mirror of https://github.com/modelscope/FunASR.git synced 2026-10-12 02:19:18 +00:00

Files

T

zhifu gao c69e69a0b7 fix(runtime): allow integrated GPU backends (#3531 )

Prefer matching discrete GPU devices while retaining a matching integrated GPU as fallback. Tested with the pinned llama.cpp revision in CPU and Vulkan builds; this addresses the iGPU selector bug without claiming to fix the separate RX 9070 XT backend crash.

2026-08-27 00:55:09 +08:00

golden

test: numerical regression harness (frozen golden vs ggml/VAD/CIF/CTC output) (#3003 )

2026-06-20 16:12:24 +08:00

README.md

test: numerical regression harness (frozen golden vs ggml/VAD/CIF/CTC output) (#3003 )

2026-06-20 16:12:24 +08:00

run_regression.sh

fix(llama.cpp): download one GGUF variant by default

2026-07-15 04:28:57 +00:00

sample.wav

test: numerical regression harness (frozen golden vs ggml/VAD/CIF/CTC output) (#3003 )

2026-06-20 16:12:24 +08:00

test_backend_flags.py

fix(runtime): allow integrated GPU backends (#3531 )

2026-08-27 00:55:09 +08:00

test_download_funasr_model.sh

docs: surface llama.cpp runtime v0.2.0 (#3484 )

2026-08-11 14:17:59 +08:00

test_gguf_server.py

feat(runtime): add GGUF HTTP transcription wrapper

2026-07-22 18:16:22 +08:00

test_release_workflow.py

docs: publish runtime v0.1.9 download links (#3399 )

2026-07-23 19:38:51 +08:00

README.md

Regression tests

Run each runtime tool on a fixed clip (sample.wav, ~6 s) and diff the output against the frozen golden in golden/. Catches regressions in the ggml graphs, the FSMN-VAD state machine, the CIF predictor and CTC decode.

Run

# from runtime/llama.cpp/, after building (cmake --build build):
./tests/run_regression.sh              # VAD (tiny model auto-fetched) + any tool whose GGUF is already local
RUN_FULL=1 ./tests/run_regression.sh   # also download the ASR GGUFs from Hugging Face and test every tool

BIN_DIR / MODELS_DIR override where binaries and GGUFs are found. Exit code is non-zero if any test fails; tools with no binary or no model are skipped.

Golden

Captured on Linux x86-64 (the reference platform) with the f16 GGUFs published at FunAudioLLM/*-GGUF. Update a golden file only with a deliberate, reviewed output change.