cstr/firered-lid-GGUF

FireRedLID -- GGUF

GGUF conversion of FireRedTeam/FireRedLID for use with CrispStrobe/CrispASR.

Available variants

File	Type	Size	Notes
firered-lid.gguf	F16	1.7 GB	Encoder (F16) + LID decoder

Model details

• Architecture: Conformer encoder (16L, d=1280, 20 heads) + 6L Transformer decoder (8 heads, cross-attention)
• Parameters: ~887M (723M shared encoder + 164M LID decoder)
• Languages: 100+ languages + 20+ Chinese dialects
• Accuracy: 97.18% on FLEURS-82, 92.07% on CommonVoice-74
• License: Apache 2.0
• Output: Language code (e.g., "en", "zh", "mandarin", "ja")

Usage with CrispASR

# Language identification
./build/bin/crispasr --backend firered-asr -m firered-lid.gguf -f audio.wav

The encoder is shared with FireRedASR2-AED. The LID decoder uses 8 attention heads (vs 20 for ASR).

Conversion

python models/convert-firered-lid-to-gguf.py --input FireRedTeam/FireRedLID --output firered-lid.gguf