'int' object is not subscriptable error on using BucketIterator

nlp
1

I am getting the following error on using BucketIterator

Screenshot_20200126_121842
Screenshot_20200126_121842941×296 34.5 KB

2

Which dataset are you using in this case?

3

Sorry sir for replying late .I was applying it on the aggregator label in WIkiSQL Dataset

4

Could you copy/paste your code so I can reproduce? You could also open an issue in pytorch/text.

In general, we would like to retire iterator in torchtext. Instead, the new abstraction uses DataLoader in torch.utils.data, like text_classification_dataset

1 Like
5

Thank you for your concern and sorry again for replying late. It was a problem in the parameters to the BucketIterator . The error has been solved. Thank you sir

6

Hi @JACOBIN-SCTCS, Could share here how you addressed the problem? Thanks!

7

Here the datasets contain only one dataset which is train_set. However in my case I had three datasets-train, valid and test. So they are given in a tuple. Thus the batch size should also be given in a tuple(even in the case of the same batch size for all datasets). This way, I solved the problem.

8

Actually, I know what was the problem. The ‘batch_size’ parameter in BucketIterator automatically changes to ‘batch_sizes’ which is wrong and there is no need to use ‘s’ at the end of the parameter.

1 Like