From 7fc9de9d2fdfdde6e5c58bb64b5d49f1450be7d1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E8=A5=BF=E9=A3=8E=E9=80=8D=E9=81=A5=E6=B8=B8?= Date: Wed, 11 Nov 2026 23:43:40 +0800 Subject: [PATCH 1/5] fix(voice): run CosyVoice2 on ROCm with shared root dependencies --- dependencies/dependencies.yaml | 14 + dependencies/linux-amdgpu-dependencies.yaml | 129 +- dependencies/linux-amdgpu.lock | 106 +- dependencies/linux-arm64-dependencies.yaml | 1317 +++++++++++++++++ dependencies/linux-arm64.lock | 164 +- dependencies/linux-dependencies.yaml | 126 +- dependencies/linux.lock | 103 +- dependencies/macosx-dependencies.yaml | 115 +- dependencies/macosx.lock | 90 +- dependencies/requirements-linux-amdgpu.txt | 12 + dependencies/requirements-linux-arm64.txt | 12 + dependencies/requirements-linux.txt | 12 + dependencies/requirements-macosx-x86_64.txt | 8 + dependencies/requirements-macosx.txt | 9 + dependencies/requirements-windows-amdgpu.txt | 4 + .../requirements-windows-arm64-cpu.txt | 4 + dependencies/requirements-windows.txt | 4 + dependencies/windows-amdgpu-dependencies.yaml | 28 + dependencies/windows-amdgpu.lock | 25 +- .../windows-arm64-cpu-dependencies.yaml | 24 + dependencies/windows-arm64-cpu.lock | 23 +- dependencies/windows-dependencies.yaml | 30 +- dependencies/windows.lock | 27 +- extensions/Voice/README.md | 85 ++ extensions/Voice/ssui_voice/Cosyvoice.py | 68 +- .../Voice/vendor/cosyvoice/cli/frontend.py | 21 +- extensions/Voice/vendor/cosyvoice/llm/llm.py | 9 + .../vendor/cosyvoice/utils/file_utils.py | 8 +- .../matcha/models/components/flow_matching.py | 4 +- .../Voice/vendor/matcha/utils/__init__.py | 28 +- package.json | 1 + scripts/probe_cosyvoice.py | 100 ++ tests/cosyvoice_adapter_test.py | 68 + tests/cosyvoice_cache_test.py | 36 + tests/model/cosyvoice_model_test.py | 55 +- tests/model_manifest.yaml | 6 +- 36 files changed, 2702 insertions(+), 173 deletions(-) create mode 100644 dependencies/linux-arm64-dependencies.yaml create mode 100644 extensions/Voice/README.md create mode 100644 scripts/probe_cosyvoice.py create mode 100644 tests/cosyvoice_adapter_test.py create mode 100644 tests/cosyvoice_cache_test.py diff --git a/dependencies/dependencies.yaml b/dependencies/dependencies.yaml index 04f92ca0..de05f42b 100644 --- a/dependencies/dependencies.yaml +++ b/dependencies/dependencies.yaml @@ -20,6 +20,20 @@ common: fastapi[standard]: ">=0.115.6" gguf: ">=0.14.0" huggingface_hub[hf-transfer,hf_xet]: ">=0.31" + # CosyVoice2 inference (cosyvoice and matcha sources are vendored in Voice). + HyperPyYAML: "==1.2.2" + # HyperPyYAML 1.2.2 combines PyYAML's Loader with ruamel internals. + ruamel.yaml: ">=0.17.28,<0.19" + conformer: ">=0.3.2" + inflect: ">=7.3,<8" + librosa: "==0.11.0" + modelscope: "==1.26.0" + omegaconf: ">=2.3,<3" + openai-whisper: "==20240930" + soundfile: ">=0.12.1" + tiktoken: ">=0.9.0" + WeTextProcessing: "@ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl" + pynini: ">=2.1.6,<2.2" igraph: ">=0.11.8" kornia: ">=0.8.0" matplotlib: ">=3.10.1" diff --git a/dependencies/linux-amdgpu-dependencies.yaml b/dependencies/linux-amdgpu-dependencies.yaml index b88a5236..7e1a7feb 100644 --- a/dependencies/linux-amdgpu-dependencies.yaml +++ b/dependencies/linux-amdgpu-dependencies.yaml @@ -67,6 +67,10 @@ attrs: - cyclopts - jsonschema - referencing +audioread: + version: "3.1.0" + dependents: + - librosa babel: version: "2.18.0" dependents: @@ -95,6 +99,7 @@ cffi: dependents: - argon2-cffi-bindings - sounddevice + - soundfile charset-normalizer: version: "2.1.1" dependents: @@ -108,6 +113,10 @@ click: clip-anytorch: version: "2.6.0" isRoot: true +cloudpickle: + version: "3.1.2" + dependents: + - joblib comm: version: "0.2.3" dependents: @@ -115,6 +124,9 @@ comm: compel: version: "2.3.1" isRoot: true +conformer: + version: "0.3.2" + isRoot: true contourpy: version: "1.3.3" dependents: @@ -142,6 +154,7 @@ decorator: version: "5.3.1" dependents: - ipython + - librosa defusedxml: version: "0.7.1" dependents: @@ -185,6 +198,7 @@ einops: version: "0.8.2" isRoot: true dependents: + - conformer - controlnet-aux - diffsynth - spandrel @@ -224,7 +238,6 @@ filelock: - datasets - diffusers - huggingface-hub - - modelscope - python-discovery - torch - transformers @@ -303,6 +316,9 @@ huggingface-hub: - timm - tokenizers - transformers +hyperpyyaml: + version: "1.2.2" + isRoot: true idna: version: "3.4" dependents: @@ -329,6 +345,13 @@ importlib-metadata: dependents: - controlnet-aux - diffusers +importlib-resources: + version: "7.1.0" + dependents: + - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true ipykernel: version: "7.3.0" dependents: @@ -345,6 +368,10 @@ isoduration: version: "20.11.0" dependents: - jsonschema +jaraco-functools: + version: "4.6.0" + dependents: + - inflect jedi: version: "0.20.0" dependents: @@ -358,6 +385,11 @@ jinja2: - jupyterlab-server - nbconvert - torch +joblib: + version: "1.6.0" + dependents: + - librosa + - scikit-learn json5: version: "0.15.0" dependents: @@ -444,7 +476,11 @@ lark: lazy-loader: version: "0.5" dependents: + - librosa - scikit-image +librosa: + version: "0.11.0" + isRoot: true llvmlite: version: "0.47.0" dependents: @@ -486,16 +522,27 @@ ml-dtypes: dependents: - onnx modelscope: - version: "1.37.1" + version: "1.26.0" + isRoot: true dependents: - diffsynth moderngl: version: "5.12.0" isRoot: true +more-itertools: + version: "11.2.1" + dependents: + - inflect + - jaraco-functools + - openai-whisper mpmath: version: "1.3.0" dependents: - sympy +msgpack: + version: "1.2.4" + dependents: + - librosa multidict: version: "6.7.1" dependents: @@ -505,6 +552,10 @@ multiprocess: version: "0.70.19" dependents: - datasets +narwhals: + version: "2.28.0" + dependents: + - scikit-learn nbclient: version: "0.11.0" dependents: @@ -540,6 +591,8 @@ notebook-shim: numba: version: "0.65.1" dependents: + - librosa + - openai-whisper - pymatting numpy: version: "2.4.4" @@ -551,12 +604,14 @@ numpy: - diffusers - gguf - imageio + - librosa - matplotlib - mediapipe - ml-dtypes - numba - onnx - onnxruntime + - openai-whisper - opencv-contrib-python - opencv-python-headless - pandas @@ -567,19 +622,28 @@ numpy: - pyvista - rembg - scikit-image + - scikit-learn - scipy + - soundfile + - soxr - spandrel - tifffile - torchsde - torchvision - transformers - trimesh +omegaconf: + version: "2.4.0" + isRoot: true onnx: version: "1.22.0" isRoot: true onnxruntime: version: "1.27.0" isRoot: true +openai-whisper: + version: "20240930" + isRoot: true opencv-contrib-python: version: "4.13.0.92" dependents: @@ -602,7 +666,6 @@ packaging: - kornia - lazy-loader - matplotlib - - modelscope - nbconvert - onnxruntime - peft @@ -660,6 +723,7 @@ plyfile: pooch: version: "1.9.0" dependents: + - librosa - pyvista - rembg prometheus-client: @@ -742,6 +806,9 @@ pymatting: pymeshfix: version: "0.18.1" isRoot: true +pynini: + version: "2.1.7" + isRoot: true pyparsing: version: "3.3.2" dependents: @@ -781,7 +848,9 @@ pyyaml: - accelerate - gguf - huggingface-hub + - hyperpyyaml - jupyter-events + - omegaconf - peft - timm - transformers @@ -803,6 +872,7 @@ regex: dependents: - clip-anytorch - diffusers + - tiktoken - transformers rembg: version: "2.0.69" @@ -817,6 +887,7 @@ requests: - jupyterlab-server - modelscope - pooch + - tiktoken - transformers restrictedpython: version: "8.3" @@ -882,6 +953,15 @@ rpds-py: dependents: - jsonschema - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml safetensors: version: "0.8.0" isRoot: true @@ -898,14 +978,20 @@ scikit-image: dependents: - controlnet-aux - rembg +scikit-learn: + version: "1.9.1" + dependents: + - librosa scipy: version: "1.18.0" isRoot: true dependents: - controlnet-aux + - librosa - pymatting - rembg - scikit-image + - scikit-learn - torchsde scooby: version: "0.11.2" @@ -926,7 +1012,6 @@ sentry-sdk: setuptools: version: "70.2.0" dependents: - - modelscope - torch shellingham: version: "1.5.4" @@ -941,10 +1026,19 @@ sounddevice: version: "0.5.5" dependents: - mediapipe +soundfile: + version: "0.14.0" + isRoot: true + dependents: + - librosa soupsieve: version: "2.8.4" dependents: - beautifulsoup4 +soxr: + version: "1.1.0" + dependents: + - librosa spandrel: version: "0.4.2" isRoot: true @@ -969,10 +1063,19 @@ texttable: version: "1.7.0" dependents: - igraph +threadpoolctl: + version: "3.8.0" + dependents: + - scikit-learn tifffile: version: "2026.6.1" dependents: - scikit-image +tiktoken: + version: "0.14.0" + isRoot: true + dependents: + - openai-whisper timm: version: "1.0.27" dependents: @@ -994,10 +1097,12 @@ torch: - accelerate - clip-anytorch - compel + - conformer - controlnet-aux - diffsynth - diffusers - kornia + - openai-whisper - peft - spandrel - timm @@ -1038,6 +1143,7 @@ tqdm: - gguf - huggingface-hub - modelscope + - openai-whisper - peft - rembg - transformers @@ -1069,6 +1175,14 @@ transformers: trimesh: version: "4.12.2" isRoot: true +triton: + version: "3.9.0" + dependents: + - openai-whisper +typeguard: + version: "4.6.0" + dependents: + - inflect typer: version: "0.25.1" dependents: @@ -1084,6 +1198,7 @@ typing-extensions: - fastapi - huggingface-hub - jupyter-client + - librosa - onnx - pydantic - pydantic-core @@ -1091,9 +1206,11 @@ typing-extensions: - pyvista - referencing - rich-toolkit + - soundfile - spandrel - starlette - torch + - typeguard - typing-inspection typing-inspection: version: "0.4.2" @@ -1159,6 +1276,10 @@ websockets: isRoot: true dependents: - uvicorn +wetextprocessing: + version: "1.0.4" + url: "https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl" + isRoot: true wrapt: version: "2.2.2" dependents: diff --git a/dependencies/linux-amdgpu.lock b/dependencies/linux-amdgpu.lock index c44ae778..da5d38e0 100644 --- a/dependencies/linux-amdgpu.lock +++ b/dependencies/linux-amdgpu.lock @@ -44,6 +44,8 @@ attrs==26.1.0 # cyclopts # jsonschema # referencing +audioread==3.1.0 + # via librosa babel==2.18.0 # via jupyterlab-server beautifulsoup4==4.15.0 @@ -63,6 +65,7 @@ cffi==2.0.0 # via # argon2-cffi-bindings # sounddevice + # soundfile charset-normalizer==2.1.1 # via requests click==8.4.2 @@ -72,10 +75,14 @@ click==8.4.2 # uvicorn clip-anytorch==2.6.0 # via -r dependencies/requirements-linux-amdgpu.txt +cloudpickle==3.1.2 + # via joblib comm==0.2.3 # via ipykernel compel==2.3.1 # via -r dependencies/requirements-linux-amdgpu.txt +conformer==0.3.2 + # via -r dependencies/requirements-linux-amdgpu.txt contourpy==1.3.3 # via matplotlib controlnet-aux==0.0.10 @@ -89,7 +96,9 @@ datasets==1.1.1 debugpy==1.8.21 # via ipykernel decorator==5.3.1 - # via ipython + # via + # ipython + # librosa defusedxml==0.7.1 # via nbconvert deprecated==1.3.1 @@ -117,6 +126,7 @@ easydict==1.13 einops==0.8.2 # via # -r dependencies/requirements-linux-amdgpu.txt + # conformer # controlnet-aux # diffsynth # spandrel @@ -144,7 +154,6 @@ filelock==3.29.0 # datasets # diffusers # huggingface-hub - # modelscope # python-discovery # torch # transformers @@ -201,6 +210,8 @@ huggingface-hub==0.36.2 # timm # tokenizers # transformers +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-linux-amdgpu.txt idna==3.4 # via # anyio @@ -221,6 +232,10 @@ importlib-metadata==7.1.0 # via # controlnet-aux # diffusers +importlib-resources==7.1.0 + # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-linux-amdgpu.txt ipykernel==7.3.0 # via jupyterlab ipython==9.15.0 @@ -229,6 +244,8 @@ ipython-pygments-lexers==1.1.1 # via ipython isoduration==20.11.0 # via jsonschema +jaraco-functools==4.6.0 + # via inflect jedi==0.20.0 # via ipython jinja2==3.1.6 @@ -239,6 +256,10 @@ jinja2==3.1.6 # jupyterlab-server # nbconvert # torch +joblib==1.6.0 + # via + # librosa + # scikit-learn json5==0.15.0 # via jupyterlab-server jsonpointer==3.1.1 @@ -295,7 +316,11 @@ kornia-rs==0.1.14 lark==1.3.1 # via rfc3987-syntax lazy-loader==0.5 - # via scikit-image + # via + # librosa + # scikit-image +librosa==0.11.0 + # via -r dependencies/requirements-linux-amdgpu.txt llvmlite==0.47.0 # via numba markdown-it-py==4.2.0 @@ -322,18 +347,29 @@ mistune==3.3.2 # via nbconvert ml-dtypes==0.5.4 # via onnx -modelscope==1.37.1 - # via diffsynth +modelscope==1.26.0 + # via + # -r dependencies/requirements-linux-amdgpu.txt + # diffsynth moderngl==5.12.0 # via -r dependencies/requirements-linux-amdgpu.txt +more-itertools==11.2.1 + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy +msgpack==1.2.4 + # via librosa multidict==6.7.1 # via # aiohttp # yarl multiprocess==0.70.19 # via datasets +narwhals==2.28.0 + # via scikit-learn nbclient==0.11.0 # via nbconvert nbconvert==7.17.1 @@ -356,7 +392,10 @@ notebook-shim==0.2.4 # jupyterlab # notebook numba==0.65.1 - # via pymatting + # via + # librosa + # openai-whisper + # pymatting numpy==2.4.4 # via # accelerate @@ -366,12 +405,14 @@ numpy==2.4.4 # diffusers # gguf # imageio + # librosa # matplotlib # mediapipe # ml-dtypes # numba # onnx # onnxruntime + # openai-whisper # opencv-contrib-python # opencv-python-headless # pandas @@ -382,17 +423,24 @@ numpy==2.4.4 # pyvista # rembg # scikit-image + # scikit-learn # scipy + # soundfile + # soxr # spandrel # tifffile # torchsde # torchvision # transformers # trimesh +omegaconf==2.4.0 + # via -r dependencies/requirements-linux-amdgpu.txt onnx==1.22.0 # via -r dependencies/requirements-linux-amdgpu.txt onnxruntime==1.27.0 # via -r dependencies/requirements-linux-amdgpu.txt +openai-whisper==20240930 + # via -r dependencies/requirements-linux-amdgpu.txt opencv-contrib-python==4.13.0.92 # via mediapipe opencv-python-headless==4.13.0.92 @@ -411,7 +459,6 @@ packaging==24.1 # kornia # lazy-loader # matplotlib - # modelscope # nbconvert # onnxruntime # peft @@ -456,6 +503,7 @@ plyfile==1.1.4 # via -r dependencies/requirements-linux-amdgpu.txt pooch==1.9.0 # via + # librosa # pyvista # rembg prometheus-client==0.25.0 @@ -514,6 +562,8 @@ pymatting==1.1.15 # via rembg pymeshfix==0.18.1 # via -r dependencies/requirements-linux-amdgpu.txt +pynini==2.1.7 + # via -r dependencies/requirements-linux-amdgpu.txt pyparsing==3.3.2 # via # compel @@ -542,7 +592,9 @@ pyyaml==6.0.3 # accelerate # gguf # huggingface-hub + # hyperpyyaml # jupyter-events + # omegaconf # peft # timm # transformers @@ -561,6 +613,7 @@ regex==2026.5.9 # via # clip-anytorch # diffusers + # tiktoken # transformers rembg==2.0.69 # via -r dependencies/requirements-linux-amdgpu.txt @@ -573,6 +626,7 @@ requests==2.28.1 # jupyterlab-server # modelscope # pooch + # tiktoken # transformers restrictedpython==8.3 # via -r dependencies/requirements-linux-amdgpu.txt @@ -618,6 +672,12 @@ rpds-py==2026.5.1 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-linux-amdgpu.