modelscope/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR.git synced 2026-10-12 02:19:18 +00:00
Code Issues Packages Projects Releases Wiki Activity
5,638 Commits 468 Branches 66 Tags
178a36718cb1110255296bdf4986ff4ebf295019
Commit Graph
453 Commits
Author SHA1 Message Date
LauraGPT 178a36718c fix(runtime): decode SenseVoice F16 query embeddings 
Signed-off-by: LauraGPT <[email protected]>
 2026-08-29 02:40:29 +00:00
zhifu gao 159f67b154 chore(release): prepare 1.4.6 (#3549) 
Publish the realtime keepalive, empty timestamp segment, and readable subtitle fixes with the verified llama.cpp v0.2.3 nine-platform runtime matrix.

Signed-off-by: LauraGPT <[email protected]>
 2026-08-29 09:23:49 +08:00
LauraGPT cabecbf133 test(runtime): track v0.2.3 release contract 
Signed-off-by: LauraGPT <[email protected]>
 2026-08-28 17:58:13 +00:00
LauraGPT 6ded40886c docs(runtime): publish v0.2.3 diagnostics 
Signed-off-by: LauraGPT <[email protected]>
 2026-08-28 17:55:23 +00:00
LauraGPT 7429dbd787 fix(runtime): trace SenseVoice backend stages 
Signed-off-by: LauraGPT <[email protected]>
 2026-08-28 17:14:09 +00:00
LauraGPT 0395c59eb7 fix(vulkan): guard backend initialization failures 
Signed-off-by: LauraGPT <[email protected]>
 2026-08-28 14:49:31 +00:00
zhifu gao 4df59cc153 chore(release): prepare 1.4.5 (#3534) 
Publish the optional-torchaudio inference path, pin the verified nine-platform llama.cpp runtime matrix, and document the protected exact-main publication gate.

Signed-off-by: LauraGPT <[email protected]>
 2026-08-28 00:59:31 +08:00
zhifu gaoandLauraGPT b454f9b7d6 docs: publish llama.cpp runtime v0.2.1 downloads (#3532) 
Signed-off-by: LauraGPT <[email protected]>
Co-authored-by: LauraGPT <[email protected]>
 2026-08-27 17:32:44 +08:00
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
Leonard Li 4999dc7632 Raise real exceptions instead of strings in the runtime Python bindings 
`raise "some message"` is invalid in Python 3 — a str is not an exception — so
the interpreter discards the message and raises

    TypeError: exceptions must derive from BaseException

instead. Paired with a bare `except:`, the original cause is lost too, and the
user is told to install something they may already have.

Concretely: funasr was installed, but importing it failed on a missing
torchaudio. The message said 'please install funasr', the traceback said
TypeError, and the actual ModuleNotFoundError was nowhere to be seen.

This converts all 30 occurrences under runtime/python to real exceptions:

  - guarded imports        -> except ImportError as e: raise ImportError(...) from e
  - everything else        -> except Exception as e:  raise RuntimeError(...) from e

`from e` keeps the original traceback, so a missing transitive dependency stays
visible. Narrowing the bare `except:` also stops unrelated failures from being
reported as a missing package — the model_dir branch in particular fired for any
failure, so a mistyped path also surfaced as TypeError.

