import torch 
import os
import torch.nn as nn
import torchvision.datasets as dsets
import torchvision.transforms as transforms
from torch.autograd import Variable
import argparse
# os.environ['CUDA_VISIBLE_DEVICES'] = "4"
parser = argparse.ArgumentParser(description='cnn_mnist')
    # learning & saving hyper parameterss
parser.add_argument('-train', default = False, help = 'train the model')
parser.add_argument('-test', default = True, help = 'test the model')
parser.add_argument('-learning_rate', type = float, default = 0.001, help = 'initial learning rate [default = 0.005')
parser.add_argument('-num_epochs', type = int, default = 5, help = 'number of epochs of training [default = 10')
parser.add_argument('-batch_size', type = int, default = 100, help = 'batch size for training')
parser.add_argument('-input_size', type = int, default = 784, help = 'input size')
parser.add_argument('-hidden_size', type = int, default = 500, help = 'hidden size')
parser.add_argument('-output_size',type = int, default = 1, help = 'output size')
parser.add_argument('-num_classes', type = int, default = 10, help = 'hidden layer number')
parser.add_argument('-cuda',default = False, help = 'enable gpu')
args = parser.parse_args()
# Hyper Parameters 
input_size = args.input_size
hidden_size = args.hidden_size
num_classes = args.num_classes
num_epochs = args.num_epochs
batch_size = args.batch_size
learning_rate = args.learning_rate

# MNIST Dataset
train_dataset = dsets.MNIST(root='./data/',
                            train=True, 
                            transform=transforms.ToTensor(),
                            download=False)

test_dataset = dsets.MNIST(root='./data/',
                           train=False, 
                           transform=transforms.ToTensor())

# Data Loader (Input Pipeline)
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size, 
                                           shuffle=True)

test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size, 
                                          shuffle=False)


# CNN Model (2 conv layer)
class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        # 100 x 1 x 28 x 28
        self.layer1 = nn.Sequential(
            nn.Conv2d(1, 16, kernel_size=5, padding=2),
            # class torch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True)
            nn.BatchNorm2d(16),
            nn.ReLU(),
            nn.MaxPool2d(2))
        # 100 x 16 x 14 x 14
        self.layer2 = nn.Sequential(
            nn.Conv2d(16, 32, kernel_size=5, padding=2),
            nn.BatchNorm2d(32),
            nn.ReLU(),
            nn.MaxPool2d(2))
        # 100 x 32 x 7 x 7
        # 100 x 1568
        self.fc = nn.Linear(7*7*32, 10)
        # 10 catagory
    def forward(self, x):
        out = self.layer1(x)
        out = self.layer2(out)
        out = out.view(out.size(0), -1) # view means reshape
        # 100 x 32 x 7 x 7 ==> 100 x 1568
        out = self.fc(out)
        return out
        
cnn = CNN()
if args.train:
    # Loss and Optimizer
    criterion = nn.CrossEntropyLoss()
    optimizer = torch.optim.Adam(cnn.parameters(), lr=learning_rate)

    # Train the Model
    for epoch in range(num_epochs):
        for i, (images, labels) in enumerate(train_loader):
            images = Variable(images)
            labels = Variable(labels)
            
            # Forward + Backward + Optimize
            optimizer.zero_grad()
            outputs = cnn(images)
            loss = criterion(outputs, labels)
            loss.backward()
            optimizer.step()
            
            if (i+1) % 100 == 0:
                print ('Epoch [%d/%d], Iter [%d/%d] Loss: %.4f' 
                       %(epoch+1, num_epochs, i+1, len(train_dataset)//batch_size, loss.data[0]))
    # Save the Trained Model
    torch.save(cnn.state_dict(), 'cnn.pkl')

if args.test:
    cnn.load_state_dict(torch.load('cnn.pkl')) 
    # Test Accuracy of the model on the 10000 test images: 99.05 %
    # [Finished in 4.8s]
    # Test the Model
    cnn.eval()  # Change model to 'eval' mode (BN uses moving mean/var).
    correct = 0.0
    total = 0.0
    for images, labels in test_loader:
        # image size is 100x1x28x28
        images = Variable(images)
        outputs = cnn(images) # 100 x 10
        _, predicted = torch.max(outputs.data, 1) # predicted.size() = 100 x 1
        """
        'predicted' is the second return 
        value of torch.max function, which is the 
        maximum value of found(argmax)
        """
        """
        outputs.data is the distribution of the 10 outputs, 
        each is a vote of probability,
         we want the highest value's index 
        """
        total += labels.size(0) # 100
        correct += (predicted == labels).sum()
        # print(outputs.size())
        x = outputs.data
        print(x)
        # print(predicted)
        # print(labels.size())
        break;

    print('Test Accuracy of the model on the 10000 test images: %.2lf %%' % (100.0 * correct / total))