Kari Ali Noriy
Kari Ali Noriy
@gmryu thank you for the reply. I've tried this and still get the same error. ```python sample['net_input']['src_tokens'] = sample['net_input']['src_tokens'].cuda() sample['net_input']['src_lengths'] = sample['net_input']['src_lengths'].cuda() sample['speaker'] = sample['speaker'].cuda() ``` Output: ``` >>> {'net_input':...
Hi @mthrok, I was wondering if there has been any update on this issue. ## Update After some looking around, the issue is caused by the `sox_io` backend. Setting torch...
Further looking into how this is implemented, the model does not need to be sent to the device. The issue arises when the model is sent to device: ```python class...
Ahh, sorry, I missed that detail in the docs. I think It would be best to require a `dict` in the case where Numpy arrays are passed in and have...
> I think we can add a warning when a bare NumPy array is passed informing the user that no resampling operations are performed and the code may be prone...
I've opened a pull request which fixes this issue