TD2: Deep learning

In this TD, you must modify this notebook to answer the questions. To do this,

1. Fork this repository
2. Clone your forked repository on your local computer
3. Answer the questions
4. Commit and push regularly

The last commit is due on Sunday, December 1, 11:59 PM. Later commits will not be taken into account.

Install and test PyTorch from https://pytorch.org/get-started/locally.

#%pip install torch torchvision

To test run the following code

import torch

N, D = 14, 10
x = torch.randn(N, D).type(torch.FloatTensor)
print(x)

from torchvision import models

alexnet = models.alexnet()
print(alexnet)

tensor([[-9.6164e-01,  7.5175e-01,  2.2068e-01, -1.9287e+00,  1.2018e+00,
          8.3169e-01,  4.4855e-01, -3.9142e-01, -5.6906e-01,  2.7513e-02],
        [-1.7256e+00,  1.1839e+00,  6.1850e-01,  4.0442e-01, -2.4731e-03,
          2.0747e-01, -8.9610e-01,  1.0929e+00,  5.2957e-01,  1.2756e+00],
        [-1.4138e+00,  4.2878e-01, -2.5560e-01, -5.0247e-01,  3.3962e-01,
          7.6361e-01, -3.9553e-01,  3.0112e-01,  8.0323e-01, -6.5776e-01],
        [-3.8860e-01, -7.1832e-01,  9.4326e-02,  1.9248e+00, -1.7670e+00,
         -2.9427e-01,  1.5313e+00, -1.3358e+00,  9.3929e-01, -7.3859e-01],
        [-1.1374e+00,  8.9347e-02,  5.6253e-01,  1.1770e+00,  1.4455e+00,
         -3.0582e-01, -1.2385e+00, -3.1167e+00,  1.5320e+00,  8.1779e-01],
        [-4.0280e-01, -6.0394e-01, -3.3574e-01,  5.9772e-01,  9.9111e-01,
          1.9502e+00,  8.0123e-01,  5.8141e-01, -2.9559e-02,  5.2823e-01],
        [-2.6291e-01,  1.4760e-01,  7.5081e-01, -1.3812e+00,  3.8280e-01,
          7.1685e-02,  6.5621e-01,  1.7026e+00,  5.1833e-02, -7.5081e-01],
        [-1.1018e+00, -1.1544e+00,  7.6059e-01, -7.3267e-01,  1.4962e+00,
          3.2975e-01,  1.4480e+00, -1.2683e+00, -1.0807e+00,  1.2137e+00],
        [ 4.5243e-01,  3.2494e-01,  1.8584e+00,  9.1640e-01, -5.0231e-01,
          2.2226e-01,  9.6508e-01, -1.3855e+00, -4.0333e-01,  2.0664e-01],
        [-8.7366e-01,  5.7875e-01,  6.7228e-01,  1.2443e+00,  5.0369e-02,
         -7.0752e-02, -9.4983e-03, -1.0352e+00, -8.9707e-01, -8.4242e-01],
        [-5.1466e-01,  9.4534e-01, -1.5424e+00,  9.0490e-01,  2.2420e+00,
          1.3246e-01, -8.2010e-01,  1.0778e-02,  8.8466e-02,  1.1311e-01],
        [ 1.3198e-02,  1.3362e-01,  4.8558e-01,  3.6620e-01, -2.1507e+00,
          1.7068e+00,  4.3327e-01,  3.7356e-02, -9.8956e-01, -7.9939e-01],
        [ 1.1221e+00, -8.4034e-01, -2.1407e+00,  4.1205e-01, -8.2332e-01,
         -1.2031e+00, -1.1830e+00,  5.4147e-02,  1.0012e-01, -3.1518e-01],
        [ 5.3398e-01, -3.4557e-01,  5.1655e-01,  9.6549e-01,  1.0638e+00,
          7.1804e-02,  2.5193e-01,  6.3295e-02, -7.5503e-01,  2.4725e-01]])

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

AlexNet(
  (features): Sequential(
    (0): Conv2d(3, 64, kernel_size=(11, 11), stride=(4, 4), padding=(2, 2))
    (1): ReLU(inplace=True)
    (2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
    (3): Conv2d(64, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (4): ReLU(inplace=True)
    (5): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
    (6): Conv2d(192, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (7): ReLU(inplace=True)
    (8): Conv2d(384, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (9): ReLU(inplace=True)
    (10): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (11): ReLU(inplace=True)
    (12): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(6, 6))
  (classifier): Sequential(
    (0): Dropout(p=0.5, inplace=False)
    (1): Linear(in_features=9216, out_features=4096, bias=True)
    (2): ReLU(inplace=True)
    (3): Dropout(p=0.5, inplace=False)
    (4): Linear(in_features=4096, out_features=4096, bias=True)
    (5): ReLU(inplace=True)
    (6): Linear(in_features=4096, out_features=1000, bias=True)
  )
)

Exercise 1: CNN on CIFAR10

The goal is to apply a Convolutional Neural Net (CNN) model on the CIFAR10 image dataset and test the accuracy of the model on the basis of image classification. Compare the Accuracy VS the neural network implemented during TD1.

Have a look at the following documentation to be familiar with PyTorch.

https://pytorch.org/tutorials/beginner/pytorch_with_examples.html

https://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html

You can test if GPU is available on your machine and thus train on it to speed up the process

import torch

# check if CUDA is available
train_on_gpu = torch.cuda.is_available()

if not train_on_gpu:
    print("CUDA is not available.  Training on CPU ...")
else:
    print("CUDA is available!  Training on GPU ...")

CUDA is not available.  Training on CPU ...

Next we load the CIFAR10 dataset

import numpy as np
from torchvision import datasets, transforms
from torch.utils.data.sampler import SubsetRandomSampler

# number of subprocesses to use for data loading
num_workers = 0
# how many samples per batch to load
batch_size = 20
# percentage of training set to use as validation
valid_size = 0.2

# convert data to a normalized torch.FloatTensor
transform = transforms.Compose(
    [transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]
)

# choose the training and test datasets
train_data = datasets.CIFAR10("data", train=True, download=True, transform=transform)
test_data = datasets.CIFAR10("data", train=False, download=True, transform=transform)

# obtain training indices that will be used for validation
num_train = len(train_data)
indices = list(range(num_train))
np.random.shuffle(indices)
split = int(np.floor(valid_size * num_train))
train_idx, valid_idx = indices[split:], indices[:split]

# define samplers for obtaining training and validation batches
train_sampler = SubsetRandomSampler(train_idx)
valid_sampler = SubsetRandomSampler(valid_idx)

