Closed zml24 closed 2 years ago
Yes, we will re-write the translation datasets later so they will be compatible with torch.utils.data.dataloader
(see discussions https://github.com/pytorch/text/issues/664)
How to solve this problem??
Closing this since torchtext.data.Iterator
has been deleted and is no longer available in the latest releases of our library.
❓ Questions and Help
Description Hey Dude, I am running a machine translation task with
torchtext.datasets.IWSLT
. However, I found that the CPU usage is always high and GPU usage is always low. When usingtorch.utils.data.dataloader
, I can usenum_workers
andpin_memory
to deal with this problem. But I can't deal with this problem withtorchtext.data.Iterator
ortorchtext.data.BucketItertor
.How can I accelerate my data loading speed via
torchtext
?