Reputation: 75
I created a custom Dataset class that inherits from PyTorch's Dataset class, in order to handle my custom dataset which i already preprocessed.
When i try to create a DataLoader object, i get this error:
/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py in __init__(self, dataset, batch_size, shuffle, sampler, batch_sampler, num_workers, collate_fn, pin_memory, drop_last, timeout, worker_init_fn)
174 if sampler is None:
175 if shuffle:
--> 176 sampler = RandomSampler(dataset)
177 else:
178 sampler = SequentialSampler(dataset)
/usr/local/lib/python3.6/dist-packages/torch/utils/data/sampler.py in __init__(self, data_source, replacement, num_samples)
62 "since a random permute will be performed.")
63
---> 64 if not isinstance(self.num_samples, int) or self.num_samples <= 0:
65 raise ValueError("num_samples should be a positive integer "
66 "value, but got num_samples={}".format(self.num_samples))
/usr/local/lib/python3.6/dist-packages/torch/utils/data/sampler.py in num_samples(self)
70 # dataset size might change at runtime
71 if self._num_samples is None:
---> 72 return len(self.data_source)
73 return self._num_samples
74
/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataset.py in __len__(self)
18
19 def __len__(self):
---> 20 raise NotImplementedError
21
22 def __add__(self, other):
NotImplementedError:
So, the error message regards to the not implementation of the len() function in the dataset.py, right? But i did implement it and the getitem(), init() as well .
How can i overcome this? Thank you
Upvotes: 2
Views: 6221
Reputation: 2200
Make sure the name is correct in your code. It should be __len__
.
Upvotes: 6