Downloads · 30 days
0
w2v-bert-VoxLingua
areffarhadi/w2v-bert-VoxLingua
w2v-bert-VoxLingua is a machine learning model from areffarhadi. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for pytorch. The card lists the license as mit.
Language identification checkpoint trained on VoxLingua107, evaluated on devvox.
Access
Public
Updated Jun 19, 2026
Repo size
2.3 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt2.3 GB · 100%
At a glance
- Library
- pytorch
- License
- mit
- Access
- Public
- Created
- Jun 19, 2026
- Updated
- Jun 19, 2026
- SHA
- 300b1e6e
Model card
From the Hugging Face model README
VoxLingua LID — WPT + MHFA (Odyssey 2026)
Language identification checkpoint trained on VoxLingua107, evaluated on dev_vox.
Related paper: Subtract to Clean, Add to Enrich: Dual-Path Disentanglement for Speaker and Language Recognition (Odyssey 2026).
Author: Aref Farhadipour
Files
| File | Description |
|---|---|
model.pt | WPT + W2V-BERT-2.0 + MHFA classifier (107 languages) |
args.json | Training hyperparameters |
- Backbone: facebook/w2v-bert-2.0
- Prompt tokens: 15 regular + 8 wavelet per layer
- Clip length: 96000 samples (6 s @ 16 kHz)
Reported performance: see the paper.
Download
huggingface-cli download areffarhadi/w2v-bert-VoxLingua --local-dir ./checkpoints/voxlingua_lid
Code
GitHub: areffarhadi/ASV-LID
Eval
Dataset layout: dev_vox/<lang>/*.wav
bash scripts/eval_voxlingua_lid.sh 0
Citation
@inproceedings{farhadipour2026subtract,
title={Subtract to Clean, Add to Enrich: Dual-Path Disentanglement for Speaker and Language Recognition},
author={Farhadipour, Aref},
booktitle={Proc. Odyssey},
year={2026}
}
- Library
- pytorch
- License
- mit
- Created
- Jun 19, 2026
- Updated
- Jun 19, 2026