ModelAufgelistetExtern

Sherpa Onnx Nemotron 3.5 Asr Streaming 0.6b 1120ms 2026 06 11

sherpa-onnx streaming transducer export of nvidia/nemotron-3.5-asr-streaming-0.6b at a 1120 ms chunk size, exported with the official k2-fsa/sherpa-onnx exportonnx.py.

Quellehuggingface.co/Masterx/sherpa-onnx-nemotron-3.5-asr-streaming-0.6b-1120ms-2026-06-11/tree/bee5a7824bb42c9ed1232d626333edc0c0030c7f

ArtOtherBasismodellnvidia/nemotron-3.5-asr-streaming-0.6bVersionRevision bee5a78CBewertung

Modellquelle

Art: Other
Dateigröße: 9,06 MB–1,16 GB
Basismodell: nvidia/nemotron-3.5-asr-streaming-0.6b
Version: vRevision bee5a78
Aktualisiert: 26. Sept. 2026
Downloads: 0
Quelle: Hugging Face

Quellenbeschreibung

sherpa-onnx Nemotron-3.5 streaming ASR (1120ms) — fp32 + int8

sherpa-onnx streaming transducer export of nvidia/nemotron-3.5-asr-streaming-0.6b at a 1120 ms chunk size, exported with the official k2-fsa/sherpa-onnx export_onnx.py.

Unlike the int8-only upstream package, this repo also ships the fp32 graphs (encoder.onnx + encoder.data, decoder.onnx, joiner.onnx) for best accuracy, alongside the *.int8.onnx variants.

The encoder is a cache-aware streaming FastConformer with a 6th prompt_index input; the encoder metadata carries prompt_dictionary (100+ languages → prompt id) and auto_prompt_id (101 = whole-utterance auto-detect). Used by WinSTT.

Multilingual (100+ languages). License CC-BY-4.0 (inherited from the base model).

Quellen

1 Quelle

Hugging FacePrimärVersion bee5a7824bb42c9ed1232d626333edc0c0030c7f

huggingface.co/Masterx/sherpa-onnx-nemotron-3.5-asr-streaming-0.6b-1120ms-2026-06-11/tree/bee5a7824bb42c9ed1232d626333edc0c0030c7f

Verifiziert 6. Sept.

Download files

Modellartefakte

9 Artefakte

decoder.fp16.onnx

onnx · 28,5 MB · SHA-256 56009566ce40…fd28 · Hugging Face

Download file

decoder.int8.onnx

onnx · 14,3 MB · SHA-256 19f9c98fc6d0…2fb0 · Hugging Face

Download file

decoder.onnx

onnx · 57,0 MB · SHA-256 3a630cfd17e0…34e7 · Hugging Face

Download file

encoder.fp16.onnx

onnx · 1,16 GB · SHA-256 e06efc7b0459…1997 · Hugging Face

Download file

encoder.int8.onnx

onnx · 627 MB · SHA-256 4260ce70aab6…068b · Hugging Face

Download file

encoder.onnx

onnx · 40,1 MB · SHA-256 51c381a0debd…0646 · Hugging Face

Download file

joiner.fp16.onnx

onnx · 18,0 MB · SHA-256 807c3732ac55…e5ee · Hugging Face

Download file

joiner.int8.onnx

onnx · 9,06 MB · SHA-256 4101c7c679a0…d7ce · Hugging Face

Download file

joiner.onnx

onnx · 36,1 MB · SHA-256 fa72e75079fa…3799 · Hugging Face

Download file

Quellenauszüge

2 Auszüge

Readme6. Sept.huggingface.co/Masterx/sherpa-onnx-nemotron-3.5-asr-streaming-0.6b-1120ms-2026-06-11

--- license: cc-by-4.0 language: - multilingual library_name: sherpa-onnx tags: - automatic-speech-recognition - streaming - nemo - nemotron base_model: - nvidia/nemotron-3.5-asr-streaming-0.6b --- # sherpa-onnx Nemotron-3.5 streaming ASR (1120ms) — fp32 + int8 sherpa-onnx streaming transducer export of [`nvidia/nemotron-3.5-asr-streaming-0.6b`](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b) at a 1120 ms chunk size, exported with the official [`k2-fsa/sherpa-onnx`](https://github.com/k2-fsa/sherpa-onnx) `export_onnx.py`. Unlike the int8-only upstream package, this repo also ships the **fp32** graphs (`encoder.onnx` + `encoder.data`, `decoder.onnx`, `joiner.onnx`) for best accuracy, alongside the `*.int8.onnx` variants. The encoder is a cache-aware streaming FastConformer with a 6th `prompt_index` input; the encoder metadata carries `prompt_dictionary` (100+ languages → prompt id) and `auto_prompt_id` (101 = whole-utterance auto-detect). Used by [WinSTT](https://github.com/dahshury/WinSTT). Multilingual (100+ languages). License CC-BY-4.0 (inherited from the base model).

Note6. Sept.huggingface.co/Masterx/sherpa-onnx-nemotron-3.5-asr-streaming-0.6b-1120ms-2026-06-11

Source context: 0 downloads · 0 likes · Pipeline automatic-speech-recognition · Library sherpa-onnx · Repo Masterx/sherpa-onnx-nemotron-3.5-asr-streaming-0.6b-1120ms-2026-06-11