# prepare data loaders (combine dataset and sampler)
train_loader = torch.utils.data.DataLoader(
    train_data, batch_size=batch_size, sampler=train_sampler, num_workers=num_workers
)
valid_loader = torch.utils.data.DataLoader(
    train_data, batch_size=batch_size, sampler=valid_sampler, num_workers=num_workers
)
test_loader = torch.utils.data.DataLoader(
    test_data, batch_size=batch_size, num_workers=num_workers
)

# specify the image classes
classes = [
    "airplane",
    "automobile",
    "bird",
    "cat",
    "deer",
    "dog",
    "frog",
    "horse",
    "ship",
    "truck",
]

Files already downloaded and verified
Files already downloaded and verified

CNN definition (this one is an example)

import torch.nn as nn
import torch.nn.functional as F

# define the CNN architecture


class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.pool = nn.MaxPool2d(2, 2)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.fc1 = nn.Linear(16 * 5 * 5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 16 * 5 * 5)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x


# create a complete CNN
model = Net()
print(model)
# move tensors to GPU if CUDA is available
if train_on_gpu:
    model.cuda()

Net(
  (conv1): Conv2d(3, 6, kernel_size=(5, 5), stride=(1, 1))
  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (conv2): Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1))
  (fc1): Linear(in_features=400, out_features=120, bias=True)
  (fc2): Linear(in_features=120, out_features=84, bias=True)
  (fc3): Linear(in_features=84, out_features=10, bias=True)
)

Loss function and training using SGD (Stochastic Gradient Descent) optimizer

import torch.optim as optim


def training(model_,name):
    criterion = nn.CrossEntropyLoss()  # specify loss function
    optimizer = optim.SGD(model_.parameters(), lr=0.01)  # specify optimizer

    n_epochs = 30  # number of epochs to train the model
    train_loss_list = []  # list to store loss to visualize
    valid_loss_min = np.Inf  # track change in validation loss

    for epoch in range(n_epochs):
        # Keep track of training and validation loss
        train_loss = 0.0
        valid_loss = 0.0

        # Train the model
        model_.train()
        for data, target in train_loader:
            # Move tensors to GPU if CUDA is available
            if train_on_gpu:
                data, target = data.cuda(), target.cuda()
            # Clear the gradients of all optimized variables
            optimizer.zero_grad()
            # Forward pass: compute predicted outputs by passing inputs to the model
            output = model_(data)
            # Calculate the batch loss
            loss = criterion(output, target)
            # Backward pass: compute gradient of the loss with respect to model parameters
            loss.backward()
            # Perform a single optimization step (parameter update)
            optimizer.step()
            # Update training loss
            train_loss += loss.item() * data.size(0)

        # Validate the model
        model_.eval()
        for data, target in valid_loader:
            # Move tensors to GPU if CUDA is available
            if train_on_gpu:
                data, target = data.cuda(), target.cuda()
            # Forward pass: compute predicted outputs by passing inputs to the model
            output = model_(data)
            # Calculate the batch loss
            loss = criterion(output, target)
            # Update average validation loss
            valid_loss += loss.item() * data.size(0)

        # Calculate average losses
        train_loss = train_loss / len(train_loader)
        valid_loss = valid_loss / len(valid_loader)
        train_loss_list.append(train_loss)

        # Print training/validation statistics
        print(
            "Epoch: {} \tTraining Loss: {:.6f} \tValidation Loss: {:.6f}".format(
                epoch, train_loss, valid_loss
            )
        )

        # Save model if validation loss has decreased
        if valid_loss <= valid_loss_min:
            print(
                "Validation loss decreased ({:.6f} --> {:.6f}).  Saving model ...".format(
                    valid_loss_min, valid_loss
                )
            )
            torch.save(model_.state_dict(), name)
            valid_loss_min = valid_loss
    return train_loss_list,n_epochs

train_loss_list,n_epochs=training(model,"model_cifar.pt")

Epoch: 0 	Training Loss: 43.322449 	Validation Loss: 37.534600
Validation loss decreased (inf --> 37.534600).  Saving model ...
Epoch: 1 	Training Loss: 34.643701 	Validation Loss: 31.136308
Validation loss decreased (37.534600 --> 31.136308).  Saving model ...
Epoch: 2 	Training Loss: 30.696749 	Validation Loss: 29.305418
Validation loss decreased (31.136308 --> 29.305418).  Saving model ...
Epoch: 3 	Training Loss: 28.533169 	Validation Loss: 28.138255
Validation loss decreased (29.305418 --> 28.138255).  Saving model ...
Epoch: 4 	Training Loss: 26.984089 	Validation Loss: 26.008851
Validation loss decreased (28.138255 --> 26.008851).  Saving model ...
Epoch: 5 	Training Loss: 25.598758 	Validation Loss: 24.788705
Validation loss decreased (26.008851 --> 24.788705).  Saving model ...
Epoch: 6 	Training Loss: 24.410424 	Validation Loss: 24.675400
Validation loss decreased (24.788705 --> 24.675400).  Saving model ...
Epoch: 7 	Training Loss: 23.331779 	Validation Loss: 24.064556
Validation loss decreased (24.675400 --> 24.064556).  Saving model ...
Epoch: 8 	Training Loss: 22.372174 	Validation Loss: 23.541685
Validation loss decreased (24.064556 --> 23.541685).  Saving model ...
Epoch: 9 	Training Loss: 21.532750 	Validation Loss: 22.832895
Validation loss decreased (23.541685 --> 22.832895).  Saving model ...
Epoch: 10 	Training Loss: 20.801076 	Validation Loss: 22.783772
Validation loss decreased (22.832895 --> 22.783772).  Saving model ...
Epoch: 11 	Training Loss: 20.057448 	Validation Loss: 21.981297
Validation loss decreased (22.783772 --> 21.981297).  Saving model ...
Epoch: 12 	Training Loss: 19.406093 	Validation Loss: 21.886178
Validation loss decreased (21.981297 --> 21.886178).  Saving model ...
Epoch: 13 	Training Loss: 18.735991 	Validation Loss: 21.499771
Validation loss decreased (21.886178 --> 21.499771).  Saving model ...
Epoch: 14 	Training Loss: 18.176201 	Validation Loss: 22.818723
Epoch: 15 	Training Loss: 17.577495 	Validation Loss: 22.793243
Epoch: 16 	Training Loss: 17.008231 	Validation Loss: 21.628658
Epoch: 17 	Training Loss: 16.398250 	Validation Loss: 22.139813
Epoch: 18 	Training Loss: 15.941069 	Validation Loss: 22.418324
Epoch: 19 	Training Loss: 15.427645 	Validation Loss: 22.229620
Epoch: 20 	Training Loss: 14.983954 	Validation Loss: 22.178547
Epoch: 21 	Training Loss: 14.419324 	Validation Loss: 23.024117
Epoch: 22 	Training Loss: 14.000020 	Validation Loss: 23.038742
Epoch: 23 	Training Loss: 13.525463 	Validation Loss: 22.881500
Epoch: 24 	Training Loss: 12.990559 	Validation Loss: 23.842180
Epoch: 25 	Training Loss: 12.650270 	Validation Loss: 23.820795
Epoch: 26 	Training Loss: 12.138135 	Validation Loss: 24.533619
Epoch: 27 	Training Loss: 11.802696 	Validation Loss: 25.346910
Epoch: 28 	Training Loss: 11.412316 	Validation Loss: 25.572966
Epoch: 29 	Training Loss: 10.919574 	Validation Loss: 26.247145

Does overfit occur? If so, do an early stopping.

