import numpy as np
import torch
from torch import nn
from torch import optim
from torchvision import datasets, transforms, models
data_dir = 'F:/Senya/Dataset'
def load_split_train_test(datadir, valid_size = .1):
    train_transforms = transforms.Compose([transforms.Resize(64),
                                           transforms.RandomHorizontalFlip(),
                                           transforms.ToTensor(),
                                       ])
    test_transforms = transforms.Compose([transforms.Resize(64),
                                      transforms.ToTensor(),
                                      ])
    train_data = datasets.ImageFolder(datadir,
                    transform=train_transforms)
    test_data = datasets.ImageFolder(datadir,
                    transform=test_transforms)
    num_train = len(train_data)
    indices = list(range(num_train))
    split = int(np.floor(valid_size * num_train))
    np.random.shuffle(indices)
    from torch.utils.data.sampler import SubsetRandomSampler
    train_idx, test_idx = indices[split:], indices[:split]
    train_sampler = SubsetRandomSampler(train_idx)
    test_sampler = SubsetRandomSampler(test_idx)
    trainloader = torch.utils.data.DataLoader(train_data,
                   sampler=train_sampler, batch_size=64)
    testloader = torch.utils.data.DataLoader(test_data,
                   sampler=test_sampler, batch_size=64)
    return trainloader, testloader
trainloader, testloader = load_split_train_test(data_dir, .1)
print(trainloader.dataset.classes)
device = torch.device("cuda" if torch.cuda.is_available()
                                  else "cpu")
model = models.mobilenet_v2(pretrained=True)
model.classifier = nn.Sequential(nn.Linear(1280, 3),
                                 nn.LogSoftmax(dim=1))
print(model)
criterion = nn.NLLLoss()
optimizer = optim.Adam(model.parameters(), lr=0.003)
model.to(device)
epochs = 5
steps = 0
running_loss = 0
print_every = 10
train_losses, test_losses = [], []
for epoch in range(epochs):
    for inputs, labels in trainloader:
        steps += 1
        inputs, labels = inputs.to(device), labels.to(device)
        optimizer.zero_grad()
        logps = model.forward(inputs)
        loss = criterion(logps, labels)
        loss.backward()
        optimizer.step()
        running_loss += loss.item()
        if steps % print_every == 0:
            test_loss = 0
            accuracy = 0
            model.eval()
            with torch.no_grad():
                for inputs, labels in testloader:
                    inputs, labels = inputs.to(device), labels.to(device)
                    logps = model.forward(inputs)
                    batch_loss = criterion(logps, labels)
                    test_loss += batch_loss.item()
                    ps = torch.exp(logps)
                    top_p, top_class = ps.topk(1, dim=1)
                    equals = top_class == labels.view(*top_class.shape)
                    accuracy += torch.mean(equals.type(torch.FloatTensor)).item()
            train_losses.append(running_loss / len(trainloader))
            test_losses.append(test_loss / len(testloader))
            print(f"Epoch {epoch + 1}/{epochs}.. "
                  f"Train loss: {running_loss / print_every:.3f}.. "
                  f"Test loss: {test_loss / len(testloader):.3f}.. "
                  f"Test accuracy: {accuracy / len(testloader):.3f}")
            running_loss = 0
            model.train()
torch.save(model, 'EyeDetector.pth')