1. Implement the classification class, the LinearClassifier class and the MLPClassifier class in models.py

import torch import torch.nn.functional as F

class ClassificationLoss(torch.nn.Module): def forward(self, input, target): """ Your code here

Compute mean(-log(softmax(input)_label))

@input: torch.Tensor((B,C)) @target: torch.Tensor((B,), dtype=torch.int64)

@return: torch.Tensor((,))

Hint: Don't be too fancy, this is a one-liner """ raise NotImplementedError('ClassificationLoss.forward')

class LinearClassifier(torch.nn.Module): def __init__(self): super().__init__()

""" Your code here """ raise NotImplementedError('LinearClassifier.__init__')

def forward(self, x): """ Your code here

@x: torch.Tensor((B,3,64,64)) @return: torch.Tensor((B,6)) """ raise NotImplementedError('LinearClassifier.forward')

class MLPClassifier(torch.nn.Module): def __init__(self): super().__init__()

""" Your code here """ raise NotImplementedError('MLPClassifier.__init__')

def forward(self, x): """ Your code here

@x: torch.Tensor((B,3,64,64)) @return: torch.Tensor((B,6)) """ raise NotImplementedError('MLPClassifier.forward')

2. Implement the __init__, __len__, and __getitem__ methods of the SuperTuxDataset class in utils.py. The __len__ method should return the size of the dataset. The __getitem__ method should return a tuple of images and labels. The image should be a torch. Tensor of size (3, 64, 64) with a range of [0, 1], and the label should be an integer. Ensure that the labels correspond to the classes: background (0), kart (1), pickup (2), nitro (3), bomb (4), and projectile (5). Use the information from the labels.csv file, where the headers are the file and label.

class SuperTuxDataset(Dataset): def __init__(self, dataset_path): """ Your code here Hint: Use the python csv library to parse labels.csv

WARNING: Do not perform data normalization here. """ raise NotImplementedError('SuperTuxDataset.__init__')

def __len__(self): """ Your code here """ raise NotImplementedError('SuperTuxDataset.__len__')

def __getitem__(self, idx): """ Your code here return a tuple: img, label """ raise NotImplementedError('SuperTuxDataset.__getitem__')

3. Implement the MLPClassifier class in models.py. The inputs and outputs to the multi-layer perceptron are the same as the linear classifier. However, now you're learning a non-linear function.

from .models import ClassificationLoss, model_factory, save_model from .utils import accuracy, load_data

def train(args): model = model_factory[args.model]()

""" Your code here

""" raise NotImplementedError('train')

save_model(model)