Downloads · 30 days
0
localingo-mms-lid-coreml
brbousnguar/localingo-mms-lid-coreml
localingo-mms-lid-coreml is a audio classification model from brbousnguar. Use it for the audio classification task on the model card, and read the license before you ship it in a product. The card lists the license as cc-by-nc-4.0.
A Core ML conversion of Meta's facebook/mms-lid-4017 spoken-language identifier, used on-device by the Localingo iOS app.
Access
Public
Updated Sep 24, 2026
Repo size
971 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin971 MB · 100%
At a glance
- Task
- Audio Classification
- License
- cc-by-nc-4.0
- Access
- Public
- Created
- Sep 24, 2026
- Updated
- Sep 24, 2026
- SHA
- 6e789ac8
Base models
Model card
From the Hugging Face model README
MMS-LID for Core ML (262 languages, 8-bit)
A Core ML conversion of Meta's facebook/mms-lid-4017 spoken-language identifier, used on-device by the Localingo iOS app.
- Languages: the 4017-class head is sliced to 262 rows: the 256 languages of
mms-lid-256, plus Inuktitut, Inuinnaqtun, Greenlandic, Iñupiaq, Tamasheq and Tawallammat.labels.jsonlists them in output order. - Input:
audio, float32[1, samples], 16 kHz mono, 1–20 s. Normalisation is built in. - Output:
probs, softmax over the 262 languages. - Weights: 8-bit linear quantisation. On test clips it gives the same top-1 as PyTorch, with probabilities within 0.03.
- Compute: float32 activations (float16 overflows on clips past ~4 s). Run it on the CPU: it loads in seconds, while the GPU rebuilds its pipeline (~40 s) on every load.
License
The weights are Meta's, released under CC BY-NC 4.0, so this conversion is too: attribution required, non-commercial use only. See the MMS paper for how the model was trained.
- Task
- Audio Classification
- License
- cc-by-nc-4.0
- Created
- Sep 24, 2026
- Updated
- Sep 24, 2026