My custom dataset has the following in the init()
...
self.train_loader = torch.utils.data.DataLoader(
self.train_set,
batch_size=batch_size, shuffle=shuffle, batch_sampler=batch_sampler,
)
now I assume, that when in my modeling process I gonna use this DataLoader, I should be able to give batch_size to the arguments
train_loader = dataset.train_loader(batch_size=self.batch_size)
where dataset is my custom Dataset. But I get
train_loader = dataset.train_loader(batch_size=self.batch_size)
TypeError: 'DataLoader' object is not callable
Why and how to fix it?
The TypeError occurs because you are trying to call the train_loader object like a function. Instead, you should create a method within your custom dataset class to return a new DataLoader with the desired batch size. Here’s how you can do it:
- In your custom dataset class, remove the initialization of
self.train_loader from the __init__ method.
- Add a new method to the custom dataset class that returns a DataLoader with the given batch size:
class MyCustomDataset(torch.utils.data.Dataset):
def __init__(self, ...):
# Your dataset initialization code here
def get_train_loader(self, batch_size, shuffle=True, batch_sampler=None):
train_loader = torch.utils.data.DataLoader(
self.train_set,
batch_size=batch_size, shuffle=shuffle, batch_sampler=batch_sampler,
)
return train_loader
- Now, in your modeling process, you can get a train_loader with the desired batch size like this:
train_loader = dataset.get_train_loader(batch_size=self.batch_size)
train_loader = dataset.get_train_loader(batch_size=self.batch_size)
By using this approach, you can create a DataLoader with the specified batch size when needed without getting the TypeError