Oui il y a de l'overfitting. A parir de l'epoche 11.

import matplotlib.pyplot as plt

plt.plot(range(n_epochs), train_loss_list)
plt.xlabel("Epoch")
plt.ylabel("Loss")
plt.title("Performance of Model 1")
plt.show()

Now loading the model with the lowest validation loss value

#on définit la fonction test
def test(model_):

    criterion = nn.CrossEntropyLoss()  # specify loss function

    # track test loss
    test_loss = 0.0
    class_correct = list(0.0 for i in range(10))
    class_total = list(0.0 for i in range(10))

    model_.eval()
    # iterate over test data
    for data, target in test_loader:
        # move tensors to GPU if CUDA is available
        if train_on_gpu:
            data, target = data.cuda(), target.cuda()
        # forward pass: compute predicted outputs by passing inputs to the model
        output = model_(data)
        # calculate the batch loss
        loss = criterion(output, target)
        # update test loss
        test_loss += loss.item() * data.size(0)
        # convert output probabilities to predicted class
        _, pred = torch.max(output, 1)
        # compare predictions to true label
        correct_tensor = pred.eq(target.data.view_as(pred))
        correct = (
            np.squeeze(correct_tensor.numpy())
            if not train_on_gpu
            else np.squeeze(correct_tensor.cpu().numpy())
        )
        # calculate test accuracy for each object class
        for i in range(batch_size):
            label = target.data[i]
            class_correct[label] += correct[i].item()
            class_total[label] += 1

    # average test loss
    test_loss = test_loss / len(test_loader)
    print("Test Loss: {:.6f}\n".format(test_loss))


    print(class_total)
    for i in range(10):
        if class_total[i] > 0:
            print(
                "Test Accuracy of %5s: %2d%% (%2d/%2d)"
                % (
                    classes[i],
                    100 * class_correct[i] / class_total[i],
                    np.sum(class_correct[i]),
                    np.sum(class_total[i]),
                )
            )
        else:
            print("Test Accuracy of %5s: N/A (no training examples)" % (classes[i]))

    print(
        "\nTest Accuracy (Overall): %2d%% (%2d/%2d)"
        % (
            100.0 * np.sum(class_correct) / np.sum(class_total),
            np.sum(class_correct),
            np.sum(class_total),
        )
    )   
    return 

# on load le model
model.load_state_dict(torch.load("./model_cifar.pt"))

#on teste le modele
test(model)

Test Loss: 21.873704

[1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0]
Test Accuracy of airplane: 61% (613/1000)
Test Accuracy of automobile: 68% (686/1000)
Test Accuracy of  bird: 57% (573/1000)
Test Accuracy of   cat: 47% (477/1000)
Test Accuracy of  deer: 48% (486/1000)
Test Accuracy of   dog: 45% (450/1000)
Test Accuracy of  frog: 64% (648/1000)
Test Accuracy of horse: 70% (702/1000)
Test Accuracy of  ship: 81% (817/1000)
Test Accuracy of truck: 71% (711/1000)

Test Accuracy (Overall): 61% (6163/10000)

Build a new network with the following structure.

• It has 3 convolutional layers of kernel size 3 and padding of 1.
• The first convolutional layer must output 16 channels, the second 32 and the third 64.
• At each convolutional layer output, we apply a ReLU activation then a MaxPool with kernel size of 2.
• Then, three fully connected layers, the first two being followed by a ReLU activation and a dropout whose value you will suggest.
• The first fully connected layer will have an output size of 512.
• The second fully connected layer will have an output size of 64.

Compare the results obtained with this new network to those obtained previously.

On crée la nouvelle architecture du réseau CNN de la même manière. On l'appellera model 2.

# define the CNN architecture

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(in_channels=3, out_channels=16, kernel_size=3,padding=1)
        self.pool = nn.MaxPool2d(kernel_size=2)
        self.conv2 = nn.Conv2d(in_channels=16, out_channels=32, kernel_size=3,padding=1)
        self.conv3 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3,padding=1)

        self.fc1 = nn.Linear(64*4*4, 512)
        self.fc2 = nn.Linear(512, 64)
        self.fc3 = nn.Linear(64, 10)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = self.pool(F.relu(self.conv3(x)))

        x = x.view(-1, 64 * 4 * 4)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x


# create a complete CNN
model2 = Net()
print(model)
# move tensors to GPU if CUDA is available
if train_on_gpu:
    model2.cuda()

Net(
  (conv1): Conv2d(3, 6, kernel_size=(5, 5), stride=(1, 1))
  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (conv2): Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1))
  (fc1): Linear(in_features=400, out_features=120, bias=True)
  (fc2): Linear(in_features=120, out_features=84, bias=True)
  (fc3): Linear(in_features=84, out_features=10, bias=True)
)

On entraine maintenant l'algorithme.

