mutisya/w2v-bert-adapters-14lang-e10-25_52-v10

W2V-BERT 2.0 ASR Adapters

This repository contains per-language bottleneck adapters for automatic speech recognition (ASR) trained on top of facebook/w2v-bert-2.0.

Model Description

• Base Model: facebook/w2v-bert-2.0 (600M parameters, frozen)
• Adapter Architecture: Bottleneck adapters (Pfeiffer-style, dim=64)
• Decoder: Lightweight transformer decoder (2 layers)
• Training: CTC loss with extended vocabulary for double vowels

Trained Adapters

Training in progress...

Adapter	Language	WER	Train Samples

Successful: 14/14

✅ Good (WER < 30%): 9
    swh_Latn_v2: 4.00%
    swh_Latn_salt: 13.25%
    kik_Latn: 16.49%
    luo_Latn: 16.50%
    swh_Latn_v1: 17.34%
    eng_Latn_tts: 21.85%
    eng_Latn_salt: 24.58%
    lug_Latn_salt: 28.02%
    ach_Latn: 28.62%

⚡ Medium (30-60% WER): 3
    kam_Latn: 30.66%
    mer_Latn: 36.49%
    teo_Latn: 58.12%

⚠️ Poor (60-90% WER): 1
    nyn_Latn: 64.39%

❌ Collapsed (WER >= 90%): 1
    ful_Latn: 99.98%

Architecture

The model uses:

1. Frozen w2v-bert-2.0 encoder - Extracts audio representations
2. Bottleneck adapters - Language-specific adaptation (trainable)
3. Lightweight decoder - Transformer decoder blocks (trainable)
4. LM head - Per-language vocabulary projection (trainable)

Usage

Each adapter folder contains:

• adapter_weights.pt - Bottleneck adapter weights
• decoder_weights.pt - Decoder block weights
• lm_head_weights.pt - Language model head weights
• final_norm_weights.pt - Final layer norm weights
• vocab.json - Language-specific vocabulary
• adapter_config.json - Adapter configuration
• metrics.json - Training metrics

Training Configuration

• Epochs: 10
• Base Learning Rate: 0.0005 (adaptive based on dataset size)
• Batch Size: 48 x 1
• Extended Vocabulary: True
• Adapter Dimension: 64

License

Apache 2.0