 2:  Loss: 0.1086 Accuracy: 0.7964
Epoch 132, CIFAR-10 Batch 3:  Loss: 0.0677 Accuracy: 0.8060
Epoch 132, CIFAR-10 Batch 4:  Loss: 0.1011 Accuracy: 0.7976
Epoch 132, CIFAR-10 Batch 5:  Loss: 0.0646 Accuracy: 0.7936
Epoch 133, CIFAR-10 Batch 1:  Loss: 0.0631 Accuracy: 0.8084
Epoch 133, CIFAR-10 Batch 2:  Loss: 0.1575 Accuracy: 0.7994
Epoch 133, CIFAR-10 Batch 3:  Loss: 0.0784 Accuracy: 0.7954
Epoch 133, CIFAR-10 Batch 4:  Loss: 0.0763 Accuracy: 0.8062
Epoch 133, CIFAR-10 Batch 5:  Loss: 0.0866 Accuracy: 0.7986
Epoch 134, CIFAR-10 Batch 1:  Loss: 0.0462 Accuracy: 0.8110
Epoch 134, CIFAR-10 Batch 2:  Loss: 0.0976 Accuracy: 0.8018
Epoch 134, CIFAR-10 Batch 3:  Loss: 0.0602 Accuracy: 0.8034
Epoch 134, CIFAR-10 Batch 4:  Loss: 0.0624 Accuracy: 0.8060
Epoch 134, CIFAR-10 Batch 5:  Loss: 0.0848 Accuracy: 0.7984
Epoch 135, CIFAR-10 Batch 1:  Loss: 0.0539 Accuracy: 0.8032
Epoch 135, CIFAR-10 Batch 2:  Loss: 0.0645 Accuracy: 0.8050
Epoch 135, CIFAR-10 Batch 3:  Loss: 0.0683 Accuracy: 0.8034
Epoch 135, CIFAR-10 Batch 4:  Loss: 0.0752 Accuracy: 0.8052
Epoch 135, CIFAR-10 Batch 5:  Loss: 0.0702 Accuracy: 0.8066
Epoch 136, CIFAR-10 Batch 1:  Loss: 0.0542 Accuracy: 0.8138
Epoch 136, CIFAR-10 Batch 2:  Loss: 0.0730 Accuracy: 0.8054
Epoch 136, CIFAR-10 Batch 3:  Loss: 0.0432 Accuracy: 0.8018
Epoch 136, CIFAR-10 Batch 4:  Loss: 0.0794 Accuracy: 0.8076
Epoch 136, CIFAR-10 Batch 5:  Loss: 0.0810 Accuracy: 0.7924
Epoch 137, CIFAR-10 Batch 1:  Loss: 0.0509 Accuracy: 0.8056
Epoch 137, CIFAR-10 Batch 2:  Loss: 0.0994 Accuracy: 0.8044
Epoch 137, CIFAR-10 Batch 3:  Loss: 0.0461 Accuracy: 0.8074
Epoch 137, CIFAR-10 Batch 4:  Loss: 0.0689 Accuracy: 0.8052
Epoch 137, CIFAR-10 Batch 5:  Loss: 0.0655 Accuracy: 0.8058
Epoch 138, CIFAR-10 Batch 1:  Loss: 0.0496 Accuracy: 0.8046
Epoch 138, CIFAR-10 Batch 2:  Loss: 0.0719 Accuracy: 0.8000
Epoch 138, CIFAR-10 Batch 3:  Loss: 0.0447 Accuracy: 0.8018
Epoch 138, CIFAR-10 Batch 4:  Loss: 0.1721 Accuracy: 0.7850
Epoch 138, CIFAR-10 Batch 5:  Loss: 0.0840 Accuracy: 0.7852
Epoch 139, CIFAR-10 Batch 1:  Loss: 0.0474 Accuracy: 0.8068
Epoch 139, CIFAR-10 Batch 2:  Loss: 0.0587 Accuracy: 0.8090
Epoch 139, CIFAR-10 Batch 3:  Loss: 0.0444 Accuracy: 0.8110
Epoch 139, CIFAR-10 Batch 4:  Loss: 0.0695 Accuracy: 0.8100
Epoch 139, CIFAR-10 Batch 5:  Loss: 0.0870 Accuracy: 0.7920
Epoch 140, CIFAR-10 Batch 1:  Loss: 0.0350 Accuracy: 0.8120
Epoch 140, CIFAR-10 Batch 2:  Loss: 0.0534 Accuracy: 0.8074
Epoch 140, CIFAR-10 Batch 3:  Loss: 0.0488 Accuracy: 0.7934
Epoch 140, CIFAR-10 Batch 4:  Loss: 0.1470 Accuracy: 0.7920
Epoch 140, CIFAR-10 Batch 5:  Loss: 0.0812 Accuracy: 0.7972
Epoch 141, CIFAR-10 Batch 1:  Loss: 0.0273 Accuracy: 0.8076
Epoch 141, CIFAR-10 Batch 2:  Loss: 0.1401 Accuracy: 0.8064
Epoch 141, CIFAR-10 Batch 3:  Loss: 0.0448 Accuracy: 0.8060
Epoch 141, CIFAR-10 Batch 4:  Loss: 0.0419 Accuracy: 0.8218
Epoch 141, CIFAR-10 Batch 5:  Loss: 0.0773 Accuracy: 0.7942
Epoch 142, CIFAR-10 Batch 1:  Loss: 0.0543 Accuracy: 0.8072