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-linux-amdgpu.txt @@ -632,13 +692,17 @@ scikit-image==0.26.0 # via # controlnet-aux # rembg +scikit-learn==1.9.1 + # via librosa scipy==1.18.0 # via # -r dependencies/requirements-linux-amdgpu.txt # controlnet-aux + # librosa # pymatting # rembg # scikit-image + # scikit-learn # torchsde scooby==0.11.2 # via pyvista @@ -649,9 +713,7 @@ sentencepiece==0.2.1 sentry-sdk==2.63.0 # via fastapi-cloud-cli setuptools==70.2.0 - # via - # modelscope - # torch + # via torch shellingham==1.5.4 # via typer six==1.17.0 @@ -660,8 +722,14 @@ six==1.17.0 # rfc3339-validator sounddevice==0.5.5 # via mediapipe +soundfile==0.14.0 + # via + # -r dependencies/requirements-linux-amdgpu.txt + # librosa soupsieve==2.8.4 # via beautifulsoup4 +soxr==1.1.0 + # via librosa spandrel==0.4.2 # via -r dependencies/requirements-linux-amdgpu.txt stack-data==0.6.3 @@ -676,8 +744,14 @@ terminado==0.18.1 # jupyter-server-terminals texttable==1.7.0 # via igraph +threadpoolctl==3.8.0 + # via scikit-learn tifffile==2026.6.1 # via scikit-image +tiktoken==0.14.0 + # via + # -r dependencies/requirements-linux-amdgpu.txt + # openai-whisper timm==1.0.27 # via controlnet-aux tinycss2==1.5.1 @@ -692,10 +766,12 @@ torch @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/torch-2.9.1%2Brocm7. # accelerate # clip-anytorch # compel + # conformer # controlnet-aux # diffsynth # diffusers # kornia + # openai-whisper # peft # spandrel # timm @@ -729,6 +805,7 @@ tqdm==4.66.5 # gguf # huggingface-hub # modelscope + # openai-whisper # peft # rembg # transformers @@ -755,6 +832,10 @@ transformers==4.57.6 # peft trimesh==4.12.2 # via -r dependencies/requirements-linux-amdgpu.txt +triton==3.9.0 + # via openai-whisper +typeguard==4.6.0 + # via inflect typer==0.25.1 # via # fastapi-cli @@ -768,6 +849,7 @@ typing-extensions==4.15.0 # fastapi # huggingface-hub # jupyter-client + # librosa # onnx # pydantic # pydantic-core @@ -775,9 +857,11 @@ typing-extensions==4.15.0 # pyvista # referencing # rich-toolkit + # soundfile # spandrel # starlette # torch + # typeguard # typing-inspection typing-inspection==0.4.2 # via @@ -822,6 +906,8 @@ websockets==16.0 # via # -r dependencies/requirements-linux-amdgpu.txt # uvicorn +wetextprocessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl + # via -r dependencies/requirements-linux-amdgpu.txt wrapt==2.2.2 # via deprecated xatlas==0.0.11 diff --git a/dependencies/linux-arm64-dependencies.yaml b/dependencies/linux-arm64-dependencies.yaml new file mode 100644 index 00000000..c74b18cd --- /dev/null +++ b/dependencies/linux-arm64-dependencies.yaml @@ -0,0 +1,1317 @@ +# 包依赖关系 +absl-py: + version: "2.5.1" + dependents: + - mediapipe +accelerate: + version: "1.15.0" + isRoot: true + dependents: + - diffsynth + - diffusers + - peft +agent-detector: + version: "2.0.0" + dependents: + - fastapi-cloud-cli +aiohappyeyeballs: + version: "2.7.1" + dependents: + - aiohttp +aiohttp: + version: "3.14.5" + isRoot: true +aioshutil: + version: "1.6" + isRoot: true +aiosignal: + version: "1.4.0" + dependents: + - aiohttp +annotated-doc: + version: "0.0.5" + dependents: + - fastapi + - typer +annotated-types: + version: "0.8.0" + dependents: + - pydantic +anyio: + version: "4.15.1" + dependents: + - httpx + - httpx2 + - jupyter-server + - starlette + - watchfiles +argon2-cffi: + version: "25.1.0" + dependents: + - jupyter-server +argon2-cffi-bindings: + version: "26.1.0" + dependents: + - argon2-cffi +arrow: + version: "1.4.0" + dependents: + - isoduration +asttokens: + version: "3.0.2" + dependents: + - stack-data +async-lru: + version: "2.4.0" + dependents: + - jupyterlab +attrs: + version: "26.1.0" + dependents: + - aiohttp + - cyclopts + - jsonschema + - referencing +audioread: + version: "3.1.0" + dependents: + - librosa +babel: + version: "2.18.0" + dependents: + - jupyterlab-server +beautifulsoup4: + version: "4.15.0" + dependents: + - nbconvert +blake3: + version: "1.0.11" + isRoot: true +bleach: + version: "6.4.0" + dependents: + - nbconvert +certifi: + version: "2022.12.7" + dependents: + - httpcore + - httpx + - mediapipe + - requests + - sentry-sdk +cffi: + version: "2.1.1" + dependents: + - argon2-cffi-bindings + - sounddevice + - soundfile +charset-normalizer: + version: "2.1.1" + dependents: + - requests +click: + version: "8.5.0" + dependents: + - rich-toolkit + - uvicorn +clip-anytorch: + version: "2.6.0" + isRoot: true +cloudpickle: + version: "3.1.2" + dependents: + - joblib +comm: + version: "0.2.3" + dependents: + - ipykernel +compel: + version: "2.3.1" + isRoot: true +conformer: + version: "0.3.2" + isRoot: true +contourpy: + version: "1.4.0" + dependents: + - matplotlib +controlnet-aux: + version: "0.0.10" + isRoot: true +cycler: + version: "0.12.1" + dependents: + - matplotlib +cyclopts: + version: "5.3.2" + dependents: + - pyvista +debugpy: + version: "1.8.22" + dependents: + - ipykernel +decorator: + version: "5.3.1" + dependents: + - librosa +defusedxml: + version: "0.7.1" + dependents: + - nbconvert +deprecated: + version: "3.0.0" + isRoot: true +detect-installer: + version: "0.2.1" + dependents: + - fastapi-cloud-cli +diffsynth: + version: "2.2.0" + isRoot: true +diffusers: + version: "0.39.0" + isRoot: true + dependents: + - compel +distlib: + version: "0.4.4" + dependents: + - virtualenv +dnspython: + version: "2.9.1" + dependents: + - email-validator +docstring-parser: + version: "0.18.0" + dependents: + - cyclopts +easydict: + version: "1.13" + isRoot: true +einops: + version: "0.8.2" + isRoot: true + dependents: + - conformer + - controlnet-aux + - diffsynth + - spandrel +email-validator: + version: "2.3.0" + dependents: + - fastapi + - pydantic +executing: + version: "2.3.0" + dependents: + - stack-data +fastapi: + version: "0.162.0" + isRoot: true +fastapi-cli: + version: "0.0.32" + dependents: + - fastapi +fastapi-cloud-cli: + version: "0.26.0" + dependents: + - fastapi-cli +fastar: + version: "0.12.0" + dependents: + - fastapi + - fastapi-cloud-cli +fastjsonschema: + version: "2.22.2" + dependents: + - nbformat +filelock: + version: "4.2.4" + dependents: + - controlnet-aux + - diffusers + - huggingface-hub + - python-discovery + - torch + - transformers + - virtualenv +flatbuffers: + version: "25.12.19" + dependents: + - mediapipe + - onnxruntime +fonttools: + version: "4.68.0" + dependents: + - matplotlib +fqdn: + version: "1.6.0" + dependents: + - jsonschema +frozenlist: + version: "1.8.0" + dependents: + - aiohttp + - aiosignal +fsspec: + version: "2026.9.0" + dependents: + - huggingface-hub + - torch +ftfy: + version: "6.3.1" + dependents: + - clip-anytorch + - diffsynth +gguf: + version: "0.19.0" + isRoot: true +glcontext: + version: "3.1.0" + dependents: + - moderngl +googleapis-common-protos: + version: "1.76.0" + dependents: + - opentelemetry-exporter-otlp-proto-http +h11: + version: "0.16.0" + dependents: + - httpcore + - httpcore2 + - uvicorn +hf-transfer: + version: "0.1.9" + dependents: + - huggingface-hub +hf-xet: + version: "1.7.1" + dependents: + - huggingface-hub +httpcore: + version: "1.0.9" + dependents: + - httpx +httpcore2: + version: "2.3.0" + dependents: + - httpx2 +httptools: + version: "0.9.0" + dependents: + - uvicorn +httpx: + version: "0.28.1" + dependents: + - diffusers + - fastapi + - fastapi-cloud-cli + - jupyterlab +httpx2: + version: "2.3.0" + dependents: + - fastapi +huggingface-hub: + version: "0.36.2" + isRoot: true + dependents: + - accelerate + - controlnet-aux + - diffusers + - peft + - timm + - tokenizers + - transformers +hyperpyyaml: + version: "1.2.2" + isRoot: true +idna: + version: "3.4" + dependents: + - anyio + - email-validator + - httpx + - httpx2 + - jsonschema + - requests + - yarl +igraph: + version: "1.0.0" + isRoot: true +imageio: + version: "2.38.1" + dependents: + - diffsynth + - scikit-image +imageio-ffmpeg: + version: "0.7.0" + dependents: + - imageio +importlib-metadata: + version: "7.1.0" + dependents: + - controlnet-aux + - diffusers +importlib-resources: + version: "7.1.0" + dependents: + - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true +ipykernel: + version: "7.4.0" + dependents: + - jupyterlab +ipython: + version: "9.17.1" + dependents: + - ipykernel +ipython-pygments-lexers: + version: "1.1.1" + dependents: + - ipython +isoduration: + version: "20.11.0" + dependents: + - jsonschema +jaraco-functools: + version: "4.6.0" + dependents: + - inflect +jedi: + version: "0.20.1" + dependents: + - ipython +jinja2: + version: "3.1.6" + dependents: + - fastapi + - jupyter-server + - jupyterlab + - jupyterlab-server + - nbconvert + - torch +joblib: + version: "1.6.0" + dependents: + - librosa + - scikit-learn +json5: + version: "0.17.3" + dependents: + - jupyterlab-server +jsonpointer: + version: "3.6" + dependents: + - jsonschema +jsonschema: + version: "4.26.0" + dependents: + - jupyter-events + - jupyterlab-server + - nbformat + - rembg +jsonschema-specifications: + version: "2025.9.1" + dependents: + - jsonschema +jupyter-client: + version: "8.10.0" + dependents: + - ipykernel + - jupyter-server + - nbclient +jupyter-core: + version: "5.9.1" + dependents: + - ipykernel + - jupyter-client + - jupyter-server + - jupyterlab + - nbclient + - nbconvert + - nbformat +jupyter-events: + version: "0.12.1" + dependents: + - jupyter-server +jupyter-lsp: + version: "2.3.1" + dependents: + - jupyterlab +jupyter-server: + version: "2.21.1" + dependents: + - jupyter-lsp + - jupyterlab + - jupyterlab-server + - notebook + - notebook-shim +jupyter-server-terminals: + version: "0.5.4" + dependents: + - jupyter-server +jupyterlab: + version: "4.1.6" + dependents: + - notebook +jupyterlab-pygments: + version: "0.3.0" + dependents: + - nbconvert +jupyterlab-server: + version: "2.24.0" + dependents: + - jupyterlab + - notebook +kiwisolver: + version: "1.5.1" + dependents: + - matplotlib +kornia: + version: "0.9.0" + isRoot: true +kornia-rs: + version: "0.2.0" + dependents: + - kornia +lark: + version: "1.3.1" + dependents: + - rfc3987-syntax +lazy-loader: + version: "0.6" + dependents: + - librosa + - scikit-image +librosa: + version: "0.11.0" + isRoot: true +llvmlite: + version: "0.50.0" + dependents: + - numba +markdown-it-py: + version: "4.2.0" + dependents: + - rich +markupsafe: + version: "3.0.3" + dependents: + - jinja2 + - nbconvert +matplotlib: + version: "3.11.2" + isRoot: true + dependents: + - mediapipe + - pyvista + - vtk +matplotlib-inline: + version: "0.2.2" + dependents: + - ipykernel + - ipython +mdurl: + version: "0.1.2" + dependents: + - markdown-it-py +mediapipe: + version: "1.2.0" + isRoot: true +mistune: + version: "3.4.0" + dependents: + - nbconvert +ml-dtypes: + version: "0.6.0" + dependents: + - onnx +modelscope: + version: "1.26.0" + isRoot: true + dependents: + - diffsynth +moderngl: + version: "5.13.0" + isRoot: true +more-itertools: + version: "11.2.1" + dependents: + - inflect + - jaraco-functools + - openai-whisper +mpmath: + version: "1.3.0" + dependents: + - sympy +msgpack: + version: "1.2.4" + dependents: + - librosa +multidict: + version: "7.2.1" + dependents: + - aiohttp + - yarl +narwhals: + version: "2.28.0" + dependents: + - scikit-learn +nbclient: + version: "0.11.0" + dependents: + - nbconvert +nbconvert: + version: "7.17.2" + dependents: + - jupyter-server +nbformat: + version: "5.11.1" + dependents: + - jupyter-server + - nbclient + - nbconvert +nest-asyncio2: + version: "1.7.4" + dependents: + - ipykernel +networkx: + version: "3.7" + dependents: + - scikit-image + - torch +notebook: + version: "7.1.3" + dependents: + - compel +notebook-shim: + version: "0.2.4" + dependents: + - jupyterlab + - notebook +numba: + version: "0.68.1" + dependents: + - librosa + - openai-whisper + - pymatting +numpy: + version: "2.5.4" + dependents: + - accelerate + - contourpy + - controlnet-aux + - diffusers + - gguf + - imageio + - kornia + - librosa + - matplotlib + - mediapipe + - ml-dtypes + - numba + - onnx + - onnxruntime + - openai-whisper + - opencv-contrib-python + - opencv-python-headless + - pandas + - peft + - plyfile + - pymatting + - pymeshfix + - pyvista + - pyvista-validation + - rembg + - scikit-image + - scikit-learn + - scipy + - soundfile + - soxr + - spandrel + - tifffile + - torchsde + - torchvision + - transformers + - trimesh +omegaconf: + version: "2.4.0" + isRoot: true +onnx: + version: "1.23.2" + isRoot: true +onnxruntime: + version: "1.31.1" + isRoot: true +openai-whisper: + version: "20240930" + isRoot: true +opencv-contrib-python: + version: "5.0.0.93" + dependents: + - mediapipe +opencv-python-headless: + version: "5.0.0.93" + dependents: + - controlnet-aux + - rembg +opentelemetry-api: + version: "1.45.2" + dependents: + - fastapi + - opentelemetry-exporter-http-transport + - opentelemetry-exporter-otlp-proto-http + - opentelemetry-sdk + - opentelemetry-semantic-conventions +opentelemetry-exporter-http-transport: + version: "0.66b2" + dependents: + - opentelemetry-exporter-otlp-proto-http +opentelemetry-exporter-otlp-common: + version: "0.66b2" + dependents: + - opentelemetry-exporter-otlp-proto-http +opentelemetry-exporter-otlp-proto-common: + version: "1.45.2" + dependents: + - opentelemetry-exporter-otlp-proto-http +opentelemetry-exporter-otlp-proto-http: + version: "1.45.2" + dependents: + - fastapi +opentelemetry-proto: + version: "1.45.2" + dependents: + - opentelemetry-exporter-otlp-proto-common + - opentelemetry-exporter-otlp-proto-http +opentelemetry-sdk: + version: "1.45.2" + dependents: + - fastapi + - opentelemetry-exporter-otlp-common + - opentelemetry-exporter-otlp-proto-http +opentelemetry-semantic-conventions: + version: "0.66b2" + dependents: + - opentelemetry-sdk +packaging: + version: "24.1" + dependents: + - accelerate + - huggingface-hub + - ipykernel + - jupyter-events + - jupyter-server + - jupyterlab + - jupyterlab-server + - lazy-loader + - matplotlib + - nbconvert + - onnxruntime + - peft + - pooch + - scikit-image + - transformers + - virtualenv +pandas: + version: "3.1.0" + dependents: + - diffsynth +pandocfilters: + version: "1.5.1" + dependents: + - nbconvert +parso: + version: "0.8.7" + dependents: + - jedi +peft: + version: "0.17.1" + isRoot: true + dependents: + - diffsynth +pexpect: + version: "4.9.0" + dependents: + - ipython +picklescan: + version: "1.0.5" + isRoot: true +pillow: + version: "13.0.0" + isRoot: true + dependents: + - controlnet-aux + - diffusers + - imageio + - matplotlib + - pymatting + - pyvista + - rembg + - scikit-image + - torchvision +platformdirs: + version: "4.14.0" + dependents: + - jupyter-core + - pooch + - virtualenv +plyfile: + version: "1.1.5" + isRoot: true +pooch: + version: "1.9.0" + dependents: + - librosa + - pyvista + - rembg +prometheus-client: + version: "0.26.0" + dependents: + - jupyter-server +prompt-toolkit: + version: "3.0.53" + dependents: + - ipython +propcache: + version: "0.5.4" + dependents: + - aiohttp + - yarl +protobuf: + version: "7.36.2" + dependents: + - googleapis-common-protos + - onnx + - onnxruntime + - opentelemetry-proto +psutil: + version: "7.2.2" + dependents: + - accelerate + - imageio + - ipython + - peft +ptyprocess: + version: "0.7.0" + dependents: + - pexpect + - terminado +pure-eval: + version: "0.2.4" + dependents: + - stack-data +pycparser: + version: "3.12" + dependents: + - cffi +pydantic: + version: "2.14.2" + isRoot: true + dependents: + - fastapi + - fastapi-cloud-cli + - pydantic-extra-types + - pydantic-settings +pydantic-core: + version: "2.50.2" + dependents: + - pydantic +pydantic-extra-types: + version: "2.11.1" + dependents: + - fastapi +pydantic-settings: + version: "2.16.2" + isRoot: true + dependents: + - fastapi +pygments: + version: "2.21.0" + dependents: + - ipython + - ipython-pygments-lexers + - nbconvert + - rich + - rich-rst +pymatting: + version: "1.1.16" + dependents: + - rembg +pymeshfix: + version: "0.18.1" + isRoot: true +pynini: + version: "2.1.7" + isRoot: true +pyparsing: + version: "3.3.3" + dependents: + - compel + - matplotlib +python-dateutil: + version: "3.0.0" + dependents: + - arrow + - jupyter-client + - matplotlib + - pandas +python-discovery: + version: "1.6.7" + dependents: + - virtualenv +python-dotenv: + version: "1.2.4" + dependents: + - pydantic-settings + - uvicorn +python-json-logger: + version: "4.2.0" + dependents: + - jupyter-events +python-multipart: + version: "0.0.32" + dependents: + - fastapi +pyvista: + version: "0.49.2" + isRoot: true +pyvista-validation: + version: "0.2.2" + dependents: + - pyvista +pyyaml: + version: "6.0.3" + isRoot: true + dependents: + - accelerate + - gguf + - huggingface-hub + - hyperpyyaml + - jupyter-events + - omegaconf + - peft + - timm + - transformers + - uvicorn +pyzmq: + version: "27.2.0" + dependents: + - ipykernel + - jupyter-client + - jupyter-server +referencing: + version: "0.37.0" + dependents: + - jsonschema + - jsonschema-specifications + - jupyter-events +regex: + version: "2026.9.29" + dependents: + - clip-anytorch + - diffusers + - tiktoken + - transformers +rembg: + version: "2.0.69" + isRoot: true +requests: + version: "2.28.1" + dependents: + - diffusers + - gguf + - huggingface-hub + - jupyterlab-server + - modelscope + - opentelemetry-exporter-http-transport + - opentelemetry-exporter-otlp-proto-http + - pooch + - tiktoken + - transformers +restrictedpython: + version: "8.5" + isRoot: true +rfc3339-validator: + version: "0.1.4" + dependents: + - jsonschema + - jupyter-events +rfc3986-validator: + version: "0.1.1" + dependents: + - jsonschema + - jupyter-events +rfc3987-syntax: + version: "1.1.0" + dependents: + - jsonschema +rich: + version: "15.0.0" + dependents: + - cyclopts + - rich-rst + - rich-toolkit + - typer +rich-rst: + version: "2.2.0" + dependents: + - cyclopts +rich-toolkit: + version: "0.20.6" + dependents: + - fastapi-cli + - fastapi-cloud-cli +rignore: + version: "0.8.1" + dependents: + - fastapi-cloud-cli +rpds-py: + version: "2026.9.1" + dependents: + - jsonschema + - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml +safetensors: + version: "0.8.0" + isRoot: true + dependents: + - accelerate + - diffsynth + - diffusers + - peft + - spandrel + - timm + - transformers +scikit-image: + version: "0.26.0" + dependents: + - controlnet-aux + - rembg +scikit-learn: + version: "1.9.1" + dependents: + - librosa +scipy: + version: "1.18.1" + isRoot: true + dependents: + - controlnet-aux + - librosa + - pymatting + - rembg + - scikit-image + - scikit-learn + - torchsde +scooby: + version: "0.12.1" + dependents: + - pyvista +send2trash: + version: "2.1.0" + dependents: + - jupyter-server +sentencepiece: + version: "0.2.0" + isRoot: true +sentry-sdk: + version: "2.73.0" + dependents: + - fastapi-cloud-cli +setuptools: + version: "78.1.0" + dependents: + - torch +shellingham: + version: "1.5.4" + dependents: + - typer +six: + version: "1.17.0" + dependents: + - python-dateutil + - rfc3339-validator +sounddevice: + version: "0.5.6" + dependents: + - mediapipe +soundfile: + version: "0.14.0" + isRoot: true + dependents: + - librosa +soupsieve: + version: "3.0.4" + dependents: + - beautifulsoup4 +soxr: + version: "1.1.0" + dependents: + - librosa +spandrel: + version: "0.4.2" + isRoot: true +stack-data: + version: "0.6.3" + dependents: + - ipython +starlette: + version: "1.10.0" + dependents: + - fastapi +sympy: + version: "1.14.0" + dependents: + - torch +terminado: + version: "0.18.1" + dependents: + - jupyter-server + - jupyter-server-terminals +texttable: + version: "1.7.2" + dependents: + - igraph +threadpoolctl: + version: "3.8.0" + dependents: + - scikit-learn +tifffile: + version: "2026.10.15" + dependents: + - scikit-image +tiktoken: + version: "0.14.0" + isRoot: true + dependents: + - openai-whisper +timm: + version: "1.0.30" + dependents: + - controlnet-aux +tinycss2: + version: "1.5.1" + dependents: + - bleach +tokenizers: + version: "0.22.2" + isRoot: true + dependents: + - transformers +torch: + version: "2.8.0+cpu" + isRoot: true + dependents: + - accelerate + - clip-anytorch + - compel + - conformer + - controlnet-aux + - diffsynth + - diffusers + - kornia + - openai-whisper + - peft + - spandrel + - timm + - torchaudio + - torchsde + - torchvision +torchaudio: + version: "2.8.0" + isRoot: true +torchsde: + version: "0.2.6" + isRoot: true +torchvision: + version: "0.23.0" + isRoot: true + dependents: + - clip-anytorch + - controlnet-aux + - diffsynth + - spandrel + - timm +tornado: + version: "6.6" + dependents: + - ipykernel + - jupyter-client + - jupyter-server + - jupyterlab + - notebook + - terminado +tqdm: + version: "4.66.5" + dependents: + - clip-anytorch + - gguf + - huggingface-hub + - modelscope + - openai-whisper + - peft + - rembg + - transformers +traitlets: + version: "5.16.1" + dependents: + - ipykernel + - ipython + - jupyter-client + - jupyter-core + - jupyter-events + - jupyter-server + - jupyterlab + - matplotlib-inline + - nbclient + - nbconvert + - nbformat +trampoline: + version: "0.1.2" + dependents: + - torchsde +transformers: + version: "4.57.6" + isRoot: true + dependents: + - compel + - diffsynth + - peft +trimesh: + version: "5.2.0" + isRoot: true +truststore: + version: "0.10.5" + dependents: + - httpcore2 + - httpx2 +typeguard: + version: "4.6.0" + dependents: + - inflect +typer: + version: "0.27.3" + dependents: + - fastapi-cli + - fastapi-cloud-cli +typing-extensions: + version: "4.16.0" + dependents: + - aiohttp + - aiosignal + - anyio + - beautifulsoup4 + - fastapi + - huggingface-hub + - jupyter-client + - librosa + - onnx + - opentelemetry-api + - opentelemetry-exporter-otlp-proto-http + - opentelemetry-sdk + - opentelemetry-semantic-conventions + - pydantic + - pydantic-core + - pydantic-extra-types + - pydantic-settings + - pyvista + - pyvista-validation + - referencing + - rich-toolkit + - soundfile + - spandrel + - starlette + - torch + - typeguard + - typing-inspection +typing-inspection: + version: "0.4.4" + dependents: + - fastapi + - pydantic + - pydantic-settings +tzdata: + version: "2026.6" + dependents: + - arrow + - python-dateutil +uri-template: + version: "1.3.0" + dependents: + - jsonschema +urllib3: + version: "1.26.13" + dependents: + - modelscope + - requests + - sentry-sdk +uvicorn: + version: "0.54.0" + dependents: + - fastapi + - fastapi-cli + - fastapi-cloud-cli +uvloop: + version: "0.23.0" + dependents: + - uvicorn +virtualenv: + version: "21.14.5" + isRoot: true +vtk: + version: "9.7.1" + dependents: + - pyvista +watchfiles: + version: "1.3.0" + dependents: + - uvicorn +wcwidth: + version: "0.9.4" + dependents: + - ftfy + - prompt-toolkit +webcolors: + version: "25.10.0" + dependents: + - jsonschema +webencodings: + version: "0.6.1" + dependents: + - bleach + - tinycss2 +websocket-client: + version: "1.9.2" + dependents: + - jupyter-server +websockets: + version: "17.3" + isRoot: true + dependents: + - uvicorn +wetextprocessing: + version: "1.0.4" + url: "https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl" + isRoot: true +wrapt: + version: "2.5.2" + dependents: + - deprecated +yarl: + version: "1.25.1" + dependents: + - aiohttp +zipp: + version: "3.19.2" + dependents: + - importlib-metadata diff --git a/dependencies/linux-arm64.lock b/dependencies/linux-arm64.lock index 7697c2ea..7911e591 100644 --- a/dependencies/linux-arm64.lock +++ b/dependencies/linux-arm64.lock @@ -1,8 +1,5 @@ # This file was autogenerated by uv via the following command: -# uv pip compile dependencies/requirements-linux-arm64.txt --python-version 3.12.8 --python-platform aarch64-manylinux_2_28 --index-strategy unsafe-best-match --emit-index-url -o dependencies/linux-arm64.lock ---index-url https://pypi.org/simple ---extra-index-url https://download.pytorch.org/whl/cpu - +# uv pip compile dependencies\requirements-linux-arm64.txt -o dependencies\linux-arm64.lock --python-platform aarch64-unknown-linux-gnu --python-version 3.12.8 absl-py==2.5.1 # via mediapipe accelerate==1.15.0 @@ -50,6 +47,8 @@ attrs==26.1.0 # cyclopts # jsonschema # referencing +audioread==3.1.0 + # via librosa babel==2.18.0 # via jupyterlab-server beautifulsoup4==4.15.0 @@ -58,7 +57,7 @@ blake3==1.0.11 # via -r dependencies/requirements-linux-arm64.txt bleach==6.4.0 # via nbconvert -certifi==2026.10.23 +certifi==2022.12.7 # via # httpcore # httpx @@ -68,9 +67,9 @@ certifi==2026.10.23 cffi==2.1.1 # via # argon2-cffi-bindings - # cryptography # sounddevice -charset-normalizer==3.5.2 + # soundfile +charset-normalizer==2.1.1 # via requests click==8.5.0 # via @@ -78,22 +77,26 @@ click==8.5.0 # uvicorn clip-anytorch==2.6.0 # via -r dependencies/requirements-linux-arm64.txt +cloudpickle==3.1.2 + # via joblib comm==0.2.3 # via ipykernel compel==2.3.1 # via -r dependencies/requirements-linux-arm64.txt +conformer==0.3.2 + # via -r dependencies/requirements-linux-arm64.txt contourpy==1.4.0 # via matplotlib controlnet-aux==0.0.10 # via -r dependencies/requirements-linux-arm64.txt -cryptography==50.0.2 - # via modelscope-hub cycler==0.12.1 # via matplotlib cyclopts==5.3.2 # via pyvista debugpy==1.8.22 # via ipykernel +decorator==5.3.1 + # via librosa defusedxml==0.7.1 # via nbconvert deprecated==3.0.0 @@ -117,6 +120,7 @@ easydict==1.13 einops==0.8.2 # via # -r dependencies/requirements-linux-arm64.txt + # conformer # controlnet-aux # diffsynth # spandrel @@ -143,8 +147,6 @@ filelock==4.2.4 # controlnet-aux # diffusers # huggingface-hub - # modelscope - # modelscope-hub # python-discovery # torch # transformers @@ -186,7 +188,7 @@ hf-xet==1.7.1 # via huggingface-hub httpcore==1.0.9 # via httpx -httpcore2==2.16.0 +httpcore2==2.3.0 # via httpx2 httptools==0.9.0 # via uvicorn @@ -196,7 +198,7 @@ httpx==0.28.1 # fastapi # fastapi-cloud-cli # jupyterlab -httpx2==2.16.0 +httpx2==2.3.0 # via fastapi huggingface-hub==0.36.2 # via @@ -208,7 +210,9 @@ huggingface-hub==0.36.2 # timm # tokenizers # transformers -idna==3.20 +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-linux-arm64.txt +idna==3.4 # via # anyio # email-validator @@ -225,10 +229,14 @@ imageio==2.38.1 # scikit-image imageio-ffmpeg==0.7.0 # via imageio -importlib-metadata==10.1.0 +importlib-metadata==7.1.0 # via # controlnet-aux # diffusers +importlib-resources==7.1.0 + # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-linux-arm64.txt ipykernel==7.4.0 # via jupyterlab ipython==9.17.1 @@ -237,6 +245,8 @@ ipython-pygments-lexers==1.1.1 # via ipython isoduration==20.11.0 # via jsonschema +jaraco-functools==4.6.0 + # via inflect jedi==0.20.1 # via ipython jinja2==3.1.6 @@ -247,6 +257,10 @@ jinja2==3.1.6 # jupyterlab-server # nbconvert # torch +joblib==1.6.0 + # via + # librosa + # scikit-learn json5==0.17.3 # via jupyterlab-server jsonpointer==3.6 @@ -259,10 +273,6 @@ jsonschema==4.26.0 # rembg jsonschema-specifications==2025.9.1 # via jsonschema -jupyter-builder==1.2.4 - # via - # jupyterlab - # notebook jupyter-client==8.10.0 # via # ipykernel @@ -271,7 +281,6 @@ jupyter-client==8.10.0 jupyter-core==5.9.1 # via # ipykernel - # jupyter-builder # jupyter-client # jupyter-server # jupyterlab @@ -291,11 +300,11 @@ jupyter-server==2.21.1 # notebook-shim jupyter-server-terminals==0.5.4 # via jupyter-server -jupyterlab==4.6.5 +jupyterlab==4.1.6 # via notebook jupyterlab-pygments==0.3.0 # via nbconvert -jupyterlab-server==2.28.1 +jupyterlab-server==2.24.0 # via # jupyterlab # notebook @@ -308,12 +317,16 @@ kornia-rs==0.2.0 lark==1.3.1 # via rfc3987-syntax lazy-loader==0.6 - # via scikit-image + # via + # librosa + # scikit-image +librosa==0.11.0 + # via -r dependencies/requirements-linux-arm64.txt llvmlite==0.50.0 # via numba markdown-it-py==4.2.0 # via rich -markupsafe==3.0.4 +markupsafe==3.0.3 # via # jinja2 # nbconvert @@ -335,18 +348,27 @@ mistune==3.4.0 # via nbconvert ml-dtypes==0.6.0 # via onnx -modelscope==1.40.2 - # via diffsynth -modelscope-hub==0.6.1 - # via modelscope +modelscope==1.26.0 + # via + # -r dependencies/requirements-linux-arm64.txt + # diffsynth moderngl==5.13.0 # via -r dependencies/requirements-linux-arm64.txt +more-itertools==11.2.1 + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy +msgpack==1.2.4 + # via librosa multidict==7.2.1 # via # aiohttp # yarl +narwhals==2.28.0 + # via scikit-learn nbclient==0.11.0 # via nbconvert nbconvert==7.17.2 @@ -362,14 +384,17 @@ networkx==3.7 # via # scikit-image # torch -notebook==7.6.3 +notebook==7.1.3 # via compel notebook-shim==0.2.4 # via # jupyterlab # notebook numba==0.68.1 - # via pymatting + # via + # librosa + # openai-whisper + # pymatting numpy==2.5.4 # via # accelerate @@ -379,12 +404,14 @@ numpy==2.5.4 # gguf # imageio # kornia + # librosa # matplotlib # mediapipe # ml-dtypes # numba # onnx # onnxruntime + # openai-whisper # opencv-contrib-python # opencv-python-headless # pandas @@ -396,21 +423,30 @@ numpy==2.5.4 # pyvista-validation # rembg # scikit-image + # scikit-learn # scipy + # soundfile + # soxr # spandrel # tifffile # torchsde # torchvision # transformers # trimesh +omegaconf==2.4.0 + # via -r dependencies/requirements-linux-arm64.txt onnx==1.23.2 # via -r dependencies/requirements-linux-arm64.txt onnxruntime==1.31.1 # via -r dependencies/requirements-linux-arm64.txt +openai-whisper==20240930 + # via -r dependencies/requirements-linux-arm64.txt opencv-contrib-python==5.0.0.93 # via mediapipe opencv-python-headless==5.0.0.93 - # via controlnet-aux + # via + # controlnet-aux + # rembg opentelemetry-api==1.45.2 # via # fastapi @@ -437,7 +473,7 @@ opentelemetry-sdk==1.45.2 # opentelemetry-exporter-otlp-proto-http opentelemetry-semantic-conventions==0.66b2 # via opentelemetry-sdk -packaging==26.3 +packaging==24.1 # via # accelerate # huggingface-hub @@ -448,7 +484,6 @@ packaging==26.3 # jupyterlab-server # lazy-loader # matplotlib - # modelscope # nbconvert # onnxruntime # peft @@ -491,6 +526,7 @@ plyfile==1.1.5 # via -r dependencies/requirements-linux-arm64.txt pooch==1.9.0 # via + # librosa # pyvista # rembg prometheus-client==0.26.0 @@ -547,6 +583,8 @@ pymatting==1.1.16 # via rembg pymeshfix==0.18.1 # via -r dependencies/requirements-linux-arm64.txt +pynini==2.1.7 + # via -r dependencies/requirements-linux-arm64.txt pyparsing==3.3.3 # via # compel @@ -577,7 +615,9 @@ pyyaml==6.0.3 # accelerate # gguf # huggingface-hub + # hyperpyyaml # jupyter-events + # omegaconf # peft # timm # transformers @@ -596,20 +636,21 @@ regex==2026.9.29 # via # clip-anytorch # diffusers + # tiktoken # transformers -rembg==2.0.86 +rembg==2.0.69 # via -r dependencies/requirements-linux-arm64.txt -requests==2.34.2 +requests==2.28.1 # via # diffusers # gguf # huggingface-hub # jupyterlab-server # modelscope - # modelscope-hub # opentelemetry-exporter-http-transport # opentelemetry-exporter-otlp-proto-http # pooch + # tiktoken # transformers restrictedpython==8.5 # via -r dependencies/requirements-linux-arm64.txt @@ -641,6 +682,12 @@ rpds-py==2026.9.1 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-linux-arm64.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-linux-arm64.txt @@ -655,13 +702,17 @@ scikit-image==0.26.0 # via # controlnet-aux # rembg +scikit-learn==1.9.1 + # via librosa scipy==1.18.1 # via # -r dependencies/requirements-linux-arm64.txt # controlnet-aux + # librosa # pymatting # rembg # scikit-image + # scikit-learn # torchsde scooby==0.12.1 # via pyvista @@ -671,10 +722,8 @@ sentencepiece==0.2.0 # via -r dependencies/requirements-linux-arm64.txt sentry-sdk==2.73.0 # via fastapi-cloud-cli -setuptools==84.2.1 - # via - # modelscope - # torch +setuptools==78.1.0 + # via torch shellingham==1.5.4 # via typer six==1.17.0 @@ -683,8 +732,14 @@ six==1.17.0 # rfc3339-validator sounddevice==0.5.6 # via mediapipe +soundfile==0.14.0 + # via + # -r dependencies/requirements-linux-arm64.txt + # librosa soupsieve==3.0.4 # via beautifulsoup4 +soxr==1.1.0 + # via librosa spandrel==0.4.2 # via -r dependencies/requirements-linux-arm64.txt stack-data==0.6.3 @@ -699,8 +754,14 @@ terminado==0.18.1 # jupyter-server-terminals texttable==1.7.2 # via igraph +threadpoolctl==3.8.0 + # via scikit-learn tifffile==2026.10.15 # via scikit-image +tiktoken==0.14.0 + # via + # -r dependencies/requirements-linux-arm64.txt + # openai-whisper timm==1.0.30 # via controlnet-aux tinycss2==1.5.1 @@ -715,10 +776,12 @@ torch==2.8.0+cpu # accelerate # clip-anytorch # compel + # conformer # controlnet-aux # diffsynth # diffusers # kornia + # openai-whisper # peft # spandrel # timm @@ -745,13 +808,13 @@ tornado==6.6 # jupyterlab # notebook # terminado -tqdm==4.70.1 +tqdm==4.66.5 # via # clip-anytorch # gguf # huggingface-hub # modelscope - # modelscope-hub + # openai-whisper # peft # rembg # transformers @@ -759,7 +822,6 @@ traitlets==5.16.1 # via # ipykernel # ipython - # jupyter-builder # jupyter-client # jupyter-core # jupyter-events @@ -783,6 +845,8 @@ truststore==0.10.5 # via # httpcore2 # httpx2 +typeguard==4.6.0 + # via inflect typer==0.27.3 # via # fastapi-cli @@ -794,9 +858,9 @@ typing-extensions==4.16.0 # anyio # beautifulsoup4 # fastapi - # httpx2 # huggingface-hub # jupyter-client + # librosa # onnx # opentelemetry-api # opentelemetry-exporter-otlp-proto-http @@ -810,9 +874,11 @@ typing-extensions==4.16.0 # pyvista-validation # referencing # rich-toolkit + # soundfile # spandrel # starlette # torch + # typeguard # typing-inspection typing-inspection==0.4.4 # via @@ -825,10 +891,9 @@ tzdata==2026.6 # python-dateutil uri-template==1.3.0 # via jsonschema -urllib3==2.8.0 +urllib3==1.26.13 # via # modelscope - # modelscope-hub # requests # sentry-sdk uvicorn==0.54.0 @@ -838,7 +903,7 @@ uvicorn==0.54.0 # fastapi-cloud-cli uvloop==0.23.0 # via uvicorn -virtualenv==21.15.1 +virtualenv==21.14.5 # via -r dependencies/requirements-linux-arm64.txt vtk==9.7.1 # via pyvista @@ -860,9 +925,12 @@ websockets==17.3 # via # -r dependencies/requirements-linux-arm64.txt # uvicorn +wetextprocessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl + # via -r dependencies/requirements-linux-arm64.txt wrapt==2.5.2 # via deprecated yarl==1.25.1 # via aiohttp -zipp==4.1.2 +zipp==3.19.2 # via importlib-metadata +--extra-index-url https://download.pytorch.org/whl/cpu diff --git a/dependencies/linux-dependencies.yaml b/dependencies/linux-dependencies.yaml index 89606e61..323aba77 100644 --- a/dependencies/linux-dependencies.yaml +++ b/dependencies/linux-dependencies.yaml @@ -41,6 +41,10 @@ attrs: - jsonschema - mediapipe - referencing +audioread: + version: "3.1.0" + dependents: + - librosa blake3: version: "1.0.5" isRoot: true @@ -54,6 +58,7 @@ cffi: version: "1.17.1" dependents: - sounddevice + - soundfile charset-normalizer: version: "2.1.1" dependents: @@ -67,6 +72,10 @@ click: clip-anytorch: version: "2.6.0" isRoot: true +cloudpickle: + version: "3.1.2" + dependents: + - joblib coloredlogs: version: "15.0.1" dependents: @@ -74,6 +83,9 @@ coloredlogs: compel: version: "2.1.1" isRoot: true +conformer: + version: "0.3.2" + isRoot: true contourpy: version: "1.3.2" dependents: @@ -85,6 +97,10 @@ cycler: version: "0.12.1" dependents: - matplotlib +decorator: + version: "5.3.1" + dependents: + - librosa deprecated: version: "1.2.18" isRoot: true @@ -111,6 +127,7 @@ einops: version: "0.8.1" isRoot: true dependents: + - conformer - controlnet-aux - diffsynth - spandrel @@ -207,6 +224,9 @@ humanfriendly: version: "10.0" dependents: - coloredlogs +hyperpyyaml: + version: "1.2.2" + isRoot: true idna: version: "3.4" dependents: @@ -232,6 +252,17 @@ importlib-metadata: dependents: - controlnet-aux - diffusers +importlib-resources: + version: "7.1.0" + dependents: + - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true +jaraco-functools: + version: "4.6.0" + dependents: + - inflect jax: version: "0.6.1" dependents: @@ -246,6 +277,11 @@ jinja2: dependents: - fastapi - torch +joblib: + version: "1.6.0" + dependents: + - librosa + - scikit-learn jsonschema: version: "4.23.0" dependents: @@ -268,7 +304,11 @@ kornia-rs: lazy-loader: version: "0.4" dependents: + - librosa - scikit-image +librosa: + version: "0.11.0" + isRoot: true llvmlite: version: "0.44.0" dependents: @@ -302,20 +342,35 @@ ml-dtypes: - jaxlib modelscope: version: "1.26.0" + isRoot: true dependents: - diffsynth moderngl: version: "5.12.0" isRoot: true +more-itertools: + version: "11.2.1" + dependents: + - inflect + - jaraco-functools + - openai-whisper mpmath: version: "1.3.0" dependents: - sympy +msgpack: + version: "1.2.4" + dependents: + - librosa multidict: version: "6.4.4" dependents: - aiohttp - yarl +narwhals: + version: "2.28.0" + dependents: + - scikit-learn networkx: version: "3.3" dependents: @@ -324,6 +379,8 @@ networkx: numba: version: "0.61.2" dependents: + - librosa + - openai-whisper - pymatting numpy: version: "1.26.3" @@ -336,12 +393,14 @@ numpy: - imageio - jax - jaxlib + - librosa - matplotlib - mediapipe - ml-dtypes - numba - onnx - onnxruntime + - openai-whisper - opencv-contrib-python - opencv-python - opencv-python-headless @@ -353,7 +412,10 @@ numpy: - pyvista - rembg - scikit-image + - scikit-learn - scipy + - soundfile + - soxr - spandrel - tifffile - torchsde @@ -414,12 +476,18 @@ nvidia-nvtx-cu12: version: "12.4.99" dependents: - torch +omegaconf: + version: "2.4.0" + isRoot: true onnx: version: "1.18.0" isRoot: true onnxruntime: version: "1.22.0" isRoot: true +openai-whisper: + version: "20240930" + isRoot: true opencv-contrib-python: version: "4.11.0.86" dependents: @@ -485,6 +553,7 @@ plyfile: pooch: version: "1.8.2" dependents: + - librosa - pyvista - rembg propcache: @@ -532,6 +601,9 @@ pymatting: pymeshfix: version: "0.17.1" isRoot: true +pynini: + version: "2.1.7" + isRoot: true pyparsing: version: "3.2.3" dependents: @@ -563,6 +635,8 @@ pyyaml: - accelerate - gguf - huggingface-hub + - hyperpyyaml + - omegaconf - peft - timm - transformers @@ -577,6 +651,7 @@ regex: dependents: - clip-anytorch - diffusers + - tiktoken - transformers rembg: version: "2.0.66" @@ -588,6 +663,7 @@ requests: - huggingface-hub - modelscope - pooch + - tiktoken - transformers restrictedpython: version: "8.0" @@ -606,6 +682,15 @@ rpds-py: dependents: - jsonschema - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml safetensors: version: "0.8.0" isRoot: true @@ -622,6 +707,10 @@ scikit-image: dependents: - controlnet-aux - rembg +scikit-learn: + version: "1.9.1" + dependents: + - librosa scipy: version: "1.15.3" isRoot: true @@ -629,9 +718,11 @@ scipy: - controlnet-aux - jax - jaxlib + - librosa - pymatting - rembg - scikit-image + - scikit-learn - torchsde scooby: version: "0.10.1" @@ -662,6 +753,15 @@ sounddevice: version: "0.5.2" dependents: - mediapipe +soundfile: + version: "0.14.0" + isRoot: true + dependents: + - librosa +soxr: + version: "1.1.0" + dependents: + - librosa spandrel: version: "0.4.1" isRoot: true @@ -678,10 +778,19 @@ texttable: version: "1.7.0" dependents: - igraph +threadpoolctl: + version: "3.8.0" + dependents: + - scikit-learn tifffile: version: "2025.5.21" dependents: - scikit-image +tiktoken: + version: "0.14.0" + isRoot: true + dependents: + - openai-whisper timm: version: "1.0.15" dependents: @@ -698,10 +807,12 @@ torch: - accelerate - clip-anytorch - compel + - conformer - controlnet-aux - diffsynth - diffusers - kornia + - openai-whisper - peft - spandrel - timm @@ -731,6 +842,7 @@ tqdm: - gguf - huggingface-hub - modelscope + - openai-whisper - peft - rembg - transformers @@ -751,25 +863,33 @@ trimesh: triton: version: "3.0.0" dependents: + - openai-whisper - torch +typeguard: + version: "4.6.0" + dependents: + - inflect typer: version: "0.15.4" dependents: - fastapi-cli typing-extensions: - version: "4.12.2" + version: "4.16.0" dependents: - anyio - fastapi - huggingface-hub + - librosa - onnx - pydantic - pydantic-core - pyvista - referencing - rich-toolkit + - soundfile - spandrel - torch + - typeguard - typer - typing-inspection typing-inspection: @@ -811,6 +931,10 @@ websockets: isRoot: true dependents: - uvicorn +wetextprocessing: + version: "1.0.4" + url: "https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl" + isRoot: true wrapt: version: "1.17.2" dependents: diff --git a/dependencies/linux.lock b/dependencies/linux.lock index 93060aa6..0871084d 100644 --- a/dependencies/linux.lock +++ b/dependencies/linux.lock @@ -29,6 +29,8 @@ attrs==25.3.0 # jsonschema # mediapipe # referencing +audioread==3.1.0 + # via librosa blake3==1.0.5 # via -r dependencies/requirements-linux.txt certifi==2022.12.7 @@ -37,7 +39,9 @@ certifi==2022.12.7 # httpx # requests cffi==1.17.1 - # via sounddevice + # via + # sounddevice + # soundfile charset-normalizer==2.1.1 # via requests click==8.1.8 @@ -47,16 +51,22 @@ click==8.1.8 # uvicorn clip-anytorch==2.6.0 # via -r dependencies/requirements-linux.txt +cloudpickle==3.1.2 + # via joblib coloredlogs==15.0.1 # via onnxruntime compel==2.1.1 # via -r dependencies/requirements-linux.txt +conformer==0.3.2 + # via -r dependencies/requirements-linux.txt contourpy==1.3.2 # via matplotlib controlnet-aux==0.0.7 # via -r dependencies/requirements-linux.txt cycler==0.12.1 # via matplotlib +decorator==5.3.1 + # via librosa deprecated==1.2.18 # via -r dependencies/requirements-linux.txt diffsynth==2.1.12 @@ -74,6 +84,7 @@ easydict==1.13 einops==0.8.1 # via # -r dependencies/requirements-linux.txt + # conformer # controlnet-aux # diffsynth # spandrel @@ -142,6 +153,8 @@ huggingface-hub==0.36.2 # transformers humanfriendly==10.0 # via coloredlogs +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-linux.txt idna==3.4 # via # anyio @@ -161,6 +174,12 @@ importlib-metadata==7.1.0 # via # controlnet-aux # diffusers +importlib-resources==7.1.0 + # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-linux.txt +jaraco-functools==4.6.0 + # via inflect jax==0.6.1 # via mediapipe jaxlib==0.6.1 @@ -171,6 +190,10 @@ jinja2==3.1.4 # via # fastapi # torch +joblib==1.6.0 + # via + # librosa + # scikit-learn jsonschema==4.23.0 # via rembg jsonschema-specifications==2025.4.1 @@ -182,7 +205,11 @@ kornia==0.8.1 kornia-rs==0.1.9 # via kornia lazy-loader==0.4 - # via scikit-image + # via + # librosa + # scikit-image +librosa==0.11.0 + # via -r dependencies/requirements-linux.txt llvmlite==0.44.0 # via numba markdown-it-py==3.0.0 @@ -204,21 +231,35 @@ ml-dtypes==0.5.1 # jax # jaxlib modelscope==1.26.0 - # via diffsynth + # via + # -r dependencies/requirements-linux.txt + # diffsynth moderngl==5.12.0 # via -r dependencies/requirements-linux.txt +more-itertools==11.2.1 + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy +msgpack==1.2.4 + # via librosa multidict==6.4.4 # via # aiohttp # yarl +narwhals==2.28.0 + # via scikit-learn networkx==3.3 # via # scikit-image # torch numba==0.61.2 - # via pymatting + # via + # librosa + # openai-whisper + # pymatting numpy==1.26.3 # via # accelerate @@ -229,12 +270,14 @@ numpy==1.26.3 # imageio # jax # jaxlib + # librosa # matplotlib # mediapipe # ml-dtypes # numba # onnx # onnxruntime + # openai-whisper # opencv-contrib-python # opencv-python # opencv-python-headless @@ -246,7 +289,10 @@ numpy==1.26.3 # pyvista # rembg # scikit-image + # scikit-learn # scipy + # soundfile + # soxr # spandrel # tifffile # torchsde @@ -286,10 +332,14 @@ nvidia-nvjitlink-cu12==12.4.99 # torch nvidia-nvtx-cu12==12.4.99 # via torch +omegaconf==2.4.0 + # via -r dependencies/requirements-linux.txt onnx==1.18.0 # via -r dependencies/requirements-linux.txt onnxruntime==1.22.0 # via -r dependencies/requirements-linux.txt +openai-whisper==20240930 + # via -r dependencies/requirements-linux.txt opencv-contrib-python==4.11.0.86 # via mediapipe opencv-python==4.11.0.86 @@ -338,6 +388,7 @@ plyfile==1.1 # via -r dependencies/requirements-linux.txt pooch==1.8.2 # via + # librosa # pyvista # rembg propcache==0.3.1 @@ -371,6 +422,8 @@ pymatting==1.1.14 # via rembg pymeshfix==0.17.1 # via -r dependencies/requirements-linux.txt +pynini==2.1.7 + # via -r dependencies/requirements-linux.txt pyparsing==3.2.3 # via # compel @@ -395,6 +448,8 @@ pyyaml==6.0.2 # accelerate # gguf # huggingface-hub + # hyperpyyaml + # omegaconf # peft # timm # transformers @@ -407,6 +462,7 @@ regex==2024.11.6 # via # clip-anytorch # diffusers + # tiktoken # transformers rembg==2.0.66 # via -r dependencies/requirements-linux.txt @@ -416,6 +472,7 @@ requests==2.28.1 # huggingface-hub # modelscope # pooch + # tiktoken # transformers restrictedpython==8.0 # via -r dependencies/requirements-linux.txt @@ -429,6 +486,12 @@ rpds-py==0.25.1 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-linux.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-linux.txt @@ -443,15 +506,19 @@ scikit-image==0.25.2 # via # controlnet-aux # rembg +scikit-learn==1.9.1 + # via librosa scipy==1.15.3 # via # -r dependencies/requirements-linux.txt # controlnet-aux # jax # jaxlib + # librosa # pymatting # rembg # scikit-image + # scikit-learn # torchsde scooby==0.10.1 # via pyvista @@ -469,6 +536,12 @@ sniffio==1.3.1 # via anyio sounddevice==0.5.2 # via mediapipe +soundfile==0.14.0 + # via + # -r dependencies/requirements-linux.txt + # librosa +soxr==1.1.0 + # via librosa spandrel==0.4.1 # via -r dependencies/requirements-linux.txt starlette==0.41.3 @@ -479,8 +552,14 @@ sympy==1.13.3 # torch texttable==1.7.0 # via igraph +threadpoolctl==3.8.0 + # via scikit-learn tifffile==2025.5.21 # via scikit-image +tiktoken==0.14.0 + # via + # -r dependencies/requirements-linux.txt + # openai-whisper timm==1.0.15 # via controlnet-aux tokenizers==0.22.2 @@ -493,10 +572,12 @@ torch==2.4.1+cu124 # accelerate # clip-anytorch # compel + # conformer # controlnet-aux # diffsynth # diffusers # kornia + # openai-whisper # peft # spandrel # timm @@ -522,6 +603,7 @@ tqdm==4.66.5 # gguf # huggingface-hub # modelscope + # openai-whisper # peft # rembg # transformers @@ -536,22 +618,29 @@ transformers==4.57.6 trimesh==4.6.10 # via -r dependencies/requirements-linux.txt triton==3.0.0 - # via torch + # via + # openai-whisper + # torch +typeguard==4.6.0 + # via inflect typer==0.15.4 # via fastapi-cli -typing-extensions==4.12.2 +typing-extensions==4.16.0 # via # anyio # fastapi # huggingface-hub + # librosa # onnx # pydantic # pydantic-core # pyvista # referencing # rich-toolkit + # soundfile # spandrel # torch + # typeguard # typer # typing-inspection typing-inspection==0.4.1 @@ -580,6 +669,8 @@ websockets==15.0.1 # via # -r dependencies/requirements-linux.txt # uvicorn +wetextprocessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl + # via -r dependencies/requirements-linux.txt wrapt==1.17.2 # via deprecated xatlas==0.0.10 diff --git a/dependencies/macosx-dependencies.yaml b/dependencies/macosx-dependencies.yaml index 8e03b855..211ac733 100644 --- a/dependencies/macosx-dependencies.yaml +++ b/dependencies/macosx-dependencies.yaml @@ -44,6 +44,10 @@ attrs: - jsonschema - mediapipe - referencing +audioread: + version: "3.1.0" + dependents: + - librosa blake3: version: "1.0.4" isRoot: true @@ -57,6 +61,7 @@ cffi: version: "1.17.1" dependents: - sounddevice + - soundfile charset-normalizer: version: "3.4.1" dependents: @@ -70,6 +75,10 @@ click: clip-anytorch: version: "2.6.0" isRoot: true +cloudpickle: + version: "3.1.2" + dependents: + - joblib coloredlogs: version: "15.0.1" dependents: @@ -77,6 +86,9 @@ coloredlogs: compel: version: "2.0.3" isRoot: true +conformer: + version: "0.3.2" + isRoot: true contourpy: version: "1.3.1" dependents: @@ -88,6 +100,10 @@ cycler: version: "0.12.1" dependents: - matplotlib +decorator: + version: "5.3.1" + dependents: + - librosa deprecated: version: "1.2.18" isRoot: true @@ -111,6 +127,7 @@ einops: version: "0.8.1" isRoot: true dependents: + - conformer - controlnet-aux - spandrel email-validator: @@ -204,6 +221,9 @@ humanfriendly: version: "10.0" dependents: - coloredlogs +hyperpyyaml: + version: "1.2.2" + isRoot: true idna: version: "3.10" dependents: @@ -229,6 +249,17 @@ importlib-metadata: dependents: - controlnet-aux - diffusers +importlib-resources: + version: "7.1.0" + dependents: + - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true +jaraco-functools: + version: "4.6.0" + dependents: + - inflect jax: version: "0.5.3" dependents: @@ -243,6 +274,11 @@ jinja2: dependents: - fastapi - torch +joblib: + version: "1.6.0" + dependents: + - librosa + - scikit-learn jsonschema: version: "4.23.0" dependents: @@ -265,7 +301,11 @@ kornia-rs: lazy-loader: version: "0.4" dependents: + - librosa - scikit-image +librosa: + version: "0.11.0" + isRoot: true llvmlite: version: "0.44.0" dependents: @@ -303,15 +343,29 @@ modelscope: moderngl: version: "5.12.0" isRoot: true +more-itertools: + version: "11.2.1" + dependents: + - inflect + - jaraco-functools + - openai-whisper mpmath: version: "1.3.0" dependents: - sympy +msgpack: + version: "1.2.4" + dependents: + - librosa multidict: version: "6.2.0" dependents: - aiohttp - yarl +narwhals: + version: "2.28.0" + dependents: + - scikit-learn networkx: version: "3.4.2" dependents: @@ -320,6 +374,8 @@ networkx: numba: version: "0.61.0" dependents: + - librosa + - openai-whisper - pymatting numpy: version: "1.26.4" @@ -332,12 +388,14 @@ numpy: - imageio - jax - jaxlib + - librosa - matplotlib - mediapipe - ml-dtypes - numba - onnx - onnxruntime + - openai-whisper - opencv-contrib-python - opencv-python - opencv-python-headless @@ -349,7 +407,10 @@ numpy: - pyvista - rembg - scikit-image + - scikit-learn - scipy + - soundfile + - soxr - spandrel - tifffile - torchsde @@ -365,6 +426,9 @@ onnx: onnxruntime: version: "1.21.0" isRoot: true +openai-whisper: + version: "20240930" + isRoot: true opencv-contrib-python: version: "4.11.0.86" dependents: @@ -427,6 +491,7 @@ plyfile: pooch: version: "1.8.2" dependents: + - librosa - pyvista - rembg propcache: @@ -474,6 +539,9 @@ pymatting: pymeshfix: version: "0.17.0" isRoot: true +pynini: + version: "2.1.7" + isRoot: true pyparsing: version: "3.2.3" dependents: @@ -505,6 +573,7 @@ pyyaml: - accelerate - gguf - huggingface-hub + - hyperpyyaml - omegaconf - peft - transformers @@ -550,6 +619,15 @@ rpds-py: dependents: - jsonschema - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml safetensors: version: "0.8.0" isRoot: true @@ -565,6 +643,10 @@ scikit-image: dependents: - controlnet-aux - rembg +scikit-learn: + version: "1.9.1" + dependents: + - librosa scipy: version: "1.15.2" isRoot: true @@ -572,9 +654,11 @@ scipy: - controlnet-aux - jax - jaxlib + - librosa - pymatting - rembg - scikit-image + - scikit-learn - torchsde scooby: version: "0.10.0" @@ -606,6 +690,15 @@ sounddevice: version: "0.5.1" dependents: - mediapipe +soundfile: + version: "0.14.0" + isRoot: true + dependents: + - librosa +soxr: + version: "1.1.0" + dependents: + - librosa spandrel: version: "0.4.1" isRoot: true @@ -622,6 +715,10 @@ texttable: version: "1.7.0" dependents: - igraph +threadpoolctl: + version: "3.8.0" + dependents: + - scikit-learn tifffile: version: "2025.3.13" dependents: @@ -629,6 +726,8 @@ tifffile: tiktoken: version: "0.9.0" isRoot: true + dependents: + - openai-whisper timm: version: "0.6.7" dependents: @@ -645,9 +744,11 @@ torch: - accelerate - clip-anytorch - compel + - conformer - controlnet-aux - diffusers - kornia + - openai-whisper - peft - spandrel - timm @@ -675,6 +776,7 @@ tqdm: - gguf - huggingface-hub - modelscope + - openai-whisper - peft - rembg - transformers @@ -695,23 +797,30 @@ transformers-stream-generator: trimesh: version: "4.6.6" isRoot: true +typeguard: + version: "4.6.0" + dependents: + - inflect typer: version: "0.15.2" dependents: - fastapi-cli typing-extensions: - version: "4.13.0" + version: "4.16.0" dependents: - anyio - fastapi - huggingface-hub + - librosa - pydantic - pydantic-core - pyvista - referencing - rich-toolkit + - soundfile - spandrel - torch + - typeguard - typer - typing-inspection typing-inspection: @@ -752,6 +861,10 @@ websockets: isRoot: true dependents: - uvicorn +wetextprocessing: + version: "1.0.4" + url: "https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl" + isRoot: true wrapt: version: "1.17.2" dependents: diff --git a/dependencies/macosx.lock b/dependencies/macosx.lock index 59a0dc3e..ad8e98bc 100644 --- a/dependencies/macosx.lock +++ b/dependencies/macosx.lock @@ -30,6 +30,8 @@ attrs==25.3.0 # jsonschema # mediapipe # referencing +audioread==3.1.0 + # via librosa blake3==1.0.4 # via -r dependencies/requirements-macosx.txt certifi==2025.1.31 @@ -38,7 +40,9 @@ certifi==2025.1.31 # httpx # requests cffi==1.17.1 - # via sounddevice + # via + # sounddevice + # soundfile charset-normalizer==3.4.1 # via requests click==8.1.8 @@ -48,16 +52,22 @@ click==8.1.8 # uvicorn clip-anytorch==2.6.0 # via -r dependencies/requirements-macosx.txt +cloudpickle==3.1.2 + # via joblib coloredlogs==15.0.1 # via onnxruntime compel==2.0.3 # via -r dependencies/requirements-macosx.txt +conformer==0.3.2 + # via -r dependencies/requirements-macosx.txt contourpy==1.3.1 # via matplotlib controlnet-aux==0.0.7 # via -r dependencies/requirements-macosx.txt cycler==0.12.1 # via matplotlib +decorator==5.3.1 + # via librosa deprecated==1.2.18 # via -r dependencies/requirements-macosx.txt diffusers==0.38.0 @@ -73,6 +83,7 @@ easydict==1.13 einops==0.8.1 # via # -r dependencies/requirements-macosx.txt + # conformer # controlnet-aux # spandrel email-validator==2.2.0 @@ -138,6 +149,8 @@ huggingface-hub==0.36.2 # transformers humanfriendly==10.0 # via coloredlogs +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-macosx.txt idna==3.10 # via # anyio @@ -157,6 +170,12 @@ importlib-metadata==8.6.1 # via # controlnet-aux # diffusers +importlib-resources==7.1.0 + # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-macosx.txt +jaraco-functools==4.6.0 + # via inflect jax==0.5.3 # via mediapipe jaxlib==0.5.3 @@ -167,6 +186,10 @@ jinja2==3.1.6 # via # fastapi # torch +joblib==1.6.0 + # via + # librosa + # scikit-learn jsonschema==4.23.0 # via rembg jsonschema-specifications==2024.10.1 @@ -178,7 +201,11 @@ kornia==0.8.0 kornia-rs==0.1.8 # via kornia lazy-loader==0.4 - # via scikit-image + # via + # librosa + # scikit-image +librosa==0.11.0 + # via -r dependencies/requirements-macosx.txt llvmlite==0.44.0 # via numba markdown-it-py==3.0.0 @@ -203,18 +230,30 @@ modelscope==1.26.0 # via -r dependencies/requirements-macosx.txt moderngl==5.12.0 # via -r dependencies/requirements-macosx.txt +more-itertools==11.2.1 + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy +msgpack==1.2.4 + # via librosa multidict==6.2.0 # via # aiohttp # yarl +narwhals==2.28.0 + # via scikit-learn networkx==3.4.2 # via # scikit-image # torch numba==0.61.0 - # via pymatting + # via + # librosa + # openai-whisper + # pymatting numpy==1.26.4 # via # accelerate @@ -225,12 +264,14 @@ numpy==1.26.4 # imageio # jax # jaxlib + # librosa # matplotlib # mediapipe # ml-dtypes # numba # onnx # onnxruntime + # openai-whisper # opencv-contrib-python # opencv-python # opencv-python-headless @@ -242,7 +283,10 @@ numpy==1.26.4 # pyvista # rembg # scikit-image + # scikit-learn # scipy + # soundfile + # soxr # spandrel # tifffile # torchsde @@ -255,6 +299,8 @@ onnx==1.17.0 # via -r dependencies/requirements-macosx.txt onnxruntime==1.21.0 # via -r dependencies/requirements-macosx.txt +openai-whisper==20240930 + # via -r dependencies/requirements-macosx.txt opencv-contrib-python==4.11.0.86 # via mediapipe opencv-python==4.11.0.86 @@ -301,6 +347,7 @@ plyfile==1.1 # via -r dependencies/requirements-macosx.txt pooch==1.8.2 # via + # librosa # pyvista # rembg propcache==0.3.1 @@ -334,6 +381,8 @@ pymatting==1.1.13 # via rembg pymeshfix==0.17.0 # via -r dependencies/requirements-macosx.txt +pynini==2.1.7 + # via -r dependencies/requirements-macosx.txt pyparsing==3.2.3 # via # compel @@ -358,6 +407,7 @@ pyyaml==6.0.2 # accelerate # gguf # huggingface-hub + # hyperpyyaml # omegaconf # peft # transformers @@ -394,6 +444,12 @@ rpds-py==0.24.0 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-macosx.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-macosx.txt @@ -407,15 +463,19 @@ scikit-image==0.25.2 # -r dependencies/requirements-macosx.txt # controlnet-aux # rembg +scikit-learn==1.9.1 + # via librosa scipy==1.15.2 # via # -r dependencies/requirements-macosx.txt # controlnet-aux # jax # jaxlib + # librosa # pymatting # rembg # scikit-image + # scikit-learn # torchsde scooby==0.10.0 # via pyvista @@ -434,6 +494,12 @@ sniffio==1.3.1 # via anyio sounddevice==0.5.1 # via mediapipe +soundfile==0.14.0 + # via + # -r dependencies/requirements-macosx.txt + # librosa +soxr==1.1.0 + # via librosa spandrel==0.4.1 # via -r dependencies/requirements-macosx.txt starlette==0.46.1 @@ -444,10 +510,14 @@ sympy==1.13.3 # torch texttable==1.7.0 # via igraph +threadpoolctl==3.8.0 + # via scikit-learn tifffile==2025.3.13 # via scikit-image tiktoken==0.9.0 - # via -r dependencies/requirements-macosx.txt + # via + # -r dependencies/requirements-macosx.txt + # openai-whisper timm==0.6.7 # via controlnet-aux tokenizers==0.22.2 @@ -460,9 +530,11 @@ torch==2.4.1 # accelerate # clip-anytorch # compel + # conformer # controlnet-aux # diffusers # kornia + # openai-whisper # peft # spandrel # timm @@ -486,6 +558,7 @@ tqdm==4.67.1 # gguf # huggingface-hub # modelscope + # openai-whisper # peft # rembg # transformers @@ -501,20 +574,25 @@ transformers-stream-generator==0.0.4 # via -r dependencies/requirements-macosx.txt trimesh==4.6.6 # via -r dependencies/requirements-macosx.txt +typeguard==4.6.0 + # via inflect typer==0.15.2 # via fastapi-cli -typing-extensions==4.13.0 +typing-extensions==4.16.0 # via # anyio # fastapi # huggingface-hub + # librosa # pydantic # pydantic-core # pyvista # referencing # rich-toolkit + # soundfile # spandrel # torch + # typeguard # typer # typing-inspection typing-inspection==0.4.0 @@ -541,6 +619,8 @@ websockets==15.0.1 # via # -r dependencies/requirements-macosx.txt # uvicorn +wetextprocessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl + # via -r dependencies/requirements-macosx.txt wrapt==1.17.2 # via deprecated yarl==1.18.3 diff --git a/dependencies/requirements-linux-amdgpu.txt b/dependencies/requirements-linux-amdgpu.txt index be798486..90c2f334 100644 --- a/dependencies/requirements-linux-amdgpu.txt +++ b/dependencies/requirements-linux-amdgpu.txt @@ -7,6 +7,7 @@ aioshutil>=1.5 blake3>=1.0.4 clip-anytorch>=2.6.0 compel>=2.0.3 +conformer>=0.3.2 controlnet-aux>=0.0.7 deprecated>=1.2.18 diffsynth>=1.1.7 @@ -16,13 +17,19 @@ einops>=0.8.0 fastapi[standard]>=0.115.6 gguf>=0.14.0 huggingface_hub[hf-transfer,hf_xet]>=0.31 +HyperPyYAML==1.2.2 igraph>=0.11.8 +inflect>=7.3,<8 kornia>=0.8.0 +librosa==0.11.0 matplotlib>=3.10.1 mediapipe>=0.10.21 +modelscope==1.26.0 moderngl>=5.12.0 +omegaconf>=2.3,<3 onnx>=1.17.0 onnxruntime>=1.21.0 +openai-whisper==20240930 peft==0.17.1 picklescan>=0.0.24 pillow>=11.0.0 @@ -30,6 +37,7 @@ plyfile>=1.1 pydantic>=2.10.4 pydantic-settings>=2.7.0 pymeshfix>=0.17.0 +pynini>=2.1.6,<2.2 pyvista>=0.44.2 pyyaml>=6.0.2 rembg>=2.0.65 @@ -38,9 +46,12 @@ rocm @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm-7.2.1.tar.gz rocm_sdk_core @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm_sdk_core-7.2.1-py3-none-win_amd64.whl rocm_sdk_devel @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm_sdk_devel-7.2.1-py3-none-win_amd64.whl rocm_sdk_libraries_custom @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm_sdk_libraries_custom-7.2.1-py3-none-win_amd64.whl +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scipy>=1.15.2 +soundfile>=0.12.1 spandrel>=0.4.1 +tiktoken>=0.9.0 tokenizers>=0.20.0 torch @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/torch-2.9.1%2Brocm7.2.1-cp312-cp312-win_amd64.whl torchaudio @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/torchaudio-2.9.1%2Brocm7.2.1-cp312-cp312-win_amd64.whl @@ -50,4 +61,5 @@ transformers>=4.51.3,<5 trimesh>=4.6.5 virtualenv>=20.28.0 websockets>=14.1 +WeTextProcessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl xatlas>=0.0.9 diff --git a/dependencies/requirements-linux-arm64.txt b/dependencies/requirements-linux-arm64.txt index 755a7bcf..a6d24711 100644 --- a/dependencies/requirements-linux-arm64.txt +++ b/dependencies/requirements-linux-arm64.txt @@ -7,6 +7,7 @@ aioshutil>=1.5 blake3>=1.0.4 clip-anytorch>=2.6.0 compel>=2.0.3 +conformer>=0.3.2 controlnet-aux>=0.0.7 deprecated>=1.2.18 diffsynth>=1.1.7 @@ -16,13 +17,19 @@ einops>=0.8.0 fastapi[standard]>=0.115.6 gguf>=0.14.0 huggingface_hub[hf-transfer,hf_xet]>=0.31 +HyperPyYAML==1.2.2 igraph>=0.11.8 +inflect>=7.3,<8 kornia>=0.8.0 +librosa==0.11.0 matplotlib>=3.10.1 mediapipe>=0.10.21 +modelscope==1.26.0 moderngl>=5.12.0 +omegaconf>=2.3,<3 onnx>=1.17.0 onnxruntime>=1.21.0 +openai-whisper==20240930 peft==0.17.1 picklescan>=0.0.24 pillow>=11.0.0 @@ -30,14 +37,18 @@ plyfile>=1.1 pydantic>=2.10.4 pydantic-settings>=2.7.0 pymeshfix>=0.17.0 +pynini>=2.1.6,<2.2 pyvista>=0.44.2 pyyaml>=6.0.2 rembg>=2.0.65 RestrictedPython>=8.0 +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scipy>=1.15.2 sentencepiece==0.2.0 +soundfile>=0.12.1 spandrel>=0.4.1 +tiktoken>=0.9.0 tokenizers>=0.20.0 torch==2.8.0 torchaudio==2.8.0 @@ -47,4 +58,5 @@ transformers>=4.51.3,<5 trimesh>=4.6.5 virtualenv>=20.28.0 websockets>=14.1 +WeTextProcessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl --extra-index-url https://download.pytorch.org/whl/cpu diff --git a/dependencies/requirements-linux.txt b/dependencies/requirements-linux.txt index c4244a5f..03be5484 100644 --- a/dependencies/requirements-linux.txt +++ b/dependencies/requirements-linux.txt @@ -7,6 +7,7 @@ aioshutil>=1.5 blake3>=1.0.4 clip-anytorch>=2.6.0 compel>=2.0.3 +conformer>=0.3.2 controlnet-aux>=0.0.7 deprecated>=1.2.18 diffsynth>=1.1.7 @@ -16,13 +17,19 @@ einops>=0.8.0 fastapi[standard]>=0.115.6 gguf>=0.14.0 huggingface_hub[hf-transfer,hf_xet]>=0.31 +HyperPyYAML==1.2.2 igraph>=0.11.8 +inflect>=7.3,<8 kornia>=0.8.0 +librosa==0.11.0 matplotlib>=3.10.1 mediapipe>=0.10.21 +modelscope==1.26.0 moderngl>=5.12.0 +omegaconf>=2.3,<3 onnx>=1.17.0 onnxruntime>=1.21.0 +openai-whisper==20240930 peft==0.17.1 picklescan>=0.0.24 pillow>=11.0.0 @@ -30,13 +37,17 @@ plyfile>=1.1 pydantic>=2.10.4 pydantic-settings>=2.7.0 pymeshfix>=0.17.0 +pynini>=2.1.6,<2.2 pyvista>=0.44.2 pyyaml>=6.0.2 rembg>=2.0.65 RestrictedPython>=8.0 +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scipy>=1.15.2 +soundfile>=0.12.1 spandrel>=0.4.1 +tiktoken>=0.9.0 tokenizers>=0.20.0 torch==2.4.1+cu124 torchaudio==2.4.1+cu124 @@ -46,6 +57,7 @@ transformers>=4.51.3,<5 trimesh>=4.6.5 virtualenv>=20.28.0 websockets>=14.1 +WeTextProcessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl xatlas>=0.0.9 xformers==0.0.28.post1 --extra-index-url https://download.pytorch.org/whl/cu124 diff --git a/dependencies/requirements-macosx-x86_64.txt b/dependencies/requirements-macosx-x86_64.txt index 184395e5..3d46434e 100644 --- a/dependencies/requirements-macosx-x86_64.txt +++ b/dependencies/requirements-macosx-x86_64.txt @@ -7,6 +7,7 @@ aioshutil>=1.5 blake3>=1.0.4 clip-anytorch>=2.6.0 compel>=2.0.3 +conformer>=0.3.2 controlnet-aux>=0.0.7 deprecated>=1.2.18 diffusers[torch]>=0.37.0,<0.41 @@ -16,9 +17,12 @@ fastapi[standard]>=0.115.6 ftfy>=6.3 gguf>=0.14.0 huggingface_hub[hf-transfer,hf_xet]>=0.31 +HyperPyYAML==1.2.2 igraph>=0.11.8 imageio[ffmpeg]>=2.37.0 +inflect>=7.3,<8 kornia>=0.8.0 +librosa==0.11.0 llvmlite==0.45.1 matplotlib>=3.10.1 mediapipe>=0.10.21 @@ -29,6 +33,7 @@ numpy<2 omegaconf==2.3.0 onnx>=1.17.0 onnxruntime>=1.21.0 +openai-whisper==20240930 pandas>=2.2.3 peft==0.17.1 picklescan>=0.0.24 @@ -37,10 +42,12 @@ plyfile>=1.1 pydantic>=2.10.4 pydantic-settings>=2.7.0 pymeshfix>=0.17.0 +pynini>=2.1.6,<2.2 pyvista>=0.44.2 pyyaml>=6.0.2 rembg>=2.0.65 RestrictedPython>=8.0 +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scikit-image>=0.25.2 scipy>=1.15.2 @@ -59,5 +66,6 @@ transformers_stream_generator==0.0.4 trimesh>=4.6.5 virtualenv>=20.28.0 websockets>=14.1 +WeTextProcessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl xatlas>=0.0.9 --extra-index-url https://download.pytorch.org/whl/cpu diff --git a/dependencies/requirements-macosx.txt b/dependencies/requirements-macosx.txt index 73717e40..43b7ef92 100644 --- a/dependencies/requirements-macosx.txt +++ b/dependencies/requirements-macosx.txt @@ -7,6 +7,7 @@ aioshutil>=1.5 blake3>=1.0.4 clip-anytorch>=2.6.0 compel>=2.0.3 +conformer>=0.3.2 controlnet-aux>=0.0.7 deprecated>=1.2.18 diffusers[torch]>=0.37.0,<0.41 @@ -16,9 +17,12 @@ fastapi[standard]>=0.115.6 ftfy>=6.3 gguf>=0.14.0 huggingface_hub[hf-transfer,hf_xet]>=0.31 +HyperPyYAML==1.2.2 igraph>=0.11.8 imageio[ffmpeg]>=2.37.0 +inflect>=7.3,<8 kornia>=0.8.0 +librosa==0.11.0 matplotlib>=3.10.1 mediapipe>=0.10.21 modelscope==1.26.0 @@ -26,6 +30,7 @@ moderngl>=5.12.0 omegaconf==2.3.0 onnx>=1.17.0 onnxruntime>=1.21.0 +openai-whisper==20240930 pandas>=2.2.3 peft==0.17.1 picklescan>=0.0.24 @@ -34,14 +39,17 @@ plyfile>=1.1 pydantic>=2.10.4 pydantic-settings>=2.7.0 pymeshfix>=0.17.0 +pynini>=2.1.6,<2.2 pyvista>=0.44.2 pyyaml>=6.0.2 rembg>=2.0.65 RestrictedPython>=8.0 +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scikit-image>=0.25.2 scipy>=1.15.2 sentencepiece==0.2.0 +soundfile>=0.12.1 spandrel>=0.4.1 tiktoken>=0.9.0 tokenizers>=0.20.0 @@ -54,3 +62,4 @@ transformers_stream_generator==0.0.4 trimesh>=4.6.5 virtualenv>=20.28.0 websockets>=14.1 +WeTextProcessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl diff --git a/dependencies/requirements-windows-amdgpu.txt b/dependencies/requirements-windows-amdgpu.txt index f3995572..b0f22113 100644 --- a/dependencies/requirements-windows-amdgpu.txt +++ b/dependencies/requirements-windows-amdgpu.txt @@ -22,8 +22,10 @@ ftfy>=6.3 gguf>=0.10.0 huggingface_hub[hf-transfer,hf_xet]>=0.36.2,<1 hydra-core==1.3.2 +HyperPyYAML==1.2.2 igraph>=0.11.8 imageio[ffmpeg]>=2.37.0 +inflect>=7.3,<8 kaolin @ https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-2.4.0_cu124/kaolin-0.17.0-cp312-cp312-win_amd64.whl kornia>=0.8.0 librosa==0.11.0 @@ -35,6 +37,7 @@ natten @ https://github.com/sunxfancy/SSUI/releases/download/rocm-wheels-2026102 networkx==3.3 nvdiffrast @ https://github.com/sunxfancy/SSUI/releases/download/rocm-wheels-20261027/nvdiffrast-0.4.0%2Bssui.rocm72-cp312-cp312-win_amd64.whl#sha256=5ead6ef29217d88d20a24cacacf05da66f810edc0b03b9567c08f1f30e8613db o-voxel @ https://github.com/sunxfancy/o-voxel.rocm72/releases/download/v0.0.1%2Brocm72/o_voxel-0.0.1%2Brocm72-cp312-cp312-win_amd64.whl +omegaconf>=2.3,<3 onnx>=1.17.0 onnxruntime>=1.21.0 openai-whisper==20240930 @@ -57,6 +60,7 @@ rocm @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm-7.2.1.tar.gz rocm_sdk_core @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm_sdk_core-7.2.1-py3-none-win_amd64.whl rocm_sdk_devel @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm_sdk_devel-7.2.1-py3-none-win_amd64.whl rocm_sdk_libraries_custom @ https://repo.radeon.com/rocm/windows/rocm-rel-7.2.1/rocm_sdk_libraries_custom-7.2.1-py3-none-win_amd64.whl +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.8.0,<0.9 scipy>=1.15.2 scipy>=1.15 diff --git a/dependencies/requirements-windows-arm64-cpu.txt b/dependencies/requirements-windows-arm64-cpu.txt index b881896b..5389cc93 100644 --- a/dependencies/requirements-windows-arm64-cpu.txt +++ b/dependencies/requirements-windows-arm64-cpu.txt @@ -21,9 +21,11 @@ glcontext==3.0.0+ssui.cpu.r1.bf1816f253e6a hf-transfer==0.1.9+ssui.cpu.r1.b0a21e114de54 huggingface_hub[hf-transfer,hf_xet]>=0.36.2,<1 hydra-core==1.3.2 +HyperPyYAML==1.2.2 igraph>=0.11.8 imageio[ffmpeg]>=2.37.0 imageio-ffmpeg==0.6.0+ssui.cpu.r2.b06fa0f77c82a +inflect>=7.3,<8 kornia>=0.8.0 librosa==0.11.0 llvmlite==0.50.0 @@ -33,6 +35,7 @@ modelscope==1.26.0 moderngl==5.12.0+ssui.cpu.r1.be987004179e0 