12,000 hour subset
#1
by
mitchelldehaven
- opened
Distil-whisper mentions it is trained on 12k hours of pseudo-labeled people's speech, however the clean
subset does not contain the full 12k hours. Where is the rest of the pseudo-labeled audio?