Epoch 142, CIFAR-10 Batch 2:  Loss: 0.0582 Accuracy: 0.7962
Epoch 142, CIFAR-10 Batch 3:  Loss: 0.0446 Accuracy: 0.7972
Epoch 142, CIFAR-10 Batch 4:  Loss: 0.0582 Accuracy: 0.8056
Epoch 142, CIFAR-10 Batch 5:  Loss: 0.0504 Accuracy: 0.7974
Epoch 143, CIFAR-10 Batch 1:  Loss: 0.0473 Accuracy: 0.8120
Epoch 143, CIFAR-10 Batch 2:  Loss: 0.0419 Accuracy: 0.8152
Epoch 143, CIFAR-10 Batch 3:  Loss: 0.0410 Accuracy: 0.8034
Epoch 143, CIFAR-10 Batch 4:  Loss: 0.0578 Accuracy: 0.8118
Epoch 143, CIFAR-10 Batch 5:  Loss: 0.0564 Accuracy: 0.8034
Epoch 144, CIFAR-10 Batch 1:  Loss: 0.0500 Accuracy: 0.8096
Epoch 144, CIFAR-10 Batch 2:  Loss: 0.0672 Accuracy: 0.8090
Epoch 144, CIFAR-10 Batch 3:  Loss: 0.0316 Accuracy: 0.7934
Epoch 144, CIFAR-10 Batch 4:  Loss: 0.1097 Accuracy: 0.7962
Epoch 144, CIFAR-10 Batch 5:  Loss: 0.0475 Accuracy: 0.8038
Epoch 145, CIFAR-10 Batch 1:  Loss: 0.0384 Accuracy: 0.8116
Epoch 145, CIFAR-10 Batch 2:  Loss: 0.0385 Accuracy: 0.8112
Epoch 145, CIFAR-10 Batch 3:  Loss: 0.0478 Accuracy: 0.7942
Epoch 145, CIFAR-10 Batch 4:  Loss: 0.0762 Accuracy: 0.8078
Epoch 145, CIFAR-10 Batch 5:  Loss: 0.0744 Accuracy: 0.7948
Epoch 146, CIFAR-10 Batch 1:  Loss: 0.0454 Accuracy: 0.8106
Epoch 146, CIFAR-10 Batch 2:  Loss: 0.1388 Accuracy: 0.7846
Epoch 146, CIFAR-10 Batch 3:  Loss: 0.0469 Accuracy: 0.7962
Epoch 146, CIFAR-10 Batch 4:  Loss: 0.1026 Accuracy: 0.7982
Epoch 146, CIFAR-10 Batch 5:  Loss: 0.0414 Accuracy: 0.8042
Epoch 147, CIFAR-10 Batch 1:  Loss: 0.0373 Accuracy: 0.8138
Epoch 147, CIFAR-10 Batch 2:  Loss: 0.0533 Accuracy: 0.8050
Epoch 147, CIFAR-10 Batch 3:  Loss: 0.0544 Accuracy: 0.8044
Epoch 147, CIFAR-10 Batch 4:  Loss: 0.0556 Accuracy: 0.8092
Epoch 147, CIFAR-10 Batch 5:  Loss: 0.0659 Accuracy: 0.7756
Epoch 148, CIFAR-10 Batch 1:  Loss: 0.0387 Accuracy: 0.8096
Epoch 148, CIFAR-10 Batch 2:  Loss: 0.1269 Accuracy: 0.7956
Epoch 148, CIFAR-10 Batch 3:  Loss: 0.0321 Accuracy: 0.8024
Epoch 148, CIFAR-10 Batch 4:  Loss: 0.0561 Accuracy: 0.8106
Epoch 148, CIFAR-10 Batch 5:  Loss: 0.0855 Accuracy: 0.7856
Epoch 149, CIFAR-10 Batch 1:  Loss: 0.0350 Accuracy: 0.8176
Epoch 149, CIFAR-10 Batch 2:  Loss: 0.0860 Accuracy: 0.8074
Epoch 149, CIFAR-10 Batch 3:  Loss: 0.0536 Accuracy: 0.7920
Epoch 149, CIFAR-10 Batch 4:  Loss: 0.0524 Accuracy: 0.8170
Epoch 149, CIFAR-10 Batch 5:  Loss: 0.0388 Accuracy: 0.8056
Epoch 150, CIFAR-10 Batch 1:  Loss: 0.0558 Accuracy: 0.7996
Epoch 150, CIFAR-10 Batch 2:  Loss: 0.0762 Accuracy: 0.8040
Epoch 150, CIFAR-10 Batch 3:  Loss: 0.0585 Accuracy: 0.8052
Epoch 150, CIFAR-10 Batch 4:  Loss: 0.0829 Accuracy: 0.8072
Epoch 150, CIFAR-10 Batch 5:  Loss: 0.0522 Accuracy: 0.7970
Epoch 151, CIFAR-10 Batch 1:  Loss: 0.0321 Accuracy: 0.8190
Epoch 151, CIFAR-10 Batch 2:  Loss: 0.1268 Accuracy: 0.7994
Epoch 151, CIFAR-10 Batch 3:  Loss: 0.0326 Accuracy: 0.8092
Epoch 151, CIFAR-10 Batch 4:  Loss: 0.0969 Accuracy: 0.8008
Epoch 151, CIFAR-10 Batch 5:  Loss: 0.0626 Accuracy: 0.7790