networkx==3.3 numba==0.68.0 +omegaconf>=2.3,<3 onnx>=1.17.0 onnxruntime>=1.21.0 openai-whisper==20240930+ssui.cpu.r1.b7b7485ae8627 @@ -51,6 +54,7 @@ pyworld==0.3.4+ssui.cpu.r1.b83bf5208b453 pyyaml>=6.0.2 rembg>=2.0.65 RestrictedPython>=8.0 +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scipy>=1.15.2 scipy>=1.15 diff --git a/dependencies/requirements-windows.txt b/dependencies/requirements-windows.txt index 9d751bab..024214c0 100644 --- a/dependencies/requirements-windows.txt +++ b/dependencies/requirements-windows.txt @@ -21,8 +21,10 @@ ftfy>=6.3 gguf>=0.10.0 huggingface_hub[hf-transfer,hf_xet]>=0.36.2,<1 hydra-core==1.3.2 +HyperPyYAML==1.2.2 igraph>=0.11.8 imageio[ffmpeg]>=2.37.0 +inflect>=7.3,<8 kaolin @ https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-2.4.0_cu124/kaolin-0.17.0-cp312-cp312-win_amd64.whl kornia>=0.8.0 librosa==0.11.0 @@ -32,6 +34,7 @@ modelscope==1.26.0 moderngl>=5.12.0 networkx==3.3 nvdiffrast @ https://github.com/sunxfancy/nvdiffrast/releases/download/v0.3.3.1/nvdiffrast-0.3.3.1-cp312-cp312-win_amd64.whl +omegaconf>=2.3,<3 onnx>=1.17.0 onnxruntime>=1.21.0 openai-whisper==20240930 @@ -50,6 +53,7 @@ pyworld==0.3.4 pyyaml>=6.0.2 rembg>=2.0.65 RestrictedPython>=8.0 +ruamel.yaml>=0.17.28,<0.19 safetensors>=0.4.5 scipy>=1.15.2 scipy>=1.15 diff --git a/dependencies/windows-amdgpu-dependencies.yaml b/dependencies/windows-amdgpu-dependencies.yaml index 3b2fa7eb..84da23a7 100644 --- a/dependencies/windows-amdgpu-dependencies.yaml +++ b/dependencies/windows-amdgpu-dependencies.yaml @@ -376,6 +376,9 @@ huggingface-hub: hydra-core: version: "1.3.2" isRoot: true +hyperpyyaml: + version: "1.2.2" + isRoot: true idna: version: "3.4" dependents: @@ -406,6 +409,9 @@ importlib-resources: version: "7.1.0" dependents: - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true ipycanvas: version: "0.14.3" dependents: @@ -441,6 +447,10 @@ itsdangerous: version: "2.2.0" dependents: - flask +jaraco-functools: + version: "4.6.0" + dependents: + - inflect jedi: version: "0.20.0" dependents: @@ -616,6 +626,8 @@ moderngl: more-itertools: version: "11.1.0" dependents: + - inflect + - jaraco-functools - openai-whisper mpmath: version: "1.3.0" @@ -740,6 +752,7 @@ o-voxel: isRoot: true omegaconf: version: "2.3.1" + isRoot: true dependents: - hydra-core onnx: @@ -979,6 +992,7 @@ pyyaml: - accelerate - gguf - huggingface-hub + - hyperpyyaml - jupyter-events - omegaconf - peft @@ -1083,6 +1097,15 @@ rpds-py: dependents: - jsonschema - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml safetensors: version: "0.8.0" isRoot: true @@ -1324,6 +1347,10 @@ triton-windows: isRoot: true dependents: - flex-gemm +typeguard: + version: "4.6.0" + dependents: + - inflect typer: version: "0.25.1" dependents: @@ -1349,6 +1376,7 @@ typing-extensions: - spandrel - starlette - torch + - typeguard - typing-inspect - typing-inspection typing-inspect: diff --git a/dependencies/windows-amdgpu.lock b/dependencies/windows-amdgpu.lock index b0b43faa..872ecbed 100644 --- a/dependencies/windows-amdgpu.lock +++ b/dependencies/windows-amdgpu.lock @@ -252,6 +252,8 @@ huggingface-hub==0.36.2 # transformers hydra-core==1.3.2 # via -r dependencies/requirements-windows-amdgpu.txt +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-windows-amdgpu.txt idna==3.4 # via # anyio @@ -274,6 +276,8 @@ importlib-metadata==7.1.0 # diffusers importlib-resources==7.1.0 # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-windows-amdgpu.txt ipycanvas==0.14.3 # via kaolin ipyevents==2.0.4 @@ -295,6 +299,8 @@ isoduration==20.11.0 # via jsonschema itsdangerous==2.2.0 # via flask +jaraco-functools==4.6.0 + # via inflect jedi==0.20.0 # via ipython jinja2==3.1.6 @@ -413,7 +419,10 @@ modelscope==1.26.0 moderngl==5.12.0 # via -r dependencies/requirements-windows-amdgpu.txt more-itertools==11.1.0 - # via openai-whisper + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy msgpack==1.2.1 @@ -507,7 +516,9 @@ nvdiffrast @ https://github.com/sunxfancy/SSUI/releases/download/rocm-wheels-202 o-voxel @ https://github.com/sunxfancy/o-voxel.rocm72/releases/download/v0.0.1%2Brocm72/o_voxel-0.0.1%2Brocm72-cp312-cp312-win_amd64.whl # via -r dependencies/requirements-windows-amdgpu.txt omegaconf==2.3.1 - # via hydra-core + # via + # -r dependencies/requirements-windows-amdgpu.txt + # hydra-core onnx==1.22.0 # via -r dependencies/requirements-windows-amdgpu.txt onnxruntime==1.27.0 @@ -683,6 +694,7 @@ pyyaml==6.0.3 # accelerate # gguf # huggingface-hub + # hyperpyyaml # jupyter-events # omegaconf # peft @@ -762,6 +774,12 @@ rpds-py==2026.5.1 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-windows-amdgpu.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-windows-amdgpu.txt @@ -940,6 +958,8 @@ triton-windows==3.1.0.post17 # via # -r dependencies/requirements-windows-amdgpu.txt # flex-gemm +typeguard==4.6.0 + # via inflect typer==0.25.1 # via # fastapi-cli @@ -963,6 +983,7 @@ typing-extensions==4.15.0 # spandrel # starlette # torch + # typeguard # typing-inspect # typing-inspection typing-inspect==0.9.0 diff --git a/dependencies/windows-arm64-cpu-dependencies.yaml b/dependencies/windows-arm64-cpu-dependencies.yaml index 0732e608..c09a2693 100644 --- a/dependencies/windows-arm64-cpu-dependencies.yaml +++ b/dependencies/windows-arm64-cpu-dependencies.yaml @@ -329,6 +329,9 @@ huggingface-hub: hydra-core: version: "1.3.2" isRoot: true +hyperpyyaml: + version: "1.2.2" + isRoot: true idna: version: "3.20" dependents: @@ -361,6 +364,9 @@ importlib-resources: version: "7.1.0" dependents: - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true ipykernel: version: "7.4.0" dependents: @@ -377,6 +383,10 @@ isoduration: version: "20.11.0" dependents: - jsonschema +jaraco-functools: + version: "4.6.0" + dependents: + - inflect jedi: version: "0.20.1" dependents: @@ -542,6 +552,8 @@ moderngl: more-itertools: version: "11.2.1" dependents: + - inflect + - jaraco-functools - openai-whisper mpmath: version: "1.3.0" @@ -638,6 +650,7 @@ numpy: - trimesh omegaconf: version: "2.3.1" + isRoot: true dependents: - hydra-core onnx: @@ -882,6 +895,7 @@ pyyaml: - accelerate - gguf - huggingface-hub + - hyperpyyaml - jupyter-events - omegaconf - peft @@ -959,6 +973,11 @@ rpds-py: dependents: - jsonschema - referencing +ruamel-yaml: + version: "0.17.33" + isRoot: true + dependents: + - hyperpyyaml safetensors: version: "0.8.0" isRoot: true @@ -1176,6 +1195,10 @@ truststore: dependents: - httpcore2 - httpx2 +typeguard: + version: "4.6.0" + dependents: + - inflect typer: version: "0.27.3" dependents: @@ -1209,6 +1232,7 @@ typing-extensions: - spandrel - starlette - torch + - typeguard - typing-inspection typing-inspection: version: "0.4.4" diff --git a/dependencies/windows-arm64-cpu.lock b/dependencies/windows-arm64-cpu.lock index 98b69a30..781c3bfd 100644 --- a/dependencies/windows-arm64-cpu.lock +++ b/dependencies/windows-arm64-cpu.lock @@ -219,6 +219,8 @@ huggingface-hub==0.36.2 # transformers hydra-core==1.3.2 # via -r dependencies/requirements-windows-arm64-cpu.txt +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-windows-arm64-cpu.txt idna==3.20 # via # anyio @@ -244,6 +246,8 @@ importlib-metadata==9.1.0 # diffusers importlib-resources==7.1.0 # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-windows-arm64-cpu.txt ipykernel==7.4.0 # via jupyterlab ipython==9.17.1 @@ -252,6 +256,8 @@ ipython-pygments-lexers==1.1.1 # via ipython isoduration==20.11.0 # via jsonschema +jaraco-functools==4.6.0 + # via inflect jedi==0.20.1 # via ipython jinja2==3.1.6 @@ -365,7 +371,10 @@ modelscope==1.26.0 moderngl==5.12.0+ssui.cpu.r1.be987004179e0 # via -r dependencies/requirements-windows-arm64-cpu.txt more-itertools==11.2.1 - # via openai-whisper + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy msgpack==1.2.3 @@ -440,7 +449,9 @@ numpy==2.5.4 # transformers # trimesh omegaconf==2.3.1 - # via hydra-core + # via + # -r dependencies/requirements-windows-arm64-cpu.txt + # hydra-core onnx==1.23.2 # via -r dependencies/requirements-windows-arm64-cpu.txt onnxruntime==1.31.0 @@ -619,6 +630,7 @@ pyyaml==6.0.3 # accelerate # gguf # huggingface-hub + # hyperpyyaml # jupyter-events # omegaconf # peft @@ -681,6 +693,10 @@ rpds-py==2026.9.1 # via # jsonschema # referencing +ruamel-yaml==0.17.33 + # via + # -r dependencies/requirements-windows-arm64-cpu.txt + # hyperpyyaml safetensors==0.8.0 # via # -r dependencies/requirements-windows-arm64-cpu.txt @@ -844,6 +860,8 @@ truststore==0.10.5 # via # httpcore2 # httpx2 +typeguard==4.6.0 + # via inflect typer==0.27.3 # via # fastapi-cli @@ -875,6 +893,7 @@ typing-extensions==4.16.0 # spandrel # starlette # torch + # typeguard # typing-inspection typing-inspection==0.4.4 # via diff --git a/dependencies/windows-dependencies.yaml b/dependencies/windows-dependencies.yaml index 0ae96257..cb287157 100644 --- a/dependencies/windows-dependencies.yaml +++ b/dependencies/windows-dependencies.yaml @@ -296,6 +296,9 @@ humanfriendly: hydra-core: version: "1.3.2" isRoot: true +hyperpyyaml: + version: "1.2.2" + isRoot: true idna: version: "3.4" dependents: @@ -325,6 +328,9 @@ importlib-resources: version: "6.5.2" dependents: - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true ipycanvas: version: "0.13.3" dependents: @@ -351,6 +357,10 @@ itsdangerous: version: "2.2.0" dependents: - flask +jaraco-functools: + version: "4.6.0" + dependents: + - inflect jax: version: "0.5.3" dependents: @@ -471,6 +481,8 @@ moderngl: more-itertools: version: "10.7.0" dependents: + - inflect + - jaraco-functools - openai-whisper mpmath: version: "1.3.0" @@ -564,6 +576,7 @@ nvdiffrast: isRoot: true omegaconf: version: "2.3.0" + isRoot: true dependents: - hydra-core onnx: @@ -773,6 +786,7 @@ pyyaml: - accelerate - gguf - huggingface-hub + - hyperpyyaml - omegaconf - peft - timm @@ -824,6 +838,15 @@ rpds-py: dependents: - jsonschema - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml safetensors: version: "0.8.0" isRoot: true @@ -1031,12 +1054,16 @@ trimesh: triton-windows: version: "3.1.0.post17" isRoot: true +typeguard: + version: "4.6.0" + dependents: + - inflect typer: version: "0.15.2" dependents: - fastapi-cli typing-extensions: - version: "4.12.2" + version: "4.16.0" dependents: - anyio - fastapi @@ -1049,6 +1076,7 @@ typing-extensions: - rich-toolkit - spandrel - torch + - typeguard - typer - typing-inspect - typing-inspection diff --git a/dependencies/windows.lock b/dependencies/windows.lock index cabaa16f..b5d2cf12 100644 --- a/dependencies/windows.lock +++ b/dependencies/windows.lock @@ -201,6 +201,8 @@ humanfriendly==10.0 # via coloredlogs hydra-core==1.3.2 # via -r dependencies/requirements-windows.txt +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-windows.txt idna==3.4 # via # anyio @@ -222,6 +224,8 @@ importlib-metadata==7.1.0 # diffusers importlib-resources==6.5.2 # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-windows.txt ipycanvas==0.13.3 # via kaolin ipyevents==2.0.2 @@ -238,6 +242,8 @@ ipywidgets==8.1.5 # ipyevents itsdangerous==2.2.0 # via flask +jaraco-functools==4.6.0 + # via inflect jax==0.5.3 # via mediapipe jaxlib==0.5.3 @@ -312,7 +318,10 @@ modelscope==1.26.0 moderngl==5.12.0 # via -r dependencies/requirements-windows.txt more-itertools==10.7.0 - # via openai-whisper + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy msgpack==1.1.0 @@ -388,7 +397,9 @@ numpy==1.26.3 nvdiffrast @ https://github.com/sunxfancy/nvdiffrast/releases/download/v0.3.3.1/nvdiffrast-0.3.3.1-cp312-cp312-win_amd64.whl # via -r dependencies/requirements-windows.txt omegaconf==2.3.0 - # via hydra-core + # via + # -r dependencies/requirements-windows.txt + # hydra-core onnx==1.17.0 # via -r dependencies/requirements-windows.txt onnxruntime==1.21.0 @@ -540,6 +551,7 @@ pyyaml==6.0.2 # accelerate # gguf # huggingface-hub + # hyperpyyaml # omegaconf # peft # timm @@ -580,6 +592,12 @@ rpds-py==0.24.0 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-windows.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-windows.txt @@ -733,9 +751,11 @@ trimesh==4.6.6 # via -r dependencies/requirements-windows.txt triton-windows==3.1.0.post17 # via -r dependencies/requirements-windows.txt +typeguard==4.6.0 + # via inflect typer==0.15.2 # via fastapi-cli -typing-extensions==4.12.2 +typing-extensions==4.16.0 # via # anyio # fastapi @@ -748,6 +768,7 @@ typing-extensions==4.12.2 # rich-toolkit # spandrel # torch + # typeguard # typer # typing-inspect # typing-inspection diff --git a/extensions/Voice/README.md b/extensions/Voice/README.md new file mode 100644 index 00000000..f7aa2c0c --- /dev/null +++ b/extensions/Voice/README.md @@ -0,0 +1,85 @@ +# CosyVoice2 + +Voice 使用仓库内的 `vendor/cosyvoice` 与 `vendor/matcha`,Python 依赖统一由根目录 +`dependencies/dependencies.yaml` 声明,并生成各平台 requirements、lock 和依赖图。 +无需进入 Voice 目录另建 venv,也不要安装上游训练 requirements 覆盖 SSUI 的 Torch。 + +## 从根目录运行 + +先按 SSUI 的标准流程安装对应运行时。Windows AMD 使用项目的 Windows x64 ROCm +配置和 ROCm PyTorch;`torch.cuda.is_available()` 在 ROCm 下也应为 `True`。 +已有环境新增依赖后使用 `yarn install-requirements`,环境路径选择见 +[`doc/DevelopmentEnvironments.md`](../../doc/DevelopmentEnvironments.md)。 + +```powershell +# 下载固定版本官方权重、参考音频,生成中文语音并检查 GPU 驻留和输出。 +yarn probe:cosyvoice --download --require-rocm + +# 使用本地模型与自己的参考音频;prompt-text 必须是参考音频的真实转写。 +yarn probe:cosyvoice --require-rocm --model-dir 'D:/models/CosyVoice2-0.5B' --prompt 'D:/voice/reference.wav' --prompt-text '参考音频中说的原文。' + +# 同一参考音色生成英文,不要求参考音频转写。 +yarn probe:cosyvoice --download --require-rocm --cross-lingual --text 'Hello, welcome to speech synthesis.' +``` + +默认模型目录为根目录 `.build/models/CosyVoice2-0.5B`,输出为 `.build/cosyvoice/` 下的 +`zero_shot.wav` 或 `cross_lingual.wav`,旁边的 JSON 保存设备、采样率、时长、耗时及幅度。 +`--output-dir` 可更改输出目录。模型 revision 固定为 +`eec1ae6c79877dbd9379285cf8789c9e0879293d`,下载仅包含推理必需文件,约 4.1 GB。 +官方参考音频通过 SHA-256 校验。首次启动需要构建中英文文本规范化 FST 缓存。 + +探针默认使用 4 个 CPU 线程,支持 `--threads` 调整;缓存写入 `.build/`。 +应用内的文本规范化缓存默认位于 `HF_HOME/ssui/cosyvoice2-tn-v1`(无 HF_HOME 时使用 +`~/.cache/ssui/cosyvoice2-tn-v1`),可通过 `SSUI_COSYVOICE_TN_CACHE` 覆盖。 + +## SSUI 接口 + +```python +from ssui.base import Audio +from ssui_voice.Cosyvoice import CosyVoiceModel, GenVoice + +model = CosyVoiceModel.load("D:/models/CosyVoice2-0.5B") +voice = GenVoice( + config("Text to Speech"), model, "你好,欢迎使用语音合成。", + Audio.load("D:/voice/reference.wav"), + prompt_text="参考音频中说的原文。", +) +``` + +`use_zero_shot=True` 时必须同时提供参考音频和 `prompt_text`,或先通过 +`model.getModel().add_zero_shot_spk(...)` 注册 `speaker_0` 等 speaker ID。 +`use_zero_shot=False` 走跨语言模式,仍需要参考音频。 +`zero_shot_prompt=False` 表示不读取传入音频,适用于已注册 speaker 的零样本模式。 +生成器会拼接所有句子的音频,返回 24 kHz `Voice`。 + +## ROCm 路径与验证 + +2026-11-07 在 Windows、AMD Radeon AI PRO R9700、PyTorch `2.9.1+rocm7.2.1` +上完成真实权重验证:LLM、Flow、HiFT 均为 `cuda:0`(PyTorch 的 HIP 设备命名), +使用 FP32;ONNX 声纹与语音分词前处理使用 CPUExecutionProvider。 +无需 NVIDIA CUDA、TensorRT、vLLM 或 FlashAttention。 +SoundFile 负责文件解码,避免 torchaudio 2.9 的 TorchCodec 依赖。 + +中文输入“你好,这是使用 AMD 显卡生成的语音。欢迎使用语音合成。”生成约 6.68 秒音频; +中文参考音色生成英文“Hello, welcome to speech synthesis.”约 2.12 秒。 +两份输出均通过有限值、非静音检查,并用 Whisper base 独立识别确认对应内容。 +这些是本机功能验证,不代表其他 GPU 或平台已完成推理验收,也不代表实时速度。 + +```powershell +yarn test_on cosyvoice_adapter_test +yarn test_on cosyvoice_cache_test +$env:SSUI_RUN_MODEL_TESTS = '1' +$env:SSUI_COSYVOICE_REQUIRE_ROCM = '1' +yarn test_on model.cosyvoice_model_test +``` + +真实模型测试默认下载到配置的 `TEST_DATA_DIR/cosyvoice`,可设置 +`SSUI_COSYVOICE_MODEL_DIR` 使用已有权重,`SSUI_COSYVOICE_PROMPT` 复用官方参考音频。 +自定义参考转写请直接使用探针的 `--prompt-text`。 + +本次已同步除 Intel macOS 外的平台锁文件。Intel macOS 仍需先取得项目要求的已验证 +原生 PyTorch wheelhouse,并设置 `SSUI_NATIVE_WHEELHOUSE` 后运行 +`yarn update-lock:no-upgrade --target macosx-x86_64`;当前 Windows 验证环境没有该产物。 + +上游:[CosyVoice](https://github.com/FunAudioLLM/CosyVoice)、 +[CosyVoice2 权重](https://huggingface.co/FunAudioLLM/CosyVoice2-0.5B)。 