memory_leak_pytorch.py

import torch
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
import time
from sklearn import datasets
import numpy
from PIL import Image

class ImageDataSet(Dataset):
    """Image dataset."""

    def __init__(self, X=None, y=None):
        self.number = range(100)
        self.loader = transforms.Compose([transforms.ToTensor()])

    def __len__(self):
        return 3000

    def __getitem__(self, idx):
        X = {}
        # for i in range(10):
        #     # X["number_{}".format(i)] = self.loader(Image.open('example.png')).float()
        #     X["number_{}".format(i)] = numpy.asarray(Image.open('example.png'))
        X  = []
        for i in range(10):
            X.append(numpy.asarray(Image.open('example.png')) )

        # image_path_set = self.X[idx]
        # label = self._label_from_path(self.X[idx][0])

        # images = {}
        # for image_path in image_path_set:
        #     kind = self._kind_from_path(image_path)
        #     image = self._load_image(image_path)
        #     images[kind] = imag
        label = numpy.expand_dims(0, axis=0)
        label = torch.from_numpy(numpy.array(label))

        return X, label

def main():
    dataset  = ImageDataSet()
    # dataloader = DataLoader(dataset, batch_size=32, shuffle=True, num_workers=4)
    dataloader = DataLoader(dataset, batch_size=32, shuffle=False, num_workers=4)
    for epoch in range(100):
        for X, y in dataloader:
            pass
            # time.sleep(0.01)
        print("epoch: {}".format(epoch))


if __name__ == '__main__':
    main()