Epoch 152, CIFAR-10 Batch 1:  Loss: 0.0243 Accuracy: 0.8166
Epoch 152, CIFAR-10 Batch 2:  Loss: 0.0886 Accuracy: 0.7914
Epoch 152, CIFAR-10 Batch 3:  Loss: 0.0380 Accuracy: 0.7924
Epoch 152, CIFAR-10 Batch 4:  Loss: 0.0602 Accuracy: 0.8044
Epoch 152, CIFAR-10 Batch 5:  Loss: 0.0723 Accuracy: 0.7802
Epoch 153, CIFAR-10 Batch 1:  Loss: 0.0234 Accuracy: 0.8212
Epoch 153, CIFAR-10 Batch 2:  Loss: 0.0552 Accuracy: 0.8100
Epoch 153, CIFAR-10 Batch 3:  Loss: 0.0588 Accuracy: 0.7928
Epoch 153, CIFAR-10 Batch 4:  Loss: 0.0715 Accuracy: 0.8080
Epoch 153, CIFAR-10 Batch 5:  Loss: 0.0429 Accuracy: 0.8122
Epoch 154, CIFAR-10 Batch 1:  Loss: 0.0398 Accuracy: 0.8136
Epoch 154, CIFAR-10 Batch 2:  Loss: 0.0465 Accuracy: 0.8166
Epoch 154, CIFAR-10 Batch 3:  Loss: 0.0544 Accuracy: 0.7968
Epoch 154, CIFAR-10 Batch 4:  Loss: 0.0860 Accuracy: 0.7994
Epoch 154, CIFAR-10 Batch 5:  Loss: 0.0390 Accuracy: 0.8096
Epoch 155, CIFAR-10 Batch 1:  Loss: 0.0537 Accuracy: 0.8036
Epoch 155, CIFAR-10 Batch 2:  Loss: 0.0680 Accuracy: 0.8024
Epoch 155, CIFAR-10 Batch 3:  Loss: 0.0353 Accuracy: 0.7992
Epoch 155, CIFAR-10 Batch 4:  Loss: 0.0501 Accuracy: 0.8084
Epoch 155, CIFAR-10 Batch 5:  Loss: 0.0416 Accuracy: 0.8002
Epoch 156, CIFAR-10 Batch 1:  Loss: 0.0481 Accuracy: 0.8002
Epoch 156, CIFAR-10 Batch 2:  Loss: 0.0488 Accuracy: 0.8116
Epoch 156, CIFAR-10 Batch 3:  Loss: 0.0195 Accuracy: 0.8058
Epoch 156, CIFAR-10 Batch 4:  Loss: 0.0306 Accuracy: 0.8120
Epoch 156, CIFAR-10 Batch 5:  Loss: 0.0354 Accuracy: 0.8098
Epoch 157, CIFAR-10 Batch 1:  Loss: 0.0506 Accuracy: 0.8146
Epoch 157, CIFAR-10 Batch 2:  Loss: 0.0482 Accuracy: 0.8218
Epoch 157, CIFAR-10 Batch 3:  Loss: 0.0295 Accuracy: 0.8090
Epoch 157, CIFAR-10 Batch 4:  Loss: 0.0352 Accuracy: 0.8196
Epoch 157, CIFAR-10 Batch 5:  Loss: 0.0468 Accuracy: 0.8028
Epoch 158, CIFAR-10 Batch 1:  Loss: 0.0410 Accuracy: 0.8126
Epoch 158, CIFAR-10 Batch 2:  Loss: 0.0424 Accuracy: 0.8068
Epoch 158, CIFAR-10 Batch 3:  Loss: 0.0394 Accuracy: 0.7872
Epoch 158, CIFAR-10 Batch 4:  Loss: 0.0586 Accuracy: 0.8128
Epoch 158, CIFAR-10 Batch 5:  Loss: 0.1323 Accuracy: 0.7652
Epoch 159, CIFAR-10 Batch 1:  Loss: 0.0475 Accuracy: 0.8046
Epoch 159, CIFAR-10 Batch 2:  Loss: 0.0476 Accuracy: 0.8136
Epoch 159, CIFAR-10 Batch 3:  Loss: 0.0423 Accuracy: 0.7984
Epoch 159, CIFAR-10 Batch 4:  Loss: 0.0372 Accuracy: 0.8190
Epoch 159, CIFAR-10 Batch 5:  Loss: 0.0470 Accuracy: 0.7972
Epoch 160, CIFAR-10 Batch 1:  Loss: 0.0324 Accuracy: 0.8116
Epoch 160, CIFAR-10 Batch 2:  Loss: 0.0728 Accuracy: 0.7826
Epoch 160, CIFAR-10 Batch 3:  Loss: 0.0602 Accuracy: 0.7768
Epoch 160, CIFAR-10 Batch 4:  Loss: 0.0482 Accuracy: 0.8062
Epoch 160, CIFAR-10 Batch 5:  Loss: 0.0434 Accuracy: 0.8110
Epoch 161, CIFAR-10 Batch 1:  Loss: 0.0558 Accuracy: 0.8050
Epoch 161, CIFAR-10 Batch 2:  Loss: 0.0642 Accuracy: 0.8006
Epoch 161, CIFAR-10 Batch 3:  Loss: 0.0526 Accuracy: 0.7848
Epoch 161, CIFAR-10 Batch 4:  Loss: 0.1503 Accuracy: 0.7774
Epoch 161, CIFAR-10 Batch 5:  Loss: 0.0406 Accuracy: 0.7930
Epoch 162, CIFAR-10 Batch 1:  Loss: 0.0274 Accuracy: 0.8072
Epoch 162, CIFAR-10 Batch 2:  Loss: 0.0279 Accuracy: 0.8218
Epoch 162, CIFAR-10 Batch 3:  Loss: 0.0335 Accuracy: 0.8120
Epoch 162, CIFAR-10 Batch 4:  Loss: 0.0746 Accuracy: 0.8044
Epoch 162, CIFAR-10 Batch 5:  Loss: 0.0404 Accuracy: 0.8020
Epoch 163, CIFAR-10 Batch 1:  Loss: 0.0407 Accuracy: 0.8056
Epoch 163, CIFAR-10 Batch 2:  Loss: 0.0707 Accuracy: 0.7926
Epoch 163, CIFAR-10 Batch 3:  Loss: 0.0356 Accuracy: 0.7998
Epoch 163, CIFAR-10 Batch 4:  Loss: 0.0450 Accuracy: 0.8186
Epoch 163, CIFAR-10 Batch 5:  Loss: 0.0483 Accuracy: 0.7924
Epoch 164, CIFAR-10 Batch 1:  Loss: 0.0877 Accuracy: 0.7928
Epoch 164, CIFAR-10 Batch 2:  Loss: 0.0505 Accuracy: 0.8126
Epoch 164, CIFAR-10 Batch 3:  Loss: 0.0205 Accuracy: 0.8136
Epoch 164, CIFAR-10 Batch 4:  Loss: 0.0553 Accuracy: 0.8074
Epoch 164, CIFAR-10 Batch 5:  Loss: 0.0643 Accuracy: 0.7944
Epoch 165, CIFAR-10 Batch 1:  Loss: 0.0324 Accuracy: 0.8108
Epoch 165, CIFAR-10 Batch 2:  Loss: 0.0940 Accuracy: 0.7864
Epoch 165, CIFAR-10 Batch 3:  Loss: 0.0387 Accuracy: 0.8036
Epoch 165, CIFAR-10 Batch 4:  Loss: 0.0435 Accuracy: 0.8036
Epoch 165, CIFAR-10 Batch 5:  Loss: 0.0443 Accuracy: 0.8014
Epoch 166, CIFAR-10 Batch 1:  Loss: 0.0329 Accuracy: 0.8098
Epoch 166, CIFAR-10 Batch 2:  Loss: 0.0658 Accuracy: 0.8094
Epoch 166, CIFAR-10 Batch 3:  Loss: 0.0337 Accuracy: 0.8120
Epoch 166, CIFAR-10 Batch 4:  Loss: 0.1041 Accuracy: 0.7812
Epoch 166, CIFAR-10 Batch 5:  Loss: 0.0449 Accuracy: 0.7956
Epoch 167, CIFAR-10 Batch 1:  Loss: 0.0303 Accuracy: 0.8066
Epoch 167, CIFAR-10 Batch 2:  Loss: 0.0597 Accuracy: 0.8124
Epoch 167, CIFAR-10 Batch 3:  Loss: 0.0303 Accuracy: 0.8058
Epoch 167, CIFAR-10 Batch 4:  Loss: 0.0861 Accuracy: 0.8030
Epoch 167, CIFAR-10 Batch 5:  Loss: 0.0408 Accuracy: 0.8052
Epoch 168, CIFAR-10 Batch 1:  Loss: 0.0233 Accuracy: 0.8182
Epoch 168, CIFAR-10 Batch 2:  Loss: 0.0700 Accuracy: 0.8000
Epoch 168, CIFAR-10 Batch 3:  Loss: 0.0326 Accuracy: 0.8102
Epoch 168, CIFAR-10 Batch 4:  Loss: 0.0477 Accuracy: 0.8182
Epoch 168, CIFAR-10 Batch 5:  Loss: 0.0559 Accuracy: 0.7990
Epoch 169, CIFAR-10 Batch 1:  Loss: 0.0240 Accuracy: 0.8112
Epoch 169, CIFAR-10 Batch 2:  Loss: 0.0720 Accuracy: 0.8044
Epoch 169, CIFAR-10 Batch 3:  Loss: 0.0349 Accuracy: 0.7932
Epoch 169, CIFAR-10 Batch 4:  Loss: 0.0352 Accuracy: 0.8150
Epoch 169, CIFAR-10 Batch 5:  Loss: 0.0440 Accuracy: 0.8092
Epoch 170, CIFAR-10 Batch 1:  Loss: 0.0354 Accuracy: 0.8092
Epoch 170, CIFAR-10 Batch 2:  Loss: 0.0386 Accuracy: 0.8094
Epoch 170, CIFAR-10 Batch 3:  Loss: 0.0509 Accuracy: 0.7872
Epoch 170, CIFAR-10 Batch 4:  Loss: 0.0274 Accuracy: 0.8190
Epoch 170, CIFAR-10 Batch 5:  Loss: 0.0440 Accuracy: 0.8046
Epoch 171, CIFAR-10 Batch 1:  Loss: 0.0246 Accuracy: 0.8164
Epoch 171, CIFAR-10 Batch 2:  Loss: 0.0840 Accuracy: 0.7952
Epoch 171, CIFAR-10 Batch 3:  Loss: 0.0421 Accuracy: 0.8058
