Fritter Nemotron 0.6B Sherpa Onnx Int8
No description provided
About This Project
Fritter-Nemotron-0.6B-sherpa-onnx-int8 is a specialized streaming ASR model based on the NVIDIA Nemotron-speech-streaming-en-0.6b architecture. It is fine-tuned to excel at technical dictation, making it particularly effective for transcribing industry-specific terminology and AI-related jargon. The model is exported for use with the sherpa-onnx framework, utilizing an INT8-quantized transducer format for efficient, low-latency deployment in real-time streaming environments.
The model features a cache-aware encoder with fixed 560 ms chunks and provides support for hotword integration via BPE vocabulary files. It serves as a portable solution for developers requiring high-accuracy, lightweight speech recognition for technical applications. The project is licensed under CC-BY-NC-4.0 for non-commercial use and is designed to provide consistent, performant speech-to-text capabilities in resource-constrained settings.