[PYTHON] "Deep Learning from scratch" Self-study memo (No. 14) Run the program in Chapter 4 on Google Colaboratory

While reading "Deep Learning from scratch" (written by Yasuki Saito, published by O'Reilly Japan), I will make a note of the sites I referred to. Part 13 ←

I used Google Colaboratory to run the program in the book.

I am doing it with such a folder structure.

content/
└ Drive name/
     └My Drive/
        └Colab Notebooks/
            └deep_learning/
                ├common/
                ｜  ├functions.py
                ｜  └two_layer_net.py
                ├dataset/
                ｜  ├mnist.py
                ｜  ├mnist.pkl
                ｜  └lena.png
                ├ch04 
                ｜  ├train_neuralnet.ipynb
                ｜  └test_neuralne.ipynb
                ｜ 

Modify the import part of two_layer_net.py in Chapter 4 and set it in the common folder.

two placed in the common folder_layer_net.py

import sys, os
sys.path.append(os.pardir)  
from functions import *
from gradient import numerical_gradient
import numpy as np

class TwoLayerNet:

    def __init__(self, input_size, hidden_size, output_size, weight_init_std=0.01):
        #Weight initialization
        self.params = {}
        self.params['W1'] = weight_init_std * np.random.randn(input_size, hidden_size)
        self.params['b1'] = np.zeros(hidden_size)
        self.params['W2'] = weight_init_std * np.random.randn(hidden_size, output_size)
        self.params['b2'] = np.zeros(output_size)

    def predict(self, x):
        W1, W2 = self.params['W1'], self.params['W2']
        b1, b2 = self.params['b1'], self.params['b2']
    
        a1 = np.dot(x, W1) + b1
        z1 = sigmoid(a1)
        a2 = np.dot(z1, W2) + b2
        y = softmax(a2)
        
        return y
        
    # x:Input data, t:Teacher data
    def loss(self, x, t):
        y = self.predict(x)
        
        return cross_entropy_error(y, t)
    
    def accuracy(self, x, t):
        y = self.predict(x)
        y = np.argmax(y, axis=1)
        t = np.argmax(t, axis=1)
        
        accuracy = np.sum(y == t) / float(x.shape[0])
        return accuracy
        
    # x:Input data, t:Teacher data
    def numerical_gradient(self, x, t):
        loss_W = lambda W: self.loss(x, t)
        
        grads = {}
        grads['W1'] = numerical_gradient(loss_W, self.params['W1'])
        grads['b1'] = numerical_gradient(loss_W, self.params['b1'])
        grads['W2'] = numerical_gradient(loss_W, self.params['W2'])
        grads['b2'] = numerical_gradient(loss_W, self.params['b2'])
        
        return grads
        
    def gradient(self, x, t):
        W1, W2 = self.params['W1'], self.params['W2']
        b1, b2 = self.params['b1'], self.params['b2']
        grads = {}
        
        batch_num = x.shape[0]
        
        # forward
        a1 = np.dot(x, W1) + b1
        z1 = sigmoid(a1)
        a2 = np.dot(z1, W2) + b2
        y = softmax(a2)
        
        # backward
        dy = (y - t) / batch_num
        grads['W2'] = np.dot(z1.T, dy)
        grads['b2'] = np.sum(dy, axis=0)
        
        dz1 = np.dot(dy, W2.T)
        da1 = sigmoid_grad(a1) * dz1
        grads['W1'] = np.dot(x.T, da1)
        grads['b1'] = np.sum(da1, axis=0)

        return grads

I split the code of train_neuralnet.py into several parts and ran them in interactive mode.

from google.colab import drive
drive.mount('/content/drive')

Mounted at /content/drive

import sys, os
sys.path

['', '/env/python', '/usr/lib/python36.zip', '/usr/lib/python3.6', '/usr/lib/python3.6/lib-dynload', '/usr/local/lib/python3.6/dist-packages', '/usr/lib/python3/dist-packages', '/usr/local/lib/python3.6/dist-packages/IPython/extensions', '/root/.ipython']

sys.path.append(os.pardir)  #Settings for importing files in the parent directory
sys.path.append('/content/drive/My Drive/Colab Notebooks/deep_learning/common')
sys.path.append('/content/drive/My Drive/Colab Notebooks/deep_learning/dataset')

sys.path

['', '/env/python', '/usr/lib/python36.zip', '/usr/lib/python3.6', '/usr/lib/python3.6/lib-dynload', '/usr/local/lib/python3.6/dist-packages', '/usr/lib/python3/dist-packages', '/usr/local/lib/python3.6/dist-packages/IPython/extensions', '/root/.ipython', '..', '/content/drive/My Drive/Colab Notebooks/deep_learning/common', '/content/drive/My Drive/Colab Notebooks/deep_learning/dataset']

import numpy as np
import matplotlib.pyplot as plt
from mnist import load_mnist
from two_layer_net import TwoLayerNet

If you run another script before running this script and the contents of the files or folders are different, you will get a ModuleNotFoundError here. If that happens, you won't be able to proceed even if the script is okay. Runtime> Session Management> Ending a running session and trying to re-execute after a while seems to work.