Epoch 171, CIFAR-10 Batch 4:  Loss: 0.0812 Accuracy: 0.7968
Epoch 171, CIFAR-10 Batch 5:  Loss: 0.0804 Accuracy: 0.7732
Epoch 172, CIFAR-10 Batch 1:  Loss: 0.0279 Accuracy: 0.8066
Epoch 172, CIFAR-10 Batch 2:  Loss: 0.0349 Accuracy: 0.8176
Epoch 172, CIFAR-10 Batch 3:  Loss: 0.0354 Accuracy: 0.8082
Epoch 172, CIFAR-10 Batch 4:  Loss: 0.0421 Accuracy: 0.8146
Epoch 172, CIFAR-10 Batch 5:  Loss: 0.0596 Accuracy: 0.7882
Epoch 173, CIFAR-10 Batch 1:  Loss: 0.0418 Accuracy: 0.8090
Epoch 173, CIFAR-10 Batch 2:  Loss: 0.0391 Accuracy: 0.8120
Epoch 173, CIFAR-10 Batch 3:  Loss: 0.0563 Accuracy: 0.7974
Epoch 173, CIFAR-10 Batch 4:  Loss: 0.0394 Accuracy: 0.8134
Epoch 173, CIFAR-10 Batch 5:  Loss: 0.0312 Accuracy: 0.8078
Epoch 174, CIFAR-10 Batch 1:  Loss: 0.0516 Accuracy: 0.7888
Epoch 174, CIFAR-10 Batch 2:  Loss: 0.0420 Accuracy: 0.8062
Epoch 174, CIFAR-10 Batch 3:  Loss: 0.0343 Accuracy: 0.7988
Epoch 174, CIFAR-10 Batch 4:  Loss: 0.0473 Accuracy: 0.8096
Epoch 174, CIFAR-10 Batch 5:  Loss: 0.0612 Accuracy: 0.7866
Epoch 175, CIFAR-10 Batch 1:  Loss: 0.0237 Accuracy: 0.8146
Epoch 175, CIFAR-10 Batch 2:  Loss: 0.1020 Accuracy: 0.8012
Epoch 175, CIFAR-10 Batch 3:  Loss: 0.0637 Accuracy: 0.7808
Epoch 175, CIFAR-10 Batch 4:  Loss: 0.0333 Accuracy: 0.8114
Epoch 175, CIFAR-10 Batch 5:  Loss: 0.0227 Accuracy: 0.8174
Epoch 176, CIFAR-10 Batch 1:  Loss: 0.0261 Accuracy: 0.8150
Epoch 176, CIFAR-10 Batch 2:  Loss: 0.0245 Accuracy: 0.8156
Epoch 176, CIFAR-10 Batch 3:  Loss: 0.0504 Accuracy: 0.8052
Epoch 176, CIFAR-10 Batch 4:  Loss: 0.0236 Accuracy: 0.8240
Epoch 176, CIFAR-10 Batch 5:  Loss: 0.0284 Accuracy: 0.8014
Epoch 177, CIFAR-10 Batch 1:  Loss: 0.0224 Accuracy: 0.8116
Epoch 177, CIFAR-10 Batch 2:  Loss: 0.0347 Accuracy: 0.8082
Epoch 177, CIFAR-10 Batch 3:  Loss: 0.0573 Accuracy: 0.7768
Epoch 177, CIFAR-10 Batch 4:  Loss: 0.0336 Accuracy: 0.8202
Epoch 177, CIFAR-10 Batch 5:  Loss: 0.0398 Accuracy: 0.7942
Epoch 178, CIFAR-10 Batch 1:  Loss: 0.0512 Accuracy: 0.8034
Epoch 178, CIFAR-10 Batch 2:  Loss: 0.0650 Accuracy: 0.7968
Epoch 178, CIFAR-10 Batch 3:  Loss: 0.0341 Accuracy: 0.8114
Epoch 178, CIFAR-10 Batch 4:  Loss: 0.0298 Accuracy: 0.8142
Epoch 178, CIFAR-10 Batch 5:  Loss: 0.0493 Accuracy: 0.7916
Epoch 179, CIFAR-10 Batch 1:  Loss: 0.0273 Accuracy: 0.8142
Epoch 179, CIFAR-10 Batch 2:  Loss: 0.0331 Accuracy: 0.8056
Epoch 179, CIFAR-10 Batch 3:  Loss: 0.0507 Accuracy: 0.8044
Epoch 179, CIFAR-10 Batch 4:  Loss: 0.0320 Accuracy: 0.8062
Epoch 179, CIFAR-10 Batch 5:  Loss: 0.0615 Accuracy: 0.7922
Epoch 180, CIFAR-10 Batch 1:  Loss: 0.0286 Accuracy: 0.8108
Epoch 180, CIFAR-10 Batch 2:  Loss: 0.0513 Accuracy: 0.8054
Epoch 180, CIFAR-10 Batch 3:  Loss: 0.0457 Accuracy: 0.7934
Epoch 180, CIFAR-10 Batch 4:  Loss: 0.0410 Accuracy: 0.8154
Epoch 180, CIFAR-10 Batch 5:  Loss: 0.0354 Accuracy: 0.7926
Epoch 181, CIFAR-10 Batch 1:  Loss: 0.0333 Accuracy: 0.8026
Epoch 181, CIFAR-10 Batch 2:  Loss: 0.0237 Accuracy: 0.8102