diff --git a/extensions/Voice/ssui_voice/Cosyvoice.py b/extensions/Voice/ssui_voice/Cosyvoice.py index e5beeed4..5964128b 100644 --- a/extensions/Voice/ssui_voice/Cosyvoice.py +++ b/extensions/Voice/ssui_voice/Cosyvoice.py @@ -1,7 +1,6 @@ from ssui.model_memory import ModelMemoryOptions -from typing import Optional, List, Union, Generator +from typing import Optional import torch -import torchaudio from ssui.config import SSUIConfig from ssui.base import Audio, Voice from ssui.annotation import param @@ -38,49 +37,42 @@ def GenVoice( model: CosyVoiceModel, text: str, prompt_audio: Optional[Audio] = None, + prompt_text: str = "", ): if config.is_prepare(): return Voice() - print("GenVoice executed") - print("seed:", config["seed"]) - print("use_zero_shot:", config["use_zero_shot"]) - print("zero_shot_prompt:", config["zero_shot_prompt"]) - print("speaker_id:", config["speaker_id"]) - - # Convert prompt audio to the required format if provided + voice = model.getModel() + if voice is None: + raise ValueError("Load a CosyVoice2 model before generating speech") + if not text.strip(): + raise ValueError("Speech text must not be empty") + torch.manual_seed(int(config["seed"])) prompt_speech_16k = None - if prompt_audio and config["zero_shot_prompt"]: + if prompt_audio is not None and config["zero_shot_prompt"]: + if not prompt_audio.path: + raise ValueError("Reference audio must have a file path") prompt_speech_16k = load_wav(prompt_audio.path, 16000) - # Generate audio based on configuration + speaker = f"speaker_{config['speaker_id']}" if config["use_zero_shot"]: - if prompt_speech_16k: - # Use zero-shot with prompt - outputs = model.getModel().inference_zero_shot( - text, - f"Speaker {config['speaker_id']}", - prompt_speech_16k, - stream=False - ) + if prompt_speech_16k is not None: + if not prompt_text.strip(): + raise ValueError("Zero-shot synthesis requires the reference audio transcript (prompt_text)") + outputs = voice.inference_zero_shot( + text, prompt_text, prompt_speech_16k, stream=False) else: - # Use zero-shot without prompt - outputs = model.getModel().inference_zero_shot( - text, - "", - "", - zero_shot_spk_id=f"speaker_{config['speaker_id']}", - stream=False - ) + if speaker not in voice.frontend.spk2info: + raise ValueError(f"Provide reference audio and prompt_text, or register {speaker} first") + outputs = voice.inference_zero_shot( + text, "", "", zero_shot_spk_id=speaker, stream=False) else: - # Use cross-lingual inference - outputs = model.getModel().inference_cross_lingual( - text, - prompt_speech_16k if prompt_speech_16k else None, - stream=False - ) - - # Get the first output audio - output_audio = next(outputs)['tts_speech'] - - return Voice(audio=output_audio, fps=model.getModel().sample_rate, text=text) + if prompt_speech_16k is None: + raise ValueError("Cross-lingual synthesis requires reference audio") + outputs = voice.inference_cross_lingual(text, prompt_speech_16k, stream=False) + + # CosyVoice yields one chunk for each normalized sentence, even without streaming. + chunks = [chunk['tts_speech'].detach().cpu() for chunk in outputs] + if not chunks: + raise RuntimeError("CosyVoice2 generated no audio") + return Voice(audio=torch.cat(chunks, dim=-1), fps=voice.sample_rate, text=text) diff --git a/extensions/Voice/vendor/cosyvoice/cli/frontend.py b/extensions/Voice/vendor/cosyvoice/cli/frontend.py index 36dcd189..c43c27be 100644 --- a/extensions/Voice/vendor/cosyvoice/cli/frontend.py +++ b/extensions/Voice/vendor/cosyvoice/cli/frontend.py @@ -52,9 +52,14 @@ def __init__(self, option.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL option.intra_op_num_threads = 1 self.campplus_session = onnxruntime.InferenceSession(campplus_model, sess_options=option, providers=["CPUExecutionProvider"]) - self.speech_tokenizer_session = onnxruntime.InferenceSession(speech_tokenizer_model, sess_options=option, - providers=["CUDAExecutionProvider" if torch.cuda.is_available() else - "CPUExecutionProvider"]) + # ROCm PyTorch also exposes torch.cuda; that does not imply that ORT + # has NVIDIA's CUDA provider. Windows ROCm uses CPU for the tokenizer. + providers = ["CPUExecutionProvider"] + if (torch.cuda.is_available() and not torch.version.hip + and "CUDAExecutionProvider" in onnxruntime.get_available_providers()): + providers.insert(0, "CUDAExecutionProvider") + self.speech_tokenizer_session = onnxruntime.InferenceSession( + speech_tokenizer_model, sess_options=option, providers=providers) if os.path.exists(spk2info): self.spk2info = torch.load(spk2info, map_location=self.device) else: @@ -68,8 +73,14 @@ def __init__(self, 'failed to initialize ttsfrd resource' self.frd.set_lang_type('pinyinvg') else: - self.zh_tn_model = ZhNormalizer(remove_erhua=False, full_to_half=False, overwrite_cache=True) - self.en_tn_model = EnNormalizer() + # Installed packages may be read-only. Keep reusable FSTs in a user + # cache, separated from other normalizer configurations. + cache_dir = os.environ.get('SSUI_COSYVOICE_TN_CACHE') or os.path.join( + os.environ.get('HF_HOME', os.path.expanduser('~/.cache')), + 'ssui', 'cosyvoice2-tn-v1') + self.zh_tn_model = ZhNormalizer(cache_dir=cache_dir, remove_erhua=False, + full_to_half=False, overwrite_cache=False) + self.en_tn_model = EnNormalizer(cache_dir=cache_dir) self.inflect_parser = inflect.engine() def _extract_text_token(self, text): diff --git a/extensions/Voice/vendor/cosyvoice/llm/llm.py b/extensions/Voice/vendor/cosyvoice/llm/llm.py index 670ae69c..e8f9bd80 100644 --- a/extensions/Voice/vendor/cosyvoice/llm/llm.py +++ b/extensions/Voice/vendor/cosyvoice/llm/llm.py @@ -241,6 +241,15 @@ def forward(self, xs: torch.Tensor, xs_lens: torch.Tensor): def forward_one_step(self, xs, masks, cache=None): input_masks = masks[:, -1, :] + # New Transformers Cache implementations require the mask to cover + # both the cached prefix and the current tokens. A one-token mask + # otherwise hides the entire prompt during incremental decoding. + if cache is not None: + past_length = (cache.get_seq_length() if hasattr(cache, 'get_seq_length') + else cache[0][0].shape[2]) + missing = past_length + xs.shape[1] - input_masks.shape[1] + if missing > 0: + input_masks = F.pad(input_masks, (missing, 0), value=True) outs = self.model( inputs_embeds=xs, attention_mask=input_masks, diff --git a/extensions/Voice/vendor/cosyvoice/utils/file_utils.py b/extensions/Voice/vendor/cosyvoice/utils/file_utils.py index ae860c95..0fa00cbe 100644 --- a/extensions/Voice/vendor/cosyvoice/utils/file_utils.py +++ b/extensions/Voice/vendor/cosyvoice/utils/file_utils.py @@ -15,6 +15,8 @@ import json import torchaudio +import soundfile +import torch import logging logging.getLogger('matplotlib').setLevel(logging.WARNING) logging.basicConfig(level=logging.DEBUG, @@ -39,10 +41,12 @@ def read_json_lists(list_file): def load_wav(wav, target_sr): - speech, sample_rate = torchaudio.load(wav, backend='soundfile') + # torchaudio >=2.9 ignores backend= and requires TorchCodec, which is not + # part of the Windows ROCm runtime. Decode with the shared SoundFile dep. + samples, sample_rate = soundfile.read(wav, dtype='float32', always_2d=True) + speech = torch.from_numpy(samples.T.copy()) speech = speech.mean(dim=0, keepdim=True) if sample_rate != target_sr: - assert sample_rate > target_sr, 'wav sample rate {} must be greater than {}'.format(sample_rate, target_sr) speech = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=target_sr)(speech) return speech diff --git a/extensions/Voice/vendor/matcha/models/components/flow_matching.py b/extensions/Voice/vendor/matcha/models/components/flow_matching.py index 5cad7431..f6d66f53 100644 --- a/extensions/Voice/vendor/matcha/models/components/flow_matching.py +++ b/extensions/Voice/vendor/matcha/models/components/flow_matching.py @@ -4,9 +4,7 @@ import torch.nn.functional as F from matcha.models.components.decoder import Decoder -from matcha.utils.pylogger import get_pylogger - -log = get_pylogger(__name__) +# Inference does not need Matcha's Lightning/Hydra training utilities. class BASECFM(torch.nn.Module, ABC): diff --git a/extensions/Voice/vendor/matcha/utils/__init__.py b/extensions/Voice/vendor/matcha/utils/__init__.py index 074db646..9db6ffd9 100644 --- a/extensions/Voice/vendor/matcha/utils/__init__.py +++ b/extensions/Voice/vendor/matcha/utils/__init__.py @@ -1,5 +1,23 @@ -from matcha.utils.instantiators import instantiate_callbacks, instantiate_loggers -from matcha.utils.logging_utils import log_hyperparameters -from matcha.utils.pylogger import get_pylogger -from matcha.utils.rich_utils import enforce_tags, print_config_tree -from matcha.utils.utils import extras, get_metric_value, task_wrapper +"""Load training utilities only when requested, keeping audio inference light.""" +from importlib import import_module + +_EXPORTS = { + "instantiate_callbacks": "instantiators", + "instantiate_loggers": "instantiators", + "log_hyperparameters": "logging_utils", + "get_pylogger": "pylogger", + "enforce_tags": "rich_utils", + "print_config_tree": "rich_utils", + "extras": "utils", + "get_metric_value": "utils", + "task_wrapper": "utils", +} +__all__ = list(_EXPORTS) + + +def __getattr__(name): + if name not in _EXPORTS: + raise AttributeError(name) + value = getattr(import_module(f"{__name__}.{_EXPORTS[name]}"), name) + globals()[name] = value + return value diff --git a/package.json b/package.json index 0086a178..6cda7433 100644 --- a/package.json +++ b/package.json @@ -32,6 +32,7 @@ "test:mcp:server": "node dependencies/venv.cjs python -m uvicorn tests.mcp_ui_app:app --host 127.0.0.1 --port 17424", "test:mcp:ui": "playwright test --config playwright.mcp.config.ts", "probe:cumesh": "node dependencies/venv.cjs python -u scripts/probe_cumesh.py", + "probe:cosyvoice": "node dependencies/venv.cjs python -u scripts/probe_cosyvoice.py", "probe:large-sparse": "node dependencies/venv.cjs python -u scripts/probe_large_sparse.py", "prepare:cumesh-rocm": "node dependencies/venv.cjs python dependencies/prepare_cumesh_rocm.py", "build:cumesh-rocm": "powershell -NoProfile -ExecutionPolicy Bypass -File dependencies/build_nvdiffrast_rocm.ps1 -Source .build/cumesh-rocm/source -DevelInclude .build/cumesh-rocm/headers", diff --git a/scripts/probe_cosyvoice.py b/scripts/probe_cosyvoice.py new file mode 100644 index 00000000..0796aeae --- /dev/null +++ b/scripts/probe_cosyvoice.py @@ -0,0 +1,100 @@ +"""Real CosyVoice2 inference through SSUI; run with yarn probe:cosyvoice.""" +import argparse +import hashlib +import json +import os +from pathlib import Path +import sys +import time +import urllib.request + +ROOT = Path(__file__).resolve().parents[1] +sys.path.insert(0, str(ROOT)) +sys.path[:0] = [str(ROOT / "extensions/Voice"), str(ROOT / "extensions/Voice/vendor")] + +MODEL_REPO = "FunAudioLLM/CosyVoice2-0.5B" +MODEL_REVISION = "eec1ae6c79877dbd9379285cf8789c9e0879293d" +PROMPT_URL = "https://raw.githubusercontent.com/FunAudioLLM/CosyVoice/main/asset/zero_shot_prompt.wav" +PROMPT_SHA256 = "c7b31d6dbe7cc6a716dded00550db5b50940bf209e424e4ad207b12e657c8ff6" +PROMPT_TEXT = "希望你以后能够做的比我还好呦。" + + +def main(): + parser = argparse.ArgumentParser(description=__doc__) + parser.add_argument("--model-dir", type=Path, default=ROOT / ".build/models/CosyVoice2-0.5B") + parser.add_argument("--prompt", type=Path) + parser.add_argument("--prompt-text", default=PROMPT_TEXT) + parser.add_argument("--text", default="你好,这是使用 AMD 显卡生成的语音。欢迎使用语音合成。") + parser.add_argument("--output-dir", type=Path, default=ROOT / ".build/cosyvoice") + parser.add_argument("--download", action="store_true", help="Download the pinned official model and example prompt") + parser.add_argument("--require-rocm", action="store_true") + parser.add_argument("--cross-lingual", action="store_true") + parser.add_argument("--threads", type=int, default=4, help="CPU threads for preprocessing (default: 4)") + args = parser.parse_args() + args.output_dir.mkdir(parents=True, exist_ok=True) + os.environ.setdefault("NUMBA_CACHE_DIR", str(ROOT / ".build/numba-cache")) + os.environ.setdefault("SSUI_COSYVOICE_TN_CACHE", str(ROOT / ".build/cosyvoice-tn")) + for name, folder in (("MIOPEN_USER_DB_PATH", "miopen-db"), + ("MIOPEN_CUSTOM_CACHE_DIR", "miopen-cache")): + cache = Path(os.environ.setdefault(name, str(ROOT / ".build" / folder))) + cache.mkdir(parents=True, exist_ok=True) + + import soundfile as sf + import torch + from ssui.base import Audio + from ssui.config import SSUIConfig + from ssui_voice.Cosyvoice import CosyVoiceModel, GenVoice + + if args.threads < 1: + raise ValueError("--threads must be positive") + torch.set_num_threads(args.threads) + if args.require_rocm and not (torch.version.hip and torch.cuda.is_available()): + raise RuntimeError("A working ROCm PyTorch GPU is required") + prompt = args.prompt or args.output_dir / "prompt.wav" + if args.download: + from huggingface_hub import snapshot_download + snapshot_download(MODEL_REPO, revision=MODEL_REVISION, local_dir=args.model_dir, + allow_patterns=["CosyVoice-BlankEN/*", "cosyvoice2.yaml", "campplus.onnx", + "speech_tokenizer_v2.onnx", "llm.pt", "flow.pt", "hift.pt"]) + if args.prompt is None and not prompt.exists(): + urllib.request.urlretrieve(PROMPT_URL, prompt) + if args.prompt is None and hashlib.sha256(prompt.read_bytes()).hexdigest() != PROMPT_SHA256: + raise RuntimeError("Official reference audio checksum mismatch; supply your own --prompt and --prompt-text") + if not prompt.is_file(): + raise FileNotFoundError(f"Reference audio missing: {prompt}; supply --prompt or --download") + + start = time.perf_counter() + model = CosyVoiceModel.load(str(args.model_dir)) + engine = model.getModel() + components = {name: str(next(getattr(engine.model, name).parameters()).device) + for name in ("llm", "flow", "hift")} + if args.require_rocm and any(not device.startswith("cuda") for device in components.values()): + raise RuntimeError(f"Model is not on the ROCm GPU: {components}") + load_seconds = time.perf_counter() - start + config = SSUIConfig()("cosyvoice") + config["use_zero_shot"] = not args.cross_lingual + start = time.perf_counter() + voice = GenVoice(config, model, args.text, Audio.load(str(prompt)), args.prompt_text) + samples = voice._audio + if not torch.isfinite(samples).all() or samples.numel() == 0 or samples.abs().max() == 0: + raise RuntimeError("Generated audio is empty, silent, or nonfinite") + elapsed = time.perf_counter() - start + output = args.output_dir / ("cross_lingual.wav" if args.cross_lingual else "zero_shot.wav") + sf.write(str(output), samples.squeeze(0).numpy(), voice.sample_rate) + report = { + "torch": torch.__version__, "hip": torch.version.hip, + "gpu": torch.cuda.get_device_name() if torch.cuda.is_available() else None, + "components": components, "onnx_providers": engine.frontend.speech_tokenizer_session.get_providers(), + "model_revision": MODEL_REVISION, "model_dir": str(args.model_dir.resolve()), + "sample_rate": voice.sample_rate, "audio_seconds": samples.shape[-1] / voice.sample_rate, + "load_seconds": load_seconds, "inference_seconds": elapsed, + "peak": samples.abs().max().item(), "rms": samples.square().mean().sqrt().item(), + "output": str(output.resolve()), "text": args.text, + } + report_path = output.with_suffix(".json") + report_path.write_text(json.dumps(report, ensure_ascii=False, indent=2), encoding="utf-8") + print(json.dumps(report, ensure_ascii=False, indent=2)) + + +if __name__ == "__main__": + main() diff --git a/tests/cosyvoice_adapter_test.py b/tests/cosyvoice_adapter_test.py new file mode 100644 index 00000000..45f885b9 --- /dev/null +++ b/tests/cosyvoice_adapter_test.py @@ -0,0 +1,68 @@ +"""SSUI adapter contracts, independent of model downloads and fake model tests.""" +import importlib.util +from pathlib import Path +import sys +import types +import unittest +from unittest.mock import Mock, patch + +import torch +from ssui.base import Audio +from ssui.config import SSUIConfig + + +class TestCosyVoiceAdapter(unittest.TestCase): + def setUp(self): + cli = types.ModuleType("cosyvoice.cli.cosyvoice") + cli.CosyVoice2 = Mock() + files = types.ModuleType("cosyvoice.utils.file_utils") + self.reference = torch.ones(1, 32000) + files.load_wav = Mock(return_value=self.reference) + path = Path(__file__).resolve().parents[1] / "extensions/Voice/ssui_voice/Cosyvoice.py" + spec = importlib.util.spec_from_file_location("cosyvoice_adapter_contract", path) + self.adapter = importlib.util.module_from_spec(spec) + with patch.dict(sys.modules, {cli.__name__: cli, files.__name__: files}): + spec.loader.exec_module(self.adapter) + self.engine = Mock(sample_rate=24000) + self.engine.frontend.spk2info = {} + self.engine.inference_zero_shot.return_value = iter([ + {"tts_speech": torch.ones(1, 20)}, {"tts_speech": torch.full((1, 30), 2.)}]) + self.engine.inference_cross_lingual.return_value = iter([{"tts_speech": torch.ones(1, 40)}]) + self.model = self.adapter.CosyVoiceModel("test", self.engine) + self.config = SSUIConfig()("voice") + + def test_reference_transcript_and_all_sentence_chunks(self): + result = self.adapter.GenVoice(self.config, self.model, "两句。