Hebrish/Shakshuka-GGUF

Shakshuka v2.0: GGUF

GGUF build of Shakshuka v2.0 (Hebrish Release v2, run r9-nemotron-all, 2026-10-22) for transcribe.cpp (converted with v0.1.3, variant profile nemotron-speech-streaming-en-0.6b): Shakshuka-F16.gguf, for desktop GPU inference (HIP/ROCm, Vulkan or CUDA). From v2.0 only the F16 build is made; the v1.0 Q8_0, Q4_K_M and F16 files are in this repo's history. On CPUs use the sherpa-onnx streaming build. Read the main card for training data and limitations.

Whether a given app (Offline Voice Input, Handy) loads this architecture depends on its transcribe.cpp version; transcribe-cli from v0.1.3 does.

Not benchmarked: Release v2 is published without a test-set evaluation. The export was checked by transcribing five held-out clips (load-and-transcribe, not an evaluation): offline GGUF and streaming sherpa agree except punctuation; both dropped the last word or two of one clip that ends right after its final word (pad audio with ~0.5 s of trailing silence).

Licence

Licensed by NVIDIA Corporation under the NVIDIA Open Model License (LICENSE, NOTICE); use is also governed by the Hebrish terms of use. Not made or endorsed by NVIDIA.

Base model and licence notices

This model is a build of Hebrish/Shakshuka, a fine-tune of nvidia/nemotron-speech-streaming-en-0.6b, by NVIDIA. Licensed by NVIDIA Corporation under the NVIDIA Open Model License. Hebrish modified the base model by fine-tuning it and converted the weights for this build.

A copy of the agreement (LICENSE) and the NOTICE file are in this repository. Use must also follow the NVIDIA Trustworthy AI terms.

The licence stated at the top of this card covers Hebrish's contribution; the base model's licence keeps applying to the base model, and nothing here restricts what that licence allows for it. Terms of use: https://hebrish.app/terms.

Hebrish is not affiliated with or endorsed by NVIDIA.