Messages are unchanged. No behaviour changes beyond the exception type.
 2026-08-22 12:35:50 -04:00
zhifu gao 65bbd29f60 docs: clarify llama.cpp speaker diarization support 
Signed-off-by: zhifu gao <[email protected]>
 2026-08-13 02:15:55 +00:00
zhifu gao 62b784b2b9 docs: surface llama.cpp runtime v0.2.0 (#3484) 
Signed-off-by: LauraGPT <[email protected]>
 2026-08-11 14:17:59 +08:00
zhifu gao 500956bc33 fix(vulkan): refresh ggml for AMD submission handling (#3482) 
Refresh the pinned llama.cpp revision with AMD Vulkan submission batching and DeviceLost diagnostics, adapt the sampler API, and document Windows troubleshooting. Refs #3479.
 2026-08-11 12:19:11 +08:00
Dingyuan Wang b0d44bd792 llama.cpp: add srt output (#3480) 
Add optional SRT output to Fun-ASR-Nano, SenseVoiceSmall, and Paraformer llama.cpp runtimes. Preserve existing non-SRT output behavior and document VAD-aligned subtitle generation.
 2026-08-11 11:27:36 +08:00
zhifu gaoandLauraGPT 6408aaa96d feat(runtime): add SenseVoice TensorRT deployment (#3463) 
Co-authored-by: LauraGPT <[email protected]>
 2026-08-04 02:22:57 +08:00
zhifu gaoandLauraGPT 4dc8e5b4f0 feat(runtime): add JSONL timestamp output (#3462) 
Signed-off-by: LauraGPT <[email protected]>
Co-authored-by: LauraGPT <[email protected]>
 2026-08-04 01:00:20 +08:00
zhifu gao cb0eb3bb40 Merge pull request #3450 from wangxiuwen/fix/audio-frame-copy-ownership 
fix(runtime): make AudioFrame copies own their sample data
 2026-08-03 13:51:15 +08:00
zhifu gao e8b75de2e5 Merge pull request #3449 from wangxiuwen/fix/ffmpeg-avio-buffer-lifetime 
fix(runtime): free custom AVIO buffers on all exits
 2026-08-03 13:51:11 +08:00
wangxiuwen 62037b10fa fix(runtime): give AudioFrame deep-copy ownership 2026-07-30 10:24:51 +08:00
wangxiuwen c3c66061ee fix(runtime): free custom AVIO buffers on all exits 2026-07-30 10:22:58 +08:00
wangxiuwen b8c60ae1c2 fix(runtime): release two-pass frame on empty result 2026-07-30 10:20:58 +08:00
zhifu gaoandLauraGPT 30caaf8e18 fix(websocket): finalize file input on end control (#3440) 
Signed-off-by: LauraGPT <[email protected]>
Co-authored-by: LauraGPT <[email protected]>
 2026-07-29 20:24:01 +08:00
zhifu gaoandzhifu gao 19a903fa87 Use QwenAudio canonical repository links (#3412) 
Co-authored-by: zhifu gao <[email protected]>
 2026-07-24 19:01:36 +08:00
zhifu gao 6c3b47c2e5 docs: publish runtime v0.1.9 download links (#3399) 2026-07-23 19:38:51 +08:00
zhifu gao 9b918d74d7 feat(runtime): publish Windows Vulkan binary (#3398) 
* feat(runtime): add Windows Vulkan release asset

* fix(ci): install SPIR-V headers for Windows Vulkan

* fix(runtime): make Windows Vulkan package standalone
 2026-07-23 18:49:28 +08:00
zhifu gao 874330a1f7 Refresh public entrypoint links 
Move high-traffic docs and website source links to current FunASR and ModelScope entry points.
 2026-07-23 07:48:42 +08:00
zhifu gaoandLauraGPT e8bdb884dc Refresh public FunASR entrypoint links (#3361) 
Co-authored-by: LauraGPT <[email protected]>
 2026-07-23 05:42:04 +08:00
zhifu gaoandLauraGPT 2aa24baa4e Refresh public docs release copy (#3337) 
Co-authored-by: LauraGPT <[email protected]>
 2026-07-23 02:55:56 +08:00
zhifu gao c812c139a1 docs: refresh remaining stale FunASR links 
Refresh remaining stale public docs and clone links, with regression coverage.
 2026-07-22 22:28:45 +08:00
zhifu gao 53e397d2fa docs: refresh stale public FunASR links 
Replace stale public docs and clone hosts with current FunASR URLs, with regression coverage.
 2026-07-22 22:23:19 +08:00
zhifu gao 82a4d67b12 docs: fix broken markdown links 
Fix stale repository-local Markdown links and add a pytest guard for future docs link regressions.
 2026-07-22 22:02:41 +08:00
zhifu gao 5882f3cf79 docs: surface Vulkan llama.cpp runtime in README 
Expose the v0.1.8 Linux Vulkan runtime asset in the top-level README docs.
 2026-07-22 19:25:17 +08:00
zhifu gao c7f6edb9e6 docs(runtime): mention Vulkan asset in release notes 
Keep generated runtime release notes aligned with the new Linux Vulkan asset.
 2026-07-22 18:44:03 +08:00
zhifu gao e3327be1bd feat(runtime): add Vulkan llama.cpp backend packaging 
Add Linux Vulkan packaging and runtime selection for the FunASR llama.cpp SenseVoice GGUF backend.
 2026-07-22 18:40:34 +08:00
zhifu gao b798e2679f feat(runtime): add GGUF HTTP transcription wrapper 
Expose the llama.cpp/GGUF command-line binaries through a lightweight OpenAI-compatible transcription endpoint.
 2026-07-22 18:16:22 +08:00
dio 724a085855 fix(runtime): close online 2pass sentence state 
Keep online mode from returning 2pass-offline text while still closing final sentence state so the next utterance starts cleanly.
 2026-07-22 18:07:37 +08:00
zhifu gao a2a6ba3ddf docs: clarify Windows CUDA GPU architecture support 
Document the current sm_86 Windows CUDA prebuilt limit and direct RTX 50 / Blackwell users to CPU binaries or source builds with CMAKE_CUDA_ARCHITECTURES=120 until a dedicated asset is available.
 2026-07-19 18:39:13 +08:00
zhifu gao b006767999 Promote llama.cpp Windows CUDA release in README (#3250) 
* Promote llama.cpp Windows CUDA release