#train
train_loss_list,n_epochs=training(model2,"model_cifar_2.pt")

Epoch: 0 	Training Loss: 45.546517 	Validation Loss: 41.965355
Validation loss decreased (inf --> 41.965355).  Saving model ...
Epoch: 1 	Training Loss: 37.506297 	Validation Loss: 32.462201
Validation loss decreased (41.965355 --> 32.462201).  Saving model ...
Epoch: 2 	Training Loss: 30.826246 	Validation Loss: 28.965965
Validation loss decreased (32.462201 --> 28.965965).  Saving model ...
Epoch: 3 	Training Loss: 27.725424 	Validation Loss: 26.256559
Validation loss decreased (28.965965 --> 26.256559).  Saving model ...
Epoch: 4 	Training Loss: 25.409111 	Validation Loss: 23.947241
Validation loss decreased (26.256559 --> 23.947241).  Saving model ...
Epoch: 5 	Training Loss: 23.199655 	Validation Loss: 22.190540
Validation loss decreased (23.947241 --> 22.190540).  Saving model ...
Epoch: 6 	Training Loss: 21.171948 	Validation Loss: 21.483634
Validation loss decreased (22.190540 --> 21.483634).  Saving model ...
Epoch: 7 	Training Loss: 19.252010 	Validation Loss: 19.125935
Validation loss decreased (21.483634 --> 19.125935).  Saving model ...
Epoch: 8 	Training Loss: 17.624332 	Validation Loss: 18.777900
Validation loss decreased (19.125935 --> 18.777900).  Saving model ...
Epoch: 9 	Training Loss: 16.178626 	Validation Loss: 17.967978
Validation loss decreased (18.777900 --> 17.967978).  Saving model ...
Epoch: 10 	Training Loss: 14.790585 	Validation Loss: 17.401410
Validation loss decreased (17.967978 --> 17.401410).  Saving model ...
Epoch: 11 	Training Loss: 13.436781 	Validation Loss: 16.391732
Validation loss decreased (17.401410 --> 16.391732).  Saving model ...
Epoch: 12 	Training Loss: 12.091579 	Validation Loss: 18.383929
Epoch: 13 	Training Loss: 10.885964 	Validation Loss: 16.164210
Validation loss decreased (16.391732 --> 16.164210).  Saving model ...
Epoch: 14 	Training Loss: 9.642637 	Validation Loss: 17.355094
Epoch: 15 	Training Loss: 8.420236 	Validation Loss: 18.205283
Epoch: 16 	Training Loss: 7.142842 	Validation Loss: 17.389541
Epoch: 17 	Training Loss: 6.121646 	Validation Loss: 19.769254
Epoch: 18 	Training Loss: 4.971805 	Validation Loss: 20.793741
Epoch: 19 	Training Loss: 4.208495 	Validation Loss: 23.089104
Epoch: 20 	Training Loss: 3.380529 	Validation Loss: 23.717688
Epoch: 21 	Training Loss: 2.708748 	Validation Loss: 25.890190
Epoch: 22 	Training Loss: 2.314826 	Validation Loss: 26.893456
Epoch: 23 	Training Loss: 1.924105 	Validation Loss: 30.196253
Epoch: 24 	Training Loss: 1.555745 	Validation Loss: 28.148667
Epoch: 25 	Training Loss: 1.266747 	Validation Loss: 30.644867
Epoch: 26 	Training Loss: 1.780205 	Validation Loss: 32.042693
Epoch: 27 	Training Loss: 1.415102 	Validation Loss: 33.207941
Epoch: 28 	Training Loss: 1.071195 	Validation Loss: 35.509902
Epoch: 29 	Training Loss: 0.989859 	Validation Loss: 37.337378

On affiche le train lost en fonction du nombre d'époches.

import matplotlib.pyplot as plt

plt.plot(range(n_epochs), train_loss_list)
plt.xlabel("Epoch")
plt.ylabel("Loss")
plt.title("Performance of Model 2")
plt.show()

On calcule la precision du modèle sur une base de données test.

model2.load_state_dict(torch.load("./model_cifar_2.pt"))

test(model2)

Test Loss: 16.504843

[1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0]
Test Accuracy of airplane: 78% (786/1000)
Test Accuracy of automobile: 78% (780/1000)
Test Accuracy of  bird: 62% (625/1000)
Test Accuracy of   cat: 57% (578/1000)
Test Accuracy of  deer: 67% (679/1000)
Test Accuracy of   dog: 51% (514/1000)
Test Accuracy of  frog: 79% (795/1000)
Test Accuracy of horse: 80% (804/1000)
Test Accuracy of  ship: 86% (863/1000)
Test Accuracy of truck: 78% (781/1000)

Test Accuracy (Overall): 72% (7205/10000)

La précision est meilleure pour presque toutes les classes (la classe truck a une précision plus faible).

Au global, la précision avec cette nouvelle architecture est meilleure (71%) que la précedente (62%).

Exercise 2: Quantization: try to compress the CNN to save space

Quantization doc is available from https://pytorch.org/docs/stable/quantization.html#torch.quantization.quantize_dynamic

The Exercise is to quantize post training the above CNN model. Compare the size reduction and the impact on the classification accuracy

The size of the model is simply the size of the file.

import os


def print_size_of_model(model, label=""):
    torch.save(model.state_dict(), "temp.p")
    size = os.path.getsize("temp.p")
    print("model: ", label, " \t", "Size (KB):", size / 1e3)
    os.remove("temp.p")
    return size


print_size_of_model(model, "fp32")

model:  fp32  	 Size (KB): 251.278

251278

La taille initial du modèle 1 est de 251.278 KB

On va compresser ce modèle :

import torch.quantization

quantized_model = torch.quantization.quantize_dynamic(model, dtype=torch.qint8)
print_size_of_model(quantized_model, "int8")

model:  int8  	 Size (KB): 76.522

76522

Le modèle compressé a une taille de 76.522 KB. Il y a donc eu une réduction de 174.756 KB.

# on teste le modèle compressé
test(quantized_model)

Test Loss: 21.884723

[1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0]
Test Accuracy of airplane: 61% (611/1000)
Test Accuracy of automobile: 68% (688/1000)
Test Accuracy of  bird: 56% (568/1000)
Test Accuracy of   cat: 47% (476/1000)
Test Accuracy of  deer: 48% (487/1000)
Test Accuracy of   dog: 45% (450/1000)
Test Accuracy of  frog: 64% (646/1000)
Test Accuracy of horse: 70% (705/1000)
Test Accuracy of  ship: 81% (812/1000)
Test Accuracy of truck: 71% (714/1000)

Test Accuracy (Overall): 61% (6157/10000)

En terme de précision, elle est la même en pourcentage pour presque toutes les classes (en comparaison avec le modèle non compressé).

Elle différe de 1% pour la classe horse.

Au global, la précision est la même.

Le modèle compressé conserve la précision du modèle tout en réduisant sa taille.

---

For each class, compare the classification test accuracy of the initial model and the quantized model. Also give the overall test accuracy for both models.

Try training aware quantization to mitigate the impact on the accuracy (doc available here https://pytorch.org/docs/stable/quantization.html#torch.quantization.quantize_dynamic)

On va essayer le training aware quantization

On recrée un nouveau modèle à partir du modèle 1.