Epoch 181, CIFAR-10 Batch 3:  Loss: 0.0524 Accuracy: 0.7900
Epoch 181, CIFAR-10 Batch 4:  Loss: 0.0321 Accuracy: 0.8174
Epoch 181, CIFAR-10 Batch 5:  Loss: 0.0346 Accuracy: 0.8060
Epoch 182, CIFAR-10 Batch 1:  Loss: 0.0203 Accuracy: 0.8182
Epoch 182, CIFAR-10 Batch 2:  Loss: 0.0300 Accuracy: 0.8118
Epoch 182, CIFAR-10 Batch 3:  Loss: 0.0277 Accuracy: 0.8040
Epoch 182, CIFAR-10 Batch 4:  Loss: 0.0278 Accuracy: 0.8174
Epoch 182, CIFAR-10 Batch 5:  Loss: 0.0406 Accuracy: 0.7892
Epoch 183, CIFAR-10 Batch 1:  Loss: 0.0318 Accuracy: 0.8150
Epoch 183, CIFAR-10 Batch 2:  Loss: 0.0484 Accuracy: 0.7988
Epoch 183, CIFAR-10 Batch 3:  Loss: 0.0550 Accuracy: 0.7908
Epoch 183, CIFAR-10 Batch 4:  Loss: 0.0204 Accuracy: 0.8134
Epoch 183, CIFAR-10 Batch 5:  Loss: 0.0284 Accuracy: 0.8022
Epoch 184, CIFAR-10 Batch 1:  Loss: 0.0472 Accuracy: 0.8016
Epoch 184, CIFAR-10 Batch 2:  Loss: 0.0448 Accuracy: 0.8016
Epoch 184, CIFAR-10 Batch 3:  Loss: 0.0288 Accuracy: 0.8082
Epoch 184, CIFAR-10 Batch 4:  Loss: 0.0712 Accuracy: 0.7934
Epoch 184, CIFAR-10 Batch 5:  Loss: 0.0572 Accuracy: 0.7850
Epoch 185, CIFAR-10 Batch 1:  Loss: 0.0213 Accuracy: 0.8100
Epoch 185, CIFAR-10 Batch 2:  Loss: 0.0171 Accuracy: 0.8118
Epoch 185, CIFAR-10 Batch 3:  Loss: 0.0507 Accuracy: 0.7924
Epoch 185, CIFAR-10 Batch 4:  Loss: 0.0493 Accuracy: 0.8006
Epoch 185, CIFAR-10 Batch 5:  Loss: 0.0349 Accuracy: 0.7978
Epoch 186, CIFAR-10 Batch 1:  Loss: 0.0275 Accuracy: 0.8160
Epoch 186, CIFAR-10 Batch 2:  Loss: 0.0539 Accuracy: 0.8000
Epoch 186, CIFAR-10 Batch 3:  Loss: 0.0453 Accuracy: 0.7952
Epoch 186, CIFAR-10 Batch 4:  Loss: 0.0484 Accuracy: 0.8060
Epoch 186, CIFAR-10 Batch 5:  Loss: 0.0591 Accuracy: 0.7882
Epoch 187, CIFAR-10 Batch 1:  Loss: 0.0393 Accuracy: 0.8008
Epoch 187, CIFAR-10 Batch 2:  Loss: 0.0429 Accuracy: 0.8084
Epoch 187, CIFAR-10 Batch 3:  Loss: 0.0310 Accuracy: 0.8028
Epoch 187, CIFAR-10 Batch 4:  Loss: 0.0394 Accuracy: 0.8108
Epoch 187, CIFAR-10 Batch 5:  Loss: 0.0357 Accuracy: 0.8022
Epoch 188, CIFAR-10 Batch 1:  Loss: 0.0420 Accuracy: 0.8100
Epoch 188, CIFAR-10 Batch 2:  Loss: 0.0981 Accuracy: 0.7998
Epoch 188, CIFAR-10 Batch 3:  Loss: 0.0274 Accuracy: 0.8052
Epoch 188, CIFAR-10 Batch 4:  Loss: 0.0422 Accuracy: 0.8058
Epoch 188, CIFAR-10 Batch 5:  Loss: 0.0223 Accuracy: 0.8142
Epoch 189, CIFAR-10 Batch 1:  Loss: 0.0436 Accuracy: 0.8050
Epoch 189, CIFAR-10 Batch 2:  Loss: 0.0494 Accuracy: 0.8052
Epoch 189, CIFAR-10 Batch 3:  Loss: 0.0271 Accuracy: 0.8062
Epoch 189, CIFAR-10 Batch 4:  Loss: 0.0295 Accuracy: 0.8092
Epoch 189, CIFAR-10 Batch 5:  Loss: 0.0417 Accuracy: 0.8064
Epoch 190, CIFAR-10 Batch 1:  Loss: 0.0239 Accuracy: 0.8026
Epoch 190, CIFAR-10 Batch 2:  Loss: 0.0402 Accuracy: 0.8028
Epoch 190, CIFAR-10 Batch 3:  Loss: 0.0530 Accuracy: 0.7804
Epoch 190, CIFAR-10 Batch 4:  Loss: 0.0293 Accuracy: 0.8054
Epoch 190, CIFAR-10 Batch 5:  Loss: 0.0279 Accuracy: 0.8030
Epoch 191, CIFAR-10 Batch 1:  Loss: 0.0147 Accuracy: 0.8198
