search

flashlight / wav2letter

Facebook AI Research's Automatic Speech Recognition Toolkit
https://github.com/facebookresearch/wav2letter/wiki
Other
6.35k stars 1.02k forks source link

AMI data preparation "AssertionError: Unknown speaker!" #976

Open alkazap opened 3 years ago

alkazap commented 3 years ago

python prepare.py fails with AssertionError: Unknown speaker! FEE005

Assertion error comes from utils.py line 117 assert speaker in speakers, f"Unknown speaker! {speaker}"

Judging by the assertion and line 119 speaker = speakers[speaker], the script expects speakers to be a dictionary, however, as we can see on line 111 speakers = {(speaker.id, speaker) for speaker in speakers}, speakers is a set of tuples.

Line 111 should be speakers = { speaker.id : speaker for speaker in speakers }, then the script works as intended.

tlikhomanenko commented 3 years ago

Hi @alkazap,

Thanks for finding the bug, could you mind to send PR on the fix?