#on crée le modèle quantizé

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
         # QuantStub converts tensors from floating point to quantized
        self.quant = torch.ao.quantization.QuantStub()

        self.conv1 = nn.Conv2d(3, 6, 5)
        self.pool = nn.MaxPool2d(2, 2)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.fc1 = nn.Linear(16 * 5 * 5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

        self.dequant = torch.ao.quantization.DeQuantStub()



    def forward(self, x):
        x = self.quant(x)
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 16 * 5 * 5)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        x = self.dequant(x)
        return x


# create a complete CNN
model_aware_quantization = Net()

model_aware_quantization

Net(
  (quant): QuantStub()
  (conv1): Conv2d(3, 6, kernel_size=(5, 5), stride=(1, 1))
  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (conv2): Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1))
  (fc1): Linear(in_features=400, out_features=120, bias=True)
  (fc2): Linear(in_features=120, out_features=84, bias=True)
  (fc3): Linear(in_features=84, out_features=10, bias=True)
  (dequant): DeQuantStub()
)

#train the model
train_loss_list,n_epochs=training(model_aware_quantization,"model_cifar_aware_quantization.pt")

Epoch: 0 	Training Loss: 44.515622 	Validation Loss: 39.651668
Validation loss decreased (inf --> 39.651668).  Saving model ...
Epoch: 1 	Training Loss: 36.187454 	Validation Loss: 33.535019
Validation loss decreased (39.651668 --> 33.535019).  Saving model ...
Epoch: 2 	Training Loss: 31.515289 	Validation Loss: 29.458190
Validation loss decreased (33.535019 --> 29.458190).  Saving model ...
Epoch: 3 	Training Loss: 28.995076 	Validation Loss: 27.946637
Validation loss decreased (29.458190 --> 27.946637).  Saving model ...
Epoch: 4 	Training Loss: 27.089252 	Validation Loss: 26.622712
Validation loss decreased (27.946637 --> 26.622712).  Saving model ...
Epoch: 5 	Training Loss: 25.505706 	Validation Loss: 25.257544
Validation loss decreased (26.622712 --> 25.257544).  Saving model ...
Epoch: 6 	Training Loss: 24.208821 	Validation Loss: 24.149545
Validation loss decreased (25.257544 --> 24.149545).  Saving model ...
Epoch: 7 	Training Loss: 23.149960 	Validation Loss: 23.239544
Validation loss decreased (24.149545 --> 23.239544).  Saving model ...
Epoch: 8 	Training Loss: 22.233548 	Validation Loss: 22.827516
Validation loss decreased (23.239544 --> 22.827516).  Saving model ...
Epoch: 9 	Training Loss: 21.358925 	Validation Loss: 23.077734
Epoch: 10 	Training Loss: 20.627038 	Validation Loss: 21.895032
Validation loss decreased (22.827516 --> 21.895032).  Saving model ...
Epoch: 11 	Training Loss: 19.869867 	Validation Loss: 21.903653
Epoch: 12 	Training Loss: 19.227587 	Validation Loss: 21.835049
Validation loss decreased (21.895032 --> 21.835049).  Saving model ...
Epoch: 13 	Training Loss: 18.584631 	Validation Loss: 21.379109
Validation loss decreased (21.835049 --> 21.379109).  Saving model ...
Epoch: 14 	Training Loss: 17.965375 	Validation Loss: 22.105847
Epoch: 15 	Training Loss: 17.418559 	Validation Loss: 21.119945
Validation loss decreased (21.379109 --> 21.119945).  Saving model ...
Epoch: 16 	Training Loss: 16.785295 	Validation Loss: 21.018267
Validation loss decreased (21.119945 --> 21.018267).  Saving model ...
Epoch: 17 	Training Loss: 16.285102 	Validation Loss: 21.647233
Epoch: 18 	Training Loss: 15.765099 	Validation Loss: 21.452533
Epoch: 19 	Training Loss: 15.246105 	Validation Loss: 21.444231
Epoch: 20 	Training Loss: 14.740839 	Validation Loss: 22.717484
Epoch: 21 	Training Loss: 14.231403 	Validation Loss: 21.937744
Epoch: 22 	Training Loss: 13.831682 	Validation Loss: 22.649114
Epoch: 23 	Training Loss: 13.302803 	Validation Loss: 22.911363
Epoch: 24 	Training Loss: 12.881577 	Validation Loss: 23.119514
Epoch: 25 	Training Loss: 12.434679 	Validation Loss: 24.015339
Epoch: 26 	Training Loss: 11.999609 	Validation Loss: 23.906202
Epoch: 27 	Training Loss: 11.535745 	Validation Loss: 24.581035
Epoch: 28 	Training Loss: 11.174014 	Validation Loss: 25.033342
Epoch: 29 	Training Loss: 10.767195 	Validation Loss: 26.522749

#quantization
model_aware_quantization = torch.quantization.quantize_dynamic(
    model_aware_quantization, {nn.Conv2d, nn.Linear}, dtype=torch.qint8
)

#on affiche la taille du nouveau modele
print_size_of_model(model_aware_quantization, "int8")
# on teste le modèle compressé
print(test(model_aware_quantization))

model:  int8  	 Size (KB): 76.522
Test Loss: 28.386250

[1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0, 1000.0]
Test Accuracy of airplane: 59% (596/1000)
Test Accuracy of automobile: 66% (666/1000)
Test Accuracy of  bird: 50% (500/1000)
Test Accuracy of   cat: 31% (313/1000)
Test Accuracy of  deer: 57% (574/1000)
Test Accuracy of   dog: 54% (549/1000)
Test Accuracy of  frog: 79% (793/1000)
Test Accuracy of horse: 57% (576/1000)
Test Accuracy of  ship: 78% (784/1000)
Test Accuracy of truck: 76% (767/1000)

Test Accuracy (Overall): 61% (6118/10000)
None

La taille du modèle compressé est de 76.522 KB. La précision globale est plus faible par rapport au modèle, elle est de 61%.

Les performances sont équivalentes à la quantization précédente.

Exercise 3: working with pre-trained models.

PyTorch offers several pre-trained models https://pytorch.org/vision/0.8/models.html
We will use ResNet50 trained on ImageNet dataset (https://www.image-net.org/index.php). Use the following code with the files imagenet-simple-labels.json that contains the imagenet labels and the image dog.png that we will use as test.

La prédiction d'une image se fait en plusieurs étapes :

-On resize l'image a la taille 224.

-On transforme en tenseur.

-On normalise

-On charge le modèle pre-entrainé.

-On fait la prédiction.

Experiments:

Study the code and the results obtained. Possibly add other images downloaded from the internet.

What is the size of the model? Quantize it and then check if the model is still able to correctly classify the other images.