Epoch 191, CIFAR-10 Batch 2:  Loss: 0.0455 Accuracy: 0.8078
Epoch 191, CIFAR-10 Batch 3:  Loss: 0.0461 Accuracy: 0.7942
Epoch 191, CIFAR-10 Batch 4:  Loss: 0.0258 Accuracy: 0.8212
Epoch 191, CIFAR-10 Batch 5:  Loss: 0.0322 Accuracy: 0.8100
Epoch 192, CIFAR-10 Batch 1:  Loss: 0.0323 Accuracy: 0.8116
Epoch 192, CIFAR-10 Batch 2:  Loss: 0.0482 Accuracy: 0.8016
Epoch 192, CIFAR-10 Batch 3:  Loss: 0.0232 Accuracy: 0.8104
Epoch 192, CIFAR-10 Batch 4:  Loss: 0.0377 Accuracy: 0.8132
Epoch 192, CIFAR-10 Batch 5:  Loss: 0.0403 Accuracy: 0.8162
Epoch 193, CIFAR-10 Batch 1:  Loss: 0.0192 Accuracy: 0.8118
Epoch 193, CIFAR-10 Batch 2:  Loss: 0.0467 Accuracy: 0.8056
Epoch 193, CIFAR-10 Batch 3:  Loss: 0.0338 Accuracy: 0.7920
Epoch 193, CIFAR-10 Batch 4:  Loss: 0.0225 Accuracy: 0.8140
Epoch 193, CIFAR-10 Batch 5:  Loss: 0.0307 Accuracy: 0.8072
Epoch 194, CIFAR-10 Batch 1:  Loss: 0.0274 Accuracy: 0.8006
Epoch 194, CIFAR-10 Batch 2:  Loss: 0.0208 Accuracy: 0.8104
Epoch 194, CIFAR-10 Batch 3:  Loss: 0.0182 Accuracy: 0.7920
Epoch 194, CIFAR-10 Batch 4:  Loss: 0.0251 Accuracy: 0.8214
Epoch 194, CIFAR-10 Batch 5:  Loss: 0.0361 Accuracy: 0.8002
Epoch 195, CIFAR-10 Batch 1:  Loss: 0.0220 Accuracy: 0.8180
Epoch 195, CIFAR-10 Batch 2:  Loss: 0.1731 Accuracy: 0.7844
Epoch 195, CIFAR-10 Batch 3:  Loss: 0.0327 Accuracy: 0.8022
Epoch 195, CIFAR-10 Batch 4:  Loss: 0.0248 Accuracy: 0.8106
Epoch 195, CIFAR-10 Batch 5:  Loss: 0.0435 Accuracy: 0.7890
Epoch 196, CIFAR-10 Batch 1:  Loss: 0.0332 Accuracy: 0.8084
Epoch 196, CIFAR-10 Batch 2:  Loss: 0.0329 Accuracy: 0.8056
Epoch 196, CIFAR-10 Batch 3:  Loss: 0.0316 Accuracy: 0.8116
Epoch 196, CIFAR-10 Batch 4:  Loss: 0.0269 Accuracy: 0.8142
Epoch 196, CIFAR-10 Batch 5:  Loss: 0.0507 Accuracy: 0.7858
Epoch 197, CIFAR-10 Batch 1:  Loss: 0.0142 Accuracy: 0.8238
Epoch 197, CIFAR-10 Batch 2:  Loss: 0.0216 Accuracy: 0.8114
Epoch 197, CIFAR-10 Batch 3:  Loss: 0.0299 Accuracy: 0.7906
Epoch 197, CIFAR-10 Batch 4:  Loss: 0.0567 Accuracy: 0.8004
Epoch 197, CIFAR-10 Batch 5:  Loss: 0.0385 Accuracy: 0.8060
Epoch 198, CIFAR-10 Batch 1:  Loss: 0.0241 Accuracy: 0.8132
Epoch 198, CIFAR-10 Batch 2:  Loss: 0.0290 Accuracy: 0.8036
Epoch 198, CIFAR-10 Batch 3:  Loss: 0.0240 Accuracy: 0.7924
Epoch 198, CIFAR-10 Batch 4:  Loss: 0.0680 Accuracy: 0.8060
Epoch 198, CIFAR-10 Batch 5:  Loss: 0.0345 Accuracy: 0.8008
Epoch 199, CIFAR-10 Batch 1:  Loss: 0.0267 Accuracy: 0.8016
Epoch 199, CIFAR-10 Batch 2:  Loss: 0.0379 Accuracy: 0.7944
Epoch 199, CIFAR-10 Batch 3:  Loss: 0.0110 Accuracy: 0.8182
Epoch 199, CIFAR-10 Batch 4:  Loss: 0.0206 Accuracy: 0.8158
Epoch 199, CIFAR-10 Batch 5:  Loss: 0.0278 Accuracy: 0.8072
Epoch 200, CIFAR-10 Batch 1:  Loss: 0.0252 Accuracy: 0.8152
Epoch 200, CIFAR-10 Batch 2:  Loss: 0.0375 Accuracy: 0.8060
Epoch 200, CIFAR-10 Batch 3:  Loss: 0.0261 Accuracy: 0.7974
Epoch 200, CIFAR-10 Batch 4:  Loss: 0.0362 Accuracy: 0.8098
Epoch 200, CIFAR-10 Batch 5:  Loss: 0.0660 Accuracy: 0.7958

