audio-cpp/audio.cpp-gguf

1---2library_name: audio.cpp3license: other4license_name: funasr-model-license-1.15license_link: https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512-hf/blob/main/README.md6base_model: FunAudioLLM/Fun-ASR-Nano-2512-hf7pipeline_tag: automatic-speech-recognition8tags:9- audio.cpp10- gguf11- speech-recognition12- multilingual13- funasr14---15 16# Fun-ASR-Nano-2512 GGUF17 18Standalone audio.cpp GGUF builds of19[FunAudioLLM/Fun-ASR-Nano-2512-hf](https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512-hf).20Each file embeds the model configuration, processor configuration, tokenizer,21chat template, and the audio.cpp model package specification.22 23## Files24 25| File | Size | SHA256 |26| --- | ---: | --- |27| `fun-asr-nano-2512-q8_0.gguf` | 1,045,334,432 bytes | `4d727357574b079b7f43336b2930f39da086ca02f5d8d50872090b4c1c3d5e0a` |28| `fun-asr-nano-2512-f16.gguf` | 1,675,708,832 bytes | `3d906c3ccfed07efef88ff53d6cc94b788b9d2edf1492a5679d041b43e98c5be` |29 30The source checkpoint is pinned to revision31`854d88f94205cd17d2afdb24332130d86fbe654a`. The source32`model.safetensors` SHA256 is33`335ca3e74917f1156690400e2c344350112950165789cf78ce3d0a367affd821`.34 35## audio.cpp36 37```bash38audiocpp_cli \39  --task asr \40  --family fun_asr_nano \41  --model fun-asr-nano-2512-q8_0.gguf \42  --backend cuda \43  --audio speech.wav44```45 46Fun-ASR-Nano currently provides offline multilingual ASR. It does not expose47streaming or timestamp output. On CUDA, audio.cpp keeps the Q8_0 encoder and48adaptor weights native and loads decoder weights as BF16 by default for stable49logits. An explicit `fun_asr_nano.decoder_weight_type` session option overrides50that default.51 52## Reproducibility53 54The files were generated with audio.cpp's `audiocpp_gguf` converter:55 56```bash57audiocpp_gguf \58  --input model.safetensors \59  --root /path/to/Fun-ASR-Nano-2512-hf \60  --output fun-asr-nano-2512-q8_0.gguf \61  --type q8_0 \62  --family fun_asr_nano \63  --model-spec model_specs/fun_asr_nano.json64```65 66Both formats were checked with `audiocpp_gguf --inspect` and full reference67audio transcription on CPU and NVIDIA H100 CUDA.68 69## License70 71The original model and these converted weights are governed by the72FunASR Model Open Source License Agreement v1.1 distributed with the source73model. Review that agreement before using or redistributing the files.74