Experiment with other pre-trained CNN models.

import json
from PIL import Image

# Choose an image to pass through the model
test_image = "dog.png"

# Configure matplotlib for pretty inline plots
#%matplotlib inline
#%config InlineBackend.figure_format = 'retina'

# Prepare the labels
with open("imagenet-simple-labels.json") as f:
    labels = json.load(f)

# First prepare the transformations: resize the image to what the model was trained on and convert it to a tensor
data_transform = transforms.Compose(
    [
        transforms.Resize((224, 224)),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
    ]
)
# Load the image

image = Image.open(test_image)
plt.imshow(image), plt.xticks([]), plt.yticks([])

print(image)

# Now apply the transformation, expand the batch dimension, and send the image to the GPU
# image = data_transform(image).unsqueeze(0).cuda()
image = data_transform(image).unsqueeze(0)

# Download the model if it's not there already. It will take a bit on the first run, after that it's fast
model = models.resnet50(pretrained=True)
# Send the model to the GPU
# model.cuda()
# Set layers such as dropout and batchnorm in evaluation mode
model.eval()

# Get the 1000-dimensional model output
out = model(image)
# Find the predicted class
print("Predicted class is: {}".format(labels[out.argmax()]))

<PIL.PngImagePlugin.PngImageFile image mode=RGB size=188x282 at 0x121968FD0>

/Users/franck/Library/Python/3.8/lib/python/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet50_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet50_Weights.DEFAULT` to get the most up-to-date weights.
  warnings.warn(msg)

Predicted class is: Golden Retriever

On va tester ce modèle sur une autre image : un king crab.

Il faudra bien que l'image soit une image RGB.

image_2 = Image.open("king_crab.png")
# conversion en image RGB
image_2 = image_2.convert("RGB")
plt.imshow(image_2), plt.xticks([]), plt.yticks([])

# Now apply the transformation, expand the batch dimension, and send the image to the GPU
# image = data_transform(image).unsqueeze(0).cuda()
image_2 = data_transform(image_2).unsqueeze(0)

# Get the 1000-dimensional model output
out = model(image_2)
# Find the predicted class
print("Predicted class is: {}".format(labels[out.argmax()]))

Predicted class is: red king crab

Les classifications pour le Golden Retriever et le King Crab sont bonnes. Le modèle semble être bon.

---

Etudions la taille du modèle

#taille du modèle
print_size_of_model(model, "int8")
#quantization
quantized_model_pretrained = torch.quantization.quantize_dynamic(model, dtype=torch.qint8)
# taille du nouveau modèle 
print_size_of_model(quantized_model_pretrained, "int8")

model:  int8  	 Size (KB): 102523.238
model:  int8  	 Size (KB): 96379.996

96379996

Le modèle pré-entrainé a une taille de 102523.238 KB. Après compression sa taille est de 96379.996 KB.

out = quantized_model_pretrained(image)
# Find the predicted class
print("Predicted class is: {}".format(labels[out.argmax()]))

out_2 = quantized_model_pretrained(image_2)
# Find the predicted class
print("Predicted class is: {}".format(labels[out_2.argmax()]))

Predicted class is: Golden Retriever
Predicted class is: red king crab

Les prédictions sont toujours bonne avec le modèle compressé.

alexnet = models.alexnet()

# Set layers such as dropout and batchnorm in evaluation mode
alexnet.eval()

out = quantized_model_pretrained(image)
# Find the predicted class
print("Predicted class is: {}".format(labels[out.argmax()]))

out = alexnet(image_2)
print("Predicted class is: {}".format(labels[out_2.argmax()]))

Predicted class is: Golden Retriever
Predicted class is: red king crab

Exercise 4: Transfer Learning

For this work, we will use a pre-trained model (ResNet18) as a descriptor extractor and will refine the classification by training only the last fully connected layer of the network. Thus, the output layer of the pre-trained network will be replaced by a layer adapted to the new classes to be recognized which will be in our case ants and bees. Download and unzip in your working directory the dataset available at the address :

https://download.pytorch.org/tutorial/hymenoptera_data.zip

Execute the following code in order to display some images of the dataset.

import os

import matplotlib.pyplot as plt
import numpy as np
import torch
import torchvision
from torchvision import datasets, transforms

# Data augmentation and normalization for training
# Just normalization for validation
data_transforms = {
    "train": transforms.Compose(
        [
            transforms.RandomResizedCrop(
                224
            ),  # ImageNet models were trained on 224x224 images
            transforms.RandomHorizontalFlip(),  # flip horizontally 50% of the time - increases train set variability
            transforms.ToTensor(),  # convert it to a PyTorch tensor
            transforms.Normalize(
                [0.485, 0.456, 0.406], [0.229, 0.224, 0.225]
            ),  # ImageNet models expect this norm
        ]
    ),
    "val": transforms.Compose(
        [
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
        ]
    ),
}

data_dir = "hymenoptera_data"
# Create train and validation datasets and loaders
image_datasets = {
    x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x])
    for x in ["train", "val"]
}
dataloaders = {
    x: torch.utils.data.DataLoader(
        image_datasets[x], batch_size=4, shuffle=True, num_workers=0
    )
    for x in ["train", "val"]
}
dataset_sizes = {x: len(image_datasets[x]) for x in ["train", "val"]}
class_names = image_datasets["train"].classes
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# Helper function for displaying images
def imshow(inp, title=None):
    """Imshow for Tensor."""
    inp = inp.numpy().transpose((1, 2, 0))
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])

    # Un-normalize the images
    inp = std * inp + mean
    # Clip just in case
    inp = np.clip(inp, 0, 1)
    plt.imshow(inp)
    if title is not None:
        plt.title(title)
    plt.pause(0.001)  # pause a bit so that plots are updated
    plt.show()


# Get a batch of training data
inputs, classes = next(iter(dataloaders["train"]))

# Make a grid from batch
out = torchvision.utils.make_grid(inputs)

imshow(out, title=[class_names[x] for x in classes])

Now, execute the following code which uses a pre-trained model ResNet18 having replaced the output layer for the ants/bees classification and performs the model training by only changing the weights of this output layer.

import copy
import os
import time

import matplotlib.pyplot as plt
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
from torch.optim import lr_scheduler
from torchvision import datasets, transforms