#Data reading
(x_train, t_train), (x_test, t_test) = load_mnist(normalize=True, one_hot_label=True)

network = TwoLayerNet(input_size=784, hidden_size=50, output_size=10)

iters_num = 10000  #Set the number of repetitions as appropriate
train_size = x_train.shape[0]
batch_size = 100
learning_rate = 0.1

train_loss_list = []
train_acc_list = []
test_acc_list = []

iter_per_epoch = max(train_size / batch_size, 1)

for i in range(iters_num):
    batch_mask = np.random.choice(train_size, batch_size)
    x_batch = x_train[batch_mask]
    t_batch = t_train[batch_mask]
    
    #Gradient calculation
    #grad = network.numerical_gradient(x_batch, t_batch)
    grad = network.gradient(x_batch, t_batch)
    
    #Parameter update
    for key in ('W1', 'b1', 'W2', 'b2'):
        network.params[key] -= learning_rate * grad[key]
    
    loss = network.loss(x_batch, t_batch)
    train_loss_list.append(loss)
    
    if i % iter_per_epoch == 0:
        train_acc = network.accuracy(x_train, t_train)
        test_acc = network.accuracy(x_test, t_test)
        train_acc_list.append(train_acc)
        test_acc_list.append(test_acc)
        print("train acc, test acc | " + str(train_acc) + ", " + str(test_acc))

#Drawing a graph
markers = {'train': 'o', 'test': 's'}
x = np.arange(len(train_acc_list))
plt.plot(x, train_acc_list, label='train acc')
plt.plot(x, test_acc_list, label='test acc', linestyle='--')
plt.xlabel("epochs")
plt.ylabel("accuracy")
plt.ylim(0, 1.0)
plt.legend(loc='lower right')
plt.show()

Furthermore, the learning result is saved in the pkl file.

#Save the network object with pickle.
import pickle
save_file = '/content/drive/My Drive/Colab Notebooks/deep_learning/dataset/TwoLayerNet_weight.pkl'   
with open(save_file, 'wb') as f:
    pickle.dump(network, f, -1) 

Let's judge the test data using the saved result.

import sys, os
sys.path.append(os.pardir) 
sys.path.append('/content/drive/My Drive/Colab Notebooks/deep_learning/common')
sys.path.append('/content/drive/My Drive/Colab Notebooks/deep_learning/dataset')

import numpy as np
import matplotlib.pyplot as plt
from mnist import load_mnist
from two_layer_net import TwoLayerNet

#Data reading
(x_train, t_train), (x_test, t_test) = load_mnist(normalize=True, one_hot_label=False)

#network = TwoLayerNet(input_size=784, hidden_size=50, output_size=10)
import pickle
weight_file = '/content/drive/My Drive/Colab Notebooks/deep_learning/dataset/TwoLayerNet_weight.pkl'
with open(weight_file, 'rb') as f:
    network = pickle.load(f)

#Check the contents of the identification result
import matplotlib.pyplot as plt

def showImg(x):
    example = x.reshape((28, 28))
    plt.figure()
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    plt.imshow(example)
    plt.show()
    return

test_size = 10
test_mask = np.random.choice(10000, test_size)
x = x_test[test_mask]
t = t_test[test_mask]

for i in range(10):
    y = network.predict(x[i])
    p= np.argmax(y)
    print("Correct answer" + str(t[i]))
    print("Judgment[ " + str(p) + " ]")
    count = 0
    for v in y:
        print("["+str(count)+"] {:.2%}".format(v))
        count += 1
    showImg(x[i])

Occasionally I get a ModuleNotFoundError, but I've confirmed that the book program works.

I also moved train_convnet.py in Chapter 7

Running train_convnet.py in Chapter 7 in interactive mode required the following tweaks:

Modify the import statement of simple_convnet.py in the folder ch07 and set it in the folder common.

simple_convnet.py

from common.layers import *
from common.gradient import numerical_gradient

#Modify the import statement as follows and set it to common
from layers import *
from gradient import numerical_gradient

Modify the import statement of the following modules in common in the same way layers.py trainer.py

Create a new Notebook, mount the drive and set the path.

from google.colab import drive
drive.mount('/content/drive')

import sys, os
sys.path.append(os.pardir)  #Settings for importing files in the parent directory
sys.path.append('/content/drive/My Drive/Colab Notebooks/deep_learning/common')
sys.path.append('/content/drive/My Drive/Colab Notebooks/deep_learning/dataset')

Execute the train_convnet.py script in order. However, modify the import statement before executing it.

from mnist import load_mnist
from simple_convnet import SimpleConvNet
from trainer import Trainer

I was able to execute it in the same procedure as the program in Chapter 4.

Part 13 ← Click here for the table of contents of the memo Unreadable Glossary