Skip to content
This repository was archived by the owner on Jun 6, 2026. It is now read-only.

feat: add FunASR Nano LLM support#125

Open
oldcai wants to merge 3 commits into
thewh1teagle:mainfrom
oldcai:main
Open

feat: add FunASR Nano LLM support#125
oldcai wants to merge 3 commits into
thewh1teagle:mainfrom
oldcai:main

Conversation

@oldcai

@oldcai oldcai commented Jan 14, 2026

Copy link
Copy Markdown
  • Update sherpa-onnx submodule to latest (includes FunASR Nano C API)
  • Add funasr_nano.rs wrapper for FunASR Nano recognizer
  • Add missing fields (funasr_nano, omnilingual, medasr) to all model configs
  • FunASR Nano uses unified KV-cache LLM decoder for better accuracy

- Update sherpa-onnx submodule to latest (includes FunASR Nano C API)
- Add funasr_nano.rs wrapper for FunASR Nano recognizer
- Add missing fields (funasr_nano, omnilingual, medasr) to all model configs
- FunASR Nano uses unified KV-cache LLM decoder for better accuracy
- Update dist.json to v1.12.22 with new onnxruntime naming format
- Add checksum for v1.12.22 macOS binary
Xun Cai added 2 commits January 15, 2026 15:47
- Update dynamic macOS builds to use onnxruntime-1.16.2 (>= 1.15 required for CoreML)
- Update static macOS builds to use default archive without explicit onnxruntime version
- Add checksums for new archives

This enables CoreML execution provider on macOS, allowing GPU/ANE acceleration
for speech recognition models.
- Update dist.json to use v1.12.23 release
- macOS now uses onnxruntime 1.23.0 (was 1.14.1)
- Remove iOS target (not available in v1.12.23)
- Update checksums for all platforms
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant