Modelpublic

cstr/firered-vad-GGUF

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

1likes432downloads

Model Card

FireRedVAD -- GGUF

GGUF conversions of `FireRedTeam/FireRedVAD` for use with [CrispStrobe/CrispASR](https://github.com/CrispStrobe/CrispASR).

File	Variant	Size	Params	Notes
`firered-vad.gguf`	VAD	2.4 MB	588K	Non-streaming, lookback+lookahead
`firered-stream-vad.gguf`	Stream-VAD	2.3 MB	568K	Streaming (no lookahead)
`firered-aed-vad.gguf`	AED	2.4 MB	589K	Multi-label: speech/singing/music

All variants are F32 (no quantization needed — models are already tiny).

—Architecture: DFSMN (Deep Feedforward Sequential Memory Network) — 8 blocks with depthwise lookback/lookahead convolutions (k=20)
—Parameters: ~588K (2.4 MB)
—Languages: 100+ (language-agnostic voice activity detection)
—F1 Score: 97.57% on FLEURS-VAD-102 (outperforms Silero-VAD, TEN-VAD, FunASR-VAD, WebRTC-VAD)
—License: Apache 2.0

bash

python models/convert-firered-vad-to-gguf.py --input FireRedTeam/FireRedVAD --variant VAD --output firered-vad.gguf

—Upstream model: FireRedTeam/FireRedVAD — published by FireRedTeam.
—Upstream licence: apache-2.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
—What was done here: format conversion and/or quantisation only (GGUF/GGML). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
—Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository. No training-content summary was found on the upstream model card at the time of writing; that documentation gap is upstream's and is not filled here.
—Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.