检查点

模型已保存到本地。

测试模型

利用测试数据集测试你的模型。这将是最终的准确率。你的准确率应该高于 50%。如果没达到，请继续调整模型结构和参数。



In [1]:

    
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
%config InlineBackend.figure_format = 'retina'

import tensorflow as tf
import pickle
import helper
import random

# Set batch size if not already set
try:
    if batch_size:
        pass
except NameError:
    batch_size = 64

save_model_path = './image_classification'
n_samples = 4
top_n_predictions = 3

def test_model():
    """
    Test the saved model against the test dataset
    """

    test_features, test_labels = pickle.load(open('preprocess_test.p', mode='rb'))
    loaded_graph = tf.Graph()

    with tf.Session(graph=loaded_graph) as sess:
        # Load model
        loader = tf.train.import_meta_graph(save_model_path + '.meta')
        loader.restore(sess, save_model_path)

        # Get Tensors from loaded model
        loaded_x = loaded_graph.get_tensor_by_name('x:0')
        loaded_y = loaded_graph.get_tensor_by_name('y:0')
        loaded_keep_prob = loaded_graph.get_tensor_by_name('keep_prob:0')
        loaded_logits = loaded_graph.get_tensor_by_name('logits:0')
        loaded_acc = loaded_graph.get_tensor_by_name('accuracy:0')
        
        # Get accuracy in batches for memory limitations
        test_batch_acc_total = 0
        test_batch_count = 0
        
        for test_feature_batch, test_label_batch in helper.batch_features_labels(test_features, test_labels, batch_size):
            test_batch_acc_total += sess.run(
                loaded_acc,
                feed_dict={loaded_x: test_feature_batch, loaded_y: test_label_batch, loaded_keep_prob: 1.0})
            test_batch_count += 1

        print('Testing Accuracy: {}\n'.format(test_batch_acc_total/test_batch_count))

        # Print Random Samples
        random_test_features, random_test_labels = tuple(zip(*random.sample(list(zip(test_features, test_labels)), n_samples)))
        random_test_predictions = sess.run(
            tf.nn.top_k(tf.nn.softmax(loaded_logits), top_n_predictions),
            feed_dict={loaded_x: random_test_features, loaded_y: random_test_labels, loaded_keep_prob: 1.0})
        helper.display_image_predictions(random_test_features, random_test_labels, random_test_predictions)


test_model()









    



INFO:tensorflow:Restoring parameters from ./image_classification
Testing Accuracy: 0.7875199044585988

为何准确率只有50-80%？

你可能想问，为何准确率不能更高了？首先，对于简单的 CNN 网络来说，50% 已经不低了。纯粹猜测的准确率为10%。但是，你可能注意到有人的准确率远远超过 80%。这是因为我们还没有介绍所有的神经网络知识。我们还需要掌握一些其他技巧。

提交项目

提交项目时，确保先运行所有单元，然后再保存记事本。将 notebook 文件另存为“dlnd_image_classification.ipynb”，再在目录 "File" -> "Download as" 另存为 HTML 格式。请在提交的项目中包含 “helper.py” 和 “problem_unittests.py” 文件。