# Data augmentation and normalization for training
# Just normalization for validation
data_transforms = {
    "train": transforms.Compose(
        [
            transforms.RandomResizedCrop(
                224
            ),  # ImageNet models were trained on 224x224 images
            transforms.RandomHorizontalFlip(),  # flip horizontally 50% of the time - increases train set variability
            transforms.ToTensor(),  # convert it to a PyTorch tensor
            transforms.Normalize(
                [0.485, 0.456, 0.406], [0.229, 0.224, 0.225]
            ),  # ImageNet models expect this norm
        ]
    ),
    "val": transforms.Compose(
        [
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
        ]
    ),
}

data_dir = "hymenoptera_data"
# Create train and validation datasets and loaders
image_datasets = {
    x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x])
    for x in ["train", "val"]
}
dataloaders = {
    x: torch.utils.data.DataLoader(
        image_datasets[x], batch_size=4, shuffle=True, num_workers=4
    )
    for x in ["train", "val"]
}
dataset_sizes = {x: len(image_datasets[x]) for x in ["train", "val"]}
class_names = image_datasets["train"].classes
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# Helper function for displaying images
def imshow(inp, title=None):
    """Imshow for Tensor."""
    inp = inp.numpy().transpose((1, 2, 0))
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])

    # Un-normalize the images
    inp = std * inp + mean
    # Clip just in case
    inp = np.clip(inp, 0, 1)
    plt.imshow(inp)
    if title is not None:
        plt.title(title)
    plt.pause(0.001)  # pause a bit so that plots are updated
    plt.show()


# Get a batch of training data
# inputs, classes = next(iter(dataloaders['train']))

# Make a grid from batch
# out = torchvision.utils.make_grid(inputs)

# imshow(out, title=[class_names[x] for x in classes])
# training


def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    since = time.time()

    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0

    epoch_time = []  # we'll keep track of the time needed for each epoch

    for epoch in range(num_epochs):
        epoch_start = time.time()
        print("Epoch {}/{}".format(epoch + 1, num_epochs))
        print("-" * 10)

        # Each epoch has a training and validation phase
        for phase in ["train", "val"]:
            if phase == "train":
                scheduler.step()
                model.train()  # Set model to training mode
            else:
                model.eval()  # Set model to evaluate mode

            running_loss = 0.0
            running_corrects = 0

            # Iterate over data.
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device)

                # zero the parameter gradients
                optimizer.zero_grad()

                # Forward
                # Track history if only in training phase
                with torch.set_grad_enabled(phase == "train"):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    loss = criterion(outputs, labels)

                    # backward + optimize only if in training phase
                    if phase == "train":
                        loss.backward()
                        optimizer.step()

                # Statistics
                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(preds == labels.data)

            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]

            print("{} Loss: {:.4f} Acc: {:.4f}".format(phase, epoch_loss, epoch_acc))

            # Deep copy the model
            if phase == "val" and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())

        # Add the epoch time
        t_epoch = time.time() - epoch_start
        epoch_time.append(t_epoch)
        print()

    time_elapsed = time.time() - since
    print(
        "Training complete in {:.0f}m {:.0f}s".format(
            time_elapsed // 60, time_elapsed % 60
        )
    )
    print("Best val Acc: {:4f}".format(best_acc))

    # Load best model weights
    model.load_state_dict(best_model_wts)
    return model, epoch_time


# Download a pre-trained ResNet18 model and freeze its weights
model = torchvision.models.resnet18(pretrained=True)
for param in model.parameters():
    param.requires_grad = False

# Replace the final fully connected layer
# Parameters of newly constructed modules have requires_grad=True by default
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 2)
# Send the model to the GPU
model = model.to(device)
# Set the loss function
criterion = nn.CrossEntropyLoss()

# Observe that only the parameters of the final layer are being optimized
optimizer_conv = optim.SGD(model.fc.parameters(), lr=0.001, momentum=0.9)
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_conv, step_size=7, gamma=0.1)
model, epoch_time = train_model(
    model, criterion, optimizer_conv, exp_lr_scheduler, num_epochs=10
)

/Users/franck/Library/Python/3.8/lib/python/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet18_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet18_Weights.DEFAULT` to get the most up-to-date weights.
  warnings.warn(msg)

Epoch 1/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/torch/optim/lr_scheduler.py:136: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`.  Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
  warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.7993 Acc: 0.5943

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.3701 Acc: 0.8696

Epoch 2/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4917 Acc: 0.7951

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2116 Acc: 0.9348

Epoch 3/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.5263 Acc: 0.7295

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2495 Acc: 0.9058

Epoch 4/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.2608 Acc: 0.8893

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.3190 Acc: 0.8841

Epoch 5/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4371 Acc: 0.8197

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2230 Acc: 0.9203

Epoch 6/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.5459 Acc: 0.7828

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.4372 Acc: 0.8333

Epoch 7/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4413 Acc: 0.8074

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.1740 Acc: 0.9493

Epoch 8/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4181 Acc: 0.8238

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.1775 Acc: 0.9348

Epoch 9/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.2983 Acc: 0.8852

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2060 Acc: 0.9203

Epoch 10/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.3563 Acc: 0.8525

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.1771 Acc: 0.9420

Training complete in 13m 36s
Best val Acc: 0.949275

Experiments: Study the code and the results obtained.

Modify the code and add an "eval_model" function to allow the evaluation of the model on a test set (different from the learning and validation sets used during the learning phase). Study the results obtained.

Now modify the code to replace the current classification layer with a set of two layers using a "relu" activation function for the middle layer, and the "dropout" mechanism for both layers. Renew the experiments and study the results obtained.

Apply ther quantization (post and quantization aware) and evaluate impact on model size and accuracy.

---

Dans cette partie, nous allons ajouter un dataset de test. On configure le data_transform pour qu'il prennent en compte les données test contenu dans le dossier test.

On va aussi configurer la sortie avec la nouvelle structure de neurones.

import copy
import os
import time

import matplotlib.pyplot as plt
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
from torch.optim import lr_scheduler
from torchvision import datasets, transforms

# Data augmentation and normalization for training
# Just normalization for validation

# on ajoute les data test
data_transforms = {
    "train": transforms.Compose(
        [
            transforms.RandomResizedCrop(
                224
            ),  # ImageNet models were trained on 224x224 images
            transforms.RandomHorizontalFlip(),  # flip horizontally 50% of the time - increases train set variability
            transforms.ToTensor(),  # convert it to a PyTorch tensor
            transforms.Normalize(
                [0.485, 0.456, 0.406], [0.229, 0.224, 0.225]
            ),  # ImageNet models expect this norm
        ]
    ),
    "val": transforms.Compose(
        [
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
        ]
    ),
    "test": transforms.Compose(
        [
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
        ]
    ),
}

data_dir = "hymenoptera_data"
# Create train and validation datasets and loaders
image_datasets = {
    x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x])
    for x in ["train", "val","test"]
}

dataloaders = {
    x: torch.utils.data.DataLoader(
        image_datasets[x], batch_size=4, shuffle=True, num_workers=4
    )
    for x in ["train", "val","test"]
}
dataset_sizes = {x: len(image_datasets[x]) for x in ["train", "val","test"]}
class_names = image_datasets["train"].classes
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# Helper function for displaying images
def imshow(inp, title=None):
    """Imshow for Tensor."""
    inp = inp.numpy().transpose((1, 2, 0))
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])