* Preserve releases link in llama.cpp README snippets
 2026-07-19 04:15:44 +08:00
zhifu gao f1e287edc2 Add Windows CUDA llama.cpp release asset 
Adds a Windows x64 CUDA release asset for the llama.cpp SenseVoice runtime, with CUDA Toolkit setup, architecture-86 build flags, focused SenseVoice target packaging, docs, and workflow regression tests. Manual workflow dispatch 29656771776 passed all build matrix jobs including build-windows-x64-cuda.
 2026-07-19 03:24:07 +08:00
zhifu gao 57721fc8ce Add SenseVoice llama.cpp backend selector (#3247) 2026-07-18 19:16:03 +08:00
zhifu gao 2792c98470 Release funasr-onnx 0.4.2 without Torch runtime dependency (#3242) 
* Release funasr-onnx 0.4.2 candidate

* Use Node 24 release validation actions

* Harden release dependency parsing
 2026-07-17 17:24:02 +08:00
zhifu gao 68a21aafbb docs(llama.cpp): fix default GGUF quickstart (#3234) 
* docs(llama.cpp): fix default GGUF quickstart

Signed-off-by: 游雁 <[email protected]>

* fix(llama.cpp): make release quickstart executable

Signed-off-by: 游雁 <[email protected]>

* fix(llama.cpp): validate Windows release layout

Signed-off-by: 游雁 <[email protected]>

* test(llama.cpp): require runnable quickstart lines

Signed-off-by: 游雁 <[email protected]>

---------

Signed-off-by: 游雁 <[email protected]>
 2026-07-15 20:15:07 +08:00
游雁 916c175ec9 test(llama.cpp): validate downloader failure paths 2026-07-15 04:35:18 +00:00
游雁 9c17931645 fix(llama.cpp): download one GGUF variant by default 2026-07-15 04:28:57 +00:00
zhifu gao ddecf48db4 Fix Java offline client connection cleanup (#3229) 
* fix(runtime): close Java client after final offline result

* test(runtime): exercise EOF through Java client API
 2026-07-15 06:07:36 +08:00
Maigic 098654e8db fix: initialize ggml timer in funasr nano cli (#3210) 
Initialize the ggml timer before the first ggml_time_us() call so Windows builds do not divide by an uninitialized timer frequency.
 2026-07-13 23:55:31 +08:00
zhifu gao de761e2be2 docs: update remaining FunASR repository links 
Update remaining CRLF Markdown and HTML docs from alibaba-damo-academy/FunASR to modelscope/FunASR while preserving line endings.
 2026-07-07 18:15:52 +08:00
zhifu gao 92aca6b259 docs: update FunASR repository links 
Update user-facing documentation links from alibaba-damo-academy/FunASR to modelscope/FunASR.
 2026-07-07 18:11:41 +08:00
zhifu gao 9f1d220432 feat: q8_0 export option (--wtype q8_0) for SenseVoice/Paraformer/Nano encoder GGUF (#3013) 2026-06-22 02:39:32 +08:00
zhifu gao 9cc0ebcc66 fix: detok review findings (null gguf_get_arr_str guard + utf-8 tokens.json read) (#3005) 2026-06-20 17:39:32 +08:00