├── requirements.txt
├── data
    ├── id-1.pt
    ├── id-2.pt
    ├── id-3.pt
    └── id-4.pt
├── my_classes.py
└── pytorch_script.py


/requirements.txt:
--------------------------------------------------------------------------------
1 | torch
2 | 


--------------------------------------------------------------------------------
/data/id-1.pt:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/shervinea/pytorch-data-generator/HEAD/data/id-1.pt


--------------------------------------------------------------------------------
/data/id-2.pt:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/shervinea/pytorch-data-generator/HEAD/data/id-2.pt


--------------------------------------------------------------------------------
/data/id-3.pt:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/shervinea/pytorch-data-generator/HEAD/data/id-3.pt


--------------------------------------------------------------------------------
/data/id-4.pt:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/shervinea/pytorch-data-generator/HEAD/data/id-4.pt


--------------------------------------------------------------------------------
/my_classes.py:
--------------------------------------------------------------------------------
 1 | import torch
 2 | from torch.utils import data
 3 | 
 4 | class Dataset(data.Dataset):
 5 |    'Characterizes a dataset for PyTorch'
 6 |    def __init__(self, list_IDs, labels):
 7 |        'Initialization'
 8 |        self.labels = labels
 9 |        self.list_IDs = list_IDs
10 | 
11 |    def __len__(self):
12 |        'Denotes the total number of samples'
13 |        return len(self.list_IDs)
14 | 
15 |    def __getitem__(self, index):
16 |        'Generates one sample of data'
17 |        # Select sample
18 |        ID = self.list_IDs[index]
19 | 
20 |        # Load data and get label
21 |        X = torch.load('data/' + ID + '.pt')
22 |        y = self.labels[ID]
23 | 
24 |        return X, y
25 | 


--------------------------------------------------------------------------------
/pytorch_script.py:
--------------------------------------------------------------------------------
 1 | import torch
 2 | from torch.utils import data
 3 | 
 4 | from my_classes import Dataset
 5 | 
 6 | 
 7 | # CUDA for PyTorch
 8 | use_cuda = torch.cuda.is_available()
 9 | device = torch.device("cuda:0" if use_cuda else "cpu")
10 | cudnn.benchmark = True
11 | 
12 | # Parameters
13 | params = {'batch_size': 1,
14 |           'shuffle': True,
15 |           'num_workers': 6}
16 | max_epochs = 100
17 | 
18 | # Datasets
19 | partition = {'train': ['id-1', 'id-2', 'id-3'], 'validation': ['id-4']} # IDs
20 | labels = {'id-1': 0, 'id-2': 1, 'id-3': 2, 'id-4': 1} # Labels
21 | 
22 | # Generators
23 | training_set = Dataset(partition['train'], labels)
24 | training_generator = data.DataLoader(training_set, **params)
25 | 
26 | validation_set = Dataset(partition['validation'], labels)
27 | validation_generator = data.DataLoader(validation_set, **params)
28 | 
29 | # Loop over epochs
30 | for epoch in range(max_epochs):
31 |     # Training
32 |     for local_batch, local_labels in training_generator:
33 |         # Transfer to GPU
34 |         local_batch, local_labels = local_batch.to(device), local_labels.to(device)
35 | 
36 |         # Model computations
37 |         [...]
38 | 
39 |     # Validation
40 |     with torch.set_grad_enabled(False):
41 |         for local_batch, local_labels in validation_generator:
42 |             # Transfer to GPU
43 |             local_batch, local_labels = local_batch.to(device), local_labels.to(device)
44 | 
45 |             # Model computations
46 |             [...]
47 | 


--------------------------------------------------------------------------------