    # Un-normalize the images
    inp = std * inp + mean
    # Clip just in case
    inp = np.clip(inp, 0, 1)
    plt.imshow(inp)
    if title is not None:
        plt.title(title)
    plt.pause(0.001)  # pause a bit so that plots are updated
    plt.show()


# Get a batch of training data
# inputs, classes = next(iter(dataloaders['train']))

# Make a grid from batch
# out = torchvision.utils.make_grid(inputs)

# imshow(out, title=[class_names[x] for x in classes])
# training


def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    since = time.time()

    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0

    epoch_time = []  # we'll keep track of the time needed for each epoch

    for epoch in range(num_epochs):
        epoch_start = time.time()
        print("Epoch {}/{}".format(epoch + 1, num_epochs))
        print("-" * 10)

        # Each epoch has a training and validation phase
        for phase in ["train", "val"]:
            if phase == "train":
                scheduler.step()
                model.train()  # Set model to training mode
            else:
                model.eval()  # Set model to evaluate mode

            running_loss = 0.0
            running_corrects = 0

            # Iterate over data.
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device)

                # zero the parameter gradients
                optimizer.zero_grad()

                # Forward
                # Track history if only in training phase
                with torch.set_grad_enabled(phase == "train"):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    loss = criterion(outputs, labels)

                    # backward + optimize only if in training phase
                    if phase == "train":
                        loss.backward()
                        optimizer.step()

                # Statistics
                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(preds == labels.data)


            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]

            print("{} Loss: {:.4f} Acc: {:.4f}".format(phase, epoch_loss, epoch_acc))

            # Deep copy the model
            if phase == "val" and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())

        # Add the epoch time
        t_epoch = time.time() - epoch_start
        epoch_time.append(t_epoch)


    time_elapsed = time.time() - since
    print(
        "Training complete in {:.0f}m {:.0f}s".format(
            time_elapsed // 60, time_elapsed % 60
        )
    )
    print("Best val Acc: {:4f}".format(best_acc))

    # Load best model weights
    model.load_state_dict(best_model_wts)
    return model, epoch_time



# Download a pre-trained ResNet18 model and freeze its weights
model = torchvision.models.resnet18(pretrained=True)
for param in model.parameters():
    param.requires_grad = False

# Replace the final fully connected layer
# Parameters of newly constructed modules have requires_grad=True by default
num_ftrs = model.fc.in_features

# on ajoute les nouvelles couches
model.fc = nn.Sequential(
    nn.Linear(num_ftrs, 256),  # Add a layer with 256 neurons
    nn.ReLU(),                 # Apply ReLU activation function
    nn.Dropout(0.5),           # Apply dropout with a probability of 0.5
    nn.Linear(256, 2)           # Add the final layer with 2 output neurons
)


# Send the model to the GPU
model = model.to(device)
# Set the loss function
criterion = nn.CrossEntropyLoss()

# Observe that only the parameters of the final layer are being optimized
optimizer_conv = optim.SGD(model.fc.parameters(), lr=0.001, momentum=0.9)
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_conv, step_size=7, gamma=0.1)
model, epoch_time = train_model(
    model, criterion, optimizer_conv, exp_lr_scheduler, num_epochs=10
)

Epoch 1/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.6493 Acc: 0.6189

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.5245 Acc: 0.7174
Epoch 2/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.5375 Acc: 0.7254

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.3046 Acc: 0.9203
Epoch 3/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.5194 Acc: 0.7500

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2810 Acc: 0.8986
Epoch 4/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4514 Acc: 0.7910

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2425 Acc: 0.9275
Epoch 5/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4021 Acc: 0.7910

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2733 Acc: 0.8841
Epoch 6/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4881 Acc: 0.7500

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2048 Acc: 0.9420
Epoch 7/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4191 Acc: 0.7951

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2108 Acc: 0.9420
Epoch 8/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.3865 Acc: 0.8484

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2127 Acc: 0.9348
Epoch 9/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.4195 Acc: 0.8033

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2022 Acc: 0.9565
Epoch 10/10
----------

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

train Loss: 0.3888 Acc: 0.8156

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

val Loss: 0.2033 Acc: 0.9493
Training complete in 13m 34s
Best val Acc: 0.956522

On a entraîné ce nouveau modèle. on définit maintenant la fonction eval_model(model), qui va évaluer les performances du modèle sur le data test.

On ajoute un test d'évaluation du modèle avec de nouvelles données

def eval_model(model):
    # initialisation des variables
    running_loss = 0.0
    running_corrects = 0
    
    #on teste sur les données test
    for inputs, labels in dataloaders["test"]:
        inputs = inputs.to(device)
        labels = labels.to(device)

        outputs = model(inputs)
        _, preds = torch.max(outputs, 1)
        loss = criterion(outputs, labels)

        # Statistics
        running_loss += loss.item() * inputs.size(0)
        running_corrects += torch.sum(preds == labels.data)

    #Calcul du loss et de l'accuracy
    loss = running_loss / dataset_sizes["test"]
    acc = running_corrects.double() / dataset_sizes["test"]
        
    return loss,acc

loss,acc=eval_model(model)
print("Loss: {:.4f} Acc: {:.4f}".format(loss, acc))

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

Loss: 0.0623 Acc: 1.0000

On obtient une précision de 1 sur l'échantillon test. (il faudrait ajouter plus de photos)

---

On va maintenant quantizer le modèle et comparer la taille des modèles.

quantized_model = torch.quantization.quantize_dynamic(model, dtype=torch.qint8)
print_size_of_model(model, "int8")
print_size_of_model(quantized_model, "int8")

model:  int8  	 Size (KB): 45304.25
model:  int8  	 Size (KB): 44911.014

44911014

La quantization permet de passer de 45304.25 à 44911.014 KB.

On étudie les performances du modèle.

loss,acc=eval_model(quantized_model)
print("Loss: {:.4f} Acc: {:.4f}".format(loss, acc))

/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(
/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

Loss: 0.2032 Acc: 1.0000

Le modèle quantizé permet aussi d'atteindre une précision de 1.

Optional

Try this at home!!

Pytorch offers a framework to export a given CNN to your selfphone (either android or iOS). Have a look at the tutorial https://pytorch.org/mobile/home/

The Exercise consists in deploying the CNN of Exercise 4 in your phone and then test it on live.

Author

Alberto BOSIO - Ph. D.