Iterable Dataset Crashing Runtime

vision
1

I have implemented a very simple Iterable Dataset in PyTorch :

class Maps(data.IterableDataset):
    def __init__(self):
        super().__init__()
        
    def yield_items(self):
        
        for i in range(150):
          file = np.random.rand(160, 256, 512)
          for j in range(150):
              img = file[j:j+10, :, :256]
              label = file[j:j+10, :, 256:]
              img = torch.tensor(img/np.max(img))
              label = torch.tensor(label / np.max(label))
              yield img.float(), label.float()
            
    def cycle_data(self):
        return cycle(self.seeItems())

    def __iter__(self):
        return self.cycle_place()

train_dataset = Maps()
data_loader = data.DataLoader(train_dataset)

for x,y in data_loader:
  continue

I am running this code on google colab and it causes the RAM to fill up real quick and the runtime crashes. Does someone know what this would be due to?
Edit : If I remove the loop on j it works fine. But in my actual use case that loop is inevitable.