YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
DSN 2026 spoken language ID - AfriHuBERT run 4
Audio classifier for 12 Nigerian languages (yor, ibo, hau, ego, anw, efi, ibb, idu, igl, nup, tiv, urh) plus other.
Base model: ajesujoba/AfriHuBERT (continued pretraining of utter-project/mHuBERT-147). See the base model cards for their licenses.
Training data: Data-Science-Nigeria/dsn-2026-lid-train only (train + validation merged). No external datasets were used.
Approach: HubertForSequenceClassification fine-tuned with 5 s random crops (tiling for short clips), class- and source-balanced sampling, channel augmentation (band-limiting, 8 kHz round trip, reverb, additive noise), SpecAugment, label smoothing 0.1, and weight averaging of the last 3 epoch checkpoints. Weights stored in fp16.
- Downloads last month
- -