测试。", Audio.load("ref.wav"), "原文") + self.engine.inference_zero_shot.assert_called_once_with("两句。测试。", "原文", self.reference, stream=False) + self.assertEqual(result.sample_rate, 24000) + self.assertEqual(result._audio.shape, (1, 50)) + self.assertTrue(torch.equal(result._audio[:, 20:], torch.full((1, 30), 2.))) + + def test_reference_requires_transcript(self): + with self.assertRaisesRegex(ValueError, "transcript"): + self.adapter.GenVoice(self.config, self.model, "测试", Audio.load("ref.wav")) + self.engine.inference_zero_shot.assert_not_called() + + def test_missing_reference_requires_registered_speaker(self): + with self.assertRaisesRegex(ValueError, "register"): + self.adapter.GenVoice(self.config, self.model, "测试") + self.engine.frontend.spk2info["speaker_0"] = {} + self.adapter.GenVoice(self.config, self.model, "测试") + self.engine.inference_zero_shot.assert_called_once_with("测试", "", "", zero_shot_spk_id="speaker_0", stream=False) + + def test_cross_lingual_accepts_audio_tensor_without_boolean_conversion(self): + self.config["use_zero_shot"] = False + result = self.adapter.GenVoice(self.config, self.model, "Hello", Audio.load("ref.wav")) + self.engine.inference_cross_lingual.assert_called_once_with("Hello", self.reference, stream=False) + self.assertEqual(result._audio.shape[-1], 40) + + def test_cross_lingual_requires_reference(self): + self.config["use_zero_shot"] = False + with self.assertRaisesRegex(ValueError, "reference audio"): + self.adapter.GenVoice(self.config, self.model, "Hello") + + def test_empty_model_output_fails(self): + self.engine.inference_cross_lingual.return_value = iter([]) + self.config["use_zero_shot"] = False + with self.assertRaisesRegex(RuntimeError, "no audio"): + self.adapter.GenVoice(self.config, self.model, "Hello", Audio.load("ref.wav")) diff --git a/tests/cosyvoice_cache_test.py b/tests/cosyvoice_cache_test.py new file mode 100644 index 00000000..5a306000 --- /dev/null +++ b/tests/cosyvoice_cache_test.py @@ -0,0 +1,36 @@ +"""Check incremental Qwen attention against full-prefix decoding on CPU.""" +import subprocess +import sys +import unittest +from pathlib import Path + + +class TestCosyVoiceCache(unittest.TestCase): + def test_cached_decoding_matches_full_prefix(self): + # Other contract suites replace Transformers in sys.modules. Use an + # isolated interpreter to exercise the actual installed library. + code = ''' +import torch +torch.set_num_threads(2) +from tests.utils import add_extension_paths +add_extension_paths() +from transformers import Qwen2Config, Qwen2ForCausalLM +from cosyvoice.llm.llm import Qwen2Encoder +torch.manual_seed(7) +cfg = Qwen2Config(vocab_size=32, hidden_size=16, intermediate_size=32, + num_hidden_layers=2, num_attention_heads=2, num_key_value_heads=2) +cfg._attn_implementation = 'eager' +encoder = Qwen2Encoder.__new__(Qwen2Encoder) +torch.nn.Module.__init__(encoder) +encoder.model = Qwen2ForCausalLM(cfg).eval() +x = torch.randn(1, 5, 16) +with torch.inference_mode(): + full, _ = encoder.forward_one_step(x, torch.ones(1, 5, 5, dtype=torch.bool).tril()) + _, cache = encoder.forward_one_step(x[:, :4], torch.ones(1, 4, 4, dtype=torch.bool).tril()) + step, _ = encoder.forward_one_step(x[:, 4:], torch.ones(1, 1, 1, dtype=torch.bool), cache) +torch.testing.assert_close(step[:, -1], full[:, -1], atol=1e-5, rtol=1e-5) +''' + result = subprocess.run([sys.executable, "-c", code], + cwd=Path(__file__).resolve().parents[1], + capture_output=True, text=True, timeout=90) + self.assertEqual(result.returncode, 0, result.stdout + result.stderr) diff --git a/tests/model/cosyvoice_model_test.py b/tests/model/cosyvoice_model_test.py index 05d7892a..b6d28c92 100644 --- a/tests/model/cosyvoice_model_test.py +++ b/tests/model/cosyvoice_model_test.py @@ -1,42 +1,29 @@ +import json import os +from pathlib import Path +import subprocess +import sys import unittest -import torch - -from tests.utils import download_if_needed, should_run_model_tests, model_output_dir +from tests.utils import get_test_data_dir, should_run_model_tests, model_output_dir @unittest.skipUnless(should_run_model_tests(), "需要 SSUI_RUN_MODEL_TESTS=1") class TestCosyVoiceModel(unittest.TestCase): def test_cosyvoice(self): - from cosyvoice.cli.cosyvoice import CosyVoice2 - - import torchaudio - - model_dir = str(download_if_needed("cosyvoice")) - cosyvoice = CosyVoice2( - model_dir, load_jit=False, load_trt=False, fp16=False - ) - prompt_speech_16k = torch.zeros(16000, dtype=torch.float32) - - assert cosyvoice.add_zero_shot_spk("你好", prompt_speech_16k, "my_spk") is True - - with model_output_dir("cosyvoice") as tmp: - saved = 0 - for i, chunk in enumerate( - cosyvoice.inference_zero_shot( - "收到好友从远方寄来的生日礼物。", - "希望你以后能够做的比我还好呦。", - prompt_speech_16k, - stream=False, - ) - ): - torchaudio.save( - os.path.join(tmp, f"zero_shot_{i}.wav"), - chunk["tts_speech"], - cosyvoice.sample_rate, - ) - saved += 1 - if saved >= 2: - break - self.assertGreaterEqual(saved, 1) + # Exercise the public adapter with real reference audio and transcript. + root = Path(__file__).resolve().parents[2] + model_dir = os.environ.get("SSUI_COSYVOICE_MODEL_DIR", str(get_test_data_dir() / "cosyvoice")) + with model_output_dir("cosyvoice") as output: + command = [sys.executable, str(root / "scripts/probe_cosyvoice.py"), + "--model-dir", model_dir, "--output-dir", str(output), "--download"] + if os.environ.get("SSUI_COSYVOICE_REQUIRE_ROCM") == "1": + command.append("--require-rocm") + if os.environ.get("SSUI_COSYVOICE_PROMPT"): + command.extend(["--prompt", os.environ["SSUI_COSYVOICE_PROMPT"]]) + subprocess.run(command, cwd=root, check=True, timeout=900) + report = json.loads((Path(output) / "zero_shot.json").read_text(encoding="utf-8")) + self.assertEqual(report["sample_rate"], 24000) + self.assertGreater(report["audio_seconds"], 1) + self.assertGreater(report["rms"], 0) + self.assertTrue(Path(report["output"]).is_file()) diff --git a/tests/model_manifest.yaml b/tests/model_manifest.yaml index 1953a62a..80ece466 100644 --- a/tests/model_manifest.yaml +++ b/tests/model_manifest.yaml @@ -63,10 +63,10 @@ flux: sha256: "afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38" cosyvoice: - repo: "iic/CosyVoice2-0.5B" + repo: "FunAudioLLM/CosyVoice2-0.5B" snapshot: true - revision: "main" - size_gb: 1.0 + revision: "eec1ae6c79877dbd9379285cf8789c9e0879293d" + size_gb: 4.5 sha256: "" llm: From 7b7e8fb81db9d0e62c28e826612d16b6ee5eaf8e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E8=A5=BF=E9=A3=8E=E9=80=8D=E9=81=A5=E6=B8=B8?= Date: Thu, 12 Nov 2026 00:08:27 +0800 Subject: [PATCH 2/5] fix(deps): supply OpenFst for macOS and constrain isolated lock builds --- .github/workflows/ci.yml | 7 ++++ .github/workflows/macos-intel-wheels.yml | 5 ++- .github/workflows/release.yml | 6 +++ .github/workflows/windows-arm64-wheels.yml | 8 +++- dependencies/dependency_config.test.cjs | 14 +++++++ dependencies/install_requirements.cjs | 3 +- dependencies/macos_prerequisites.cjs | 47 ++++++++++++++++++++++ dependencies/update_lock.cjs | 1 + extensions/Voice/README.md | 6 +++ scripts/check_runtime.py | 7 ++++ 10 files changed, 101 insertions(+), 3 deletions(-) create mode 100644 dependencies/macos_prerequisites.cjs diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index 0cb8fed3..fc367506 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -320,8 +320,15 @@ jobs: timeout-minutes: 5 run: uv pip install packaging==24.1 + - name: Prepare macOS OpenFst GRM prerequisites + if: runner.os == 'macOS' + timeout-minutes: 15 + run: node dependencies/macos_prerequisites.cjs --github-env + - name: Install requirements timeout-minutes: 20 + # macOS has no upstream Pynini wheel; expose the real OpenFst GRM SDK. + # Keep the full lock installation and native build isolation below. run: uv pip install -r dependencies/${{ steps.runtime.outputs.lock }} --index-strategy unsafe-best-match --build-constraint dependencies/build-constraints.txt - name: Install system dependencies diff --git a/.github/workflows/macos-intel-wheels.yml b/.github/workflows/macos-intel-wheels.yml index ffaee382..fd19cf27 100644 --- a/.github/workflows/macos-intel-wheels.yml +++ b/.github/workflows/macos-intel-wheels.yml @@ -46,12 +46,15 @@ jobs: run: node dependencies/native_wheels.cjs native-wheels - name: Verify portable native libraries run: python scripts/verify_macos_wheels.py + - name: Prepare macOS OpenFst GRM prerequisites + timeout-minutes: 15 + run: node dependencies/macos_prerequisites.cjs --github-env - name: Resolve the Intel dependency lock with the native wheels env: MACOSX_DEPLOYMENT_TARGET: '14.0' run: | cp dependencies/macosx-x86_64.lock native-wheels/macosx-x86_64.lock - uv pip compile dependencies/requirements-macosx-x86_64.txt --find-links native-wheels --python-version 3.12.8 --python-platform x86_64-apple-darwin --index-strategy unsafe-best-match --emit-index-url -o native-wheels/macosx-x86_64.lock + uv pip compile dependencies/requirements-macosx-x86_64.txt --find-links native-wheels --python-version 3.12.8 --python-platform x86_64-apple-darwin --index-strategy unsafe-best-match --build-constraints dependencies/build-constraints.txt --emit-index-url -o native-wheels/macosx-x86_64.lock - name: Preserve native wheels if: always() uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml index ad85f3d0..9b59554f 100644 --- a/.github/workflows/release.yml +++ b/.github/workflows/release.yml @@ -182,8 +182,14 @@ jobs: if: matrix.id == 'macos-x86_64' run: | node dependencies/native_wheels.cjs "$RUNNER_TEMP/native-wheels" + cp "$RUNNER_TEMP/native-wheels/macosx-x86_64.lock" dependencies/macosx-x86_64.lock echo "SSUI_NATIVE_WHEELHOUSE=$RUNNER_TEMP/native-wheels" >> "$GITHUB_ENV" + - name: Prepare macOS OpenFst GRM prerequisites + if: runner.os == 'macOS' + timeout-minutes: 15 + run: node dependencies/macos_prerequisites.cjs --github-env + - name: Install requirements timeout-minutes: 30 shell: bash diff --git a/.github/workflows/windows-arm64-wheels.yml b/.github/workflows/windows-arm64-wheels.yml index 8f726f02..9679d1ae 100644 --- a/.github/workflows/windows-arm64-wheels.yml +++ b/.github/workflows/windows-arm64-wheels.yml @@ -121,7 +121,13 @@ jobs: path: native-wheels - name: Check Intel wheel manifest if: matrix.id == 'macos-x86_64' - run: node dependencies/native_wheels.cjs native-wheels + run: | + node dependencies/native_wheels.cjs native-wheels + cp native-wheels/macosx-x86_64.lock dependencies/macosx-x86_64.lock + - name: Prepare macOS OpenFst GRM prerequisites + if: runner.os == 'macOS' + timeout-minutes: 15 + run: node dependencies/macos_prerequisites.cjs --github-env - name: Install the full native dependency lock timeout-minutes: 25 env: diff --git a/dependencies/dependency_config.test.cjs b/dependencies/dependency_config.test.cjs index d68b0bdc..0283d941 100644 --- a/dependencies/dependency_config.test.cjs +++ b/dependencies/dependency_config.test.cjs @@ -6,6 +6,19 @@ const { loadConfig, validateConfig, resolveDependencies, renderRequirements, gen const { extensionRequirements } = require('./combine_requirements.cjs'); const { parseArgs, compileArgs, parseLockFile } = require('./update_lock.cjs'); const { main: generate } = require('./generate_requirements.cjs'); +const { prepare: prepareMacos } = require('./macos_prerequisites.cjs'); + +test('macOS native Pynini build preserves caller flags and requires GRM libraries', () => { + const env = { CPLUS_INCLUDE_PATH: '/custom/include', LIBRARY_PATH: '/custom/lib', LDFLAGS: '-Wl,-rpath,/custom/lib' }; + const run = (_command, args) => args[0] === '--prefix' ? '/opt/homebrew/opt/openfst\n' : 'openfst 1.8.4\n'; + const result = prepareMacos({ platform: 'darwin', env, run, exists: () => true }); + assert.equal(result.CPLUS_INCLUDE_PATH, '/opt/homebrew/opt/openfst/include:/custom/include'); + assert.equal(result.LIBRARY_PATH, '/opt/homebrew/opt/openfst/lib:/custom/lib'); + assert.equal(result.LDFLAGS, '-L/opt/homebrew/opt/openfst/lib -Wl,-rpath,/custom/lib'); + assert.throws(() => prepareMacos({ platform: 'darwin', env, run, + exists: file => !file.endsWith('libfstfar.dylib') }), /GRM prerequisite missing/); + assert.deepEqual(prepareMacos({ platform: 'win32', env, run: () => { throw new Error('must not run brew'); } }), env); +}); function sample() { return { @@ -110,6 +123,7 @@ test('lock compilation uses the requested target, Python version and unchanged l assert.throws(() => parseArgs(['--target']), /缺少/); assert.throws(() => parseArgs(['--target', 'linux-arm64-qnn']), /尚未提供/); const args = compileArgs('macosx', 'in file.txt', 'macosx.lock', true); + assert.equal(args[args.indexOf('--build-constraints') + 1], path.join(__dirname, 'build-constraints.txt')); assert.equal(args[args.indexOf('--python-platform') + 1], 'aarch64-apple-darwin'); assert.equal(args[args.indexOf('--python-version') + 1], config.python.version); assert.ok(args.includes('in file.txt')); diff --git a/dependencies/install_requirements.cjs b/dependencies/install_requirements.cjs index 17a5f706..e07546b6 100644 --- a/dependencies/install_requirements.cjs +++ b/dependencies/install_requirements.cjs @@ -1,5 +1,5 @@ const { resolveConfig, root, executable } = require('./dev_config.cjs'); -const devEnv = resolveConfig(); +let devEnv = resolveConfig(); const { execSync, execFileSync } = require('child_process'); const path = require('path'); const fs = require('fs'); @@ -17,6 +17,7 @@ function getRequirementsFile() { } try { + devEnv = require('./macos_prerequisites.cjs').prepare({ env: devEnv }); const requirementsFile = getRequirementsFile(); const pythonRuntime = JSON.parse(execFileSync(executable('python', devEnv), [ '-c', 'import json, platform, sys; print(json.dumps({"os": sys.platform, "arch": platform.machine()}))' diff --git a/dependencies/macos_prerequisites.cjs b/dependencies/macos_prerequisites.cjs new file mode 100644 index 00000000..7db1f484 --- /dev/null +++ b/dependencies/macos_prerequisites.cjs @@ -0,0 +1,47 @@ +// Pynini's macOS sdist requires OpenFst 1.8.4 with GRM headers/libraries. +// Homebrew's formula enables GRM; do not hide missing native dependencies by +// dropping Pynini or disabling build isolation. +const fs = require('node:fs'); +const path = require('node:path').posix; +const { execFileSync } = require('node:child_process'); + +function prepare({ platform = process.platform, env = process.env, + run = execFileSync, exists = fs.existsSync } = {}) { + if (platform !== 'darwin') return { ...env }; + const brew = (...args) => run('brew', args, { encoding: 'utf8', env }).trim(); + let prefix; + try { prefix = brew('--prefix', 'openfst'); } + catch { brew('install', 'openfst'); prefix = brew('--prefix', 'openfst'); } + // `brew --prefix` can succeed for an uninstalled formula. + if (!exists(path.join(prefix, 'include/fst/util.h'))) { + brew('install', 'openfst'); + prefix = brew('--prefix', 'openfst'); + } + const versions = brew('list', '--versions', 'openfst').split(/\s+/).slice(1); + if (!versions.some(v => /^1\.8\.4(?:_\d+)?$/.test(v))) { + throw new Error('Pynini 2.1.7 requires OpenFst 1.8.4; check the installed Homebrew formula'); + } + for (const file of ['include/fst/util.h', 'include/fst/extensions/far/far.h', + 'lib/libfst.dylib', 'lib/libfstfar.dylib', 'lib/libfstscript.dylib']) { + if (!exists(path.join(prefix, file))) throw new Error(`OpenFst GRM prerequisite missing: ${file}`); + } + const prepend = (name, value, separator) => value + (env[name] ? separator + env[name] : ''); + return { ...env, + CPLUS_INCLUDE_PATH: prepend('CPLUS_INCLUDE_PATH', path.join(prefix, 'include'), ':'), + LIBRARY_PATH: prepend('LIBRARY_PATH', path.join(prefix, 'lib'), ':'), + LDFLAGS: prepend('LDFLAGS', `-L${path.join(prefix, 'lib')}`, ' '), + }; +} + +if (require.main === module) { + const env = prepare(); + if (process.argv.includes('--github-env')) { + for (const key of ['CPLUS_INCLUDE_PATH', 'LIBRARY_PATH', 'LDFLAGS']) { + if (env[key] !== undefined) { + if (/[\r\n]/.test(env[key])) throw new Error(`Invalid newline in ${key}`); + fs.appendFileSync(process.env.GITHUB_ENV, `${key}=${env[key]}\n`); + } + } + } +} +module.exports = { prepare }; diff --git a/dependencies/update_lock.cjs b/dependencies/update_lock.cjs index 5a9ba3bc..cb07d7b8 100644 --- a/dependencies/update_lock.cjs +++ b/dependencies/update_lock.cjs @@ -27,6 +27,7 @@ function compileArgs(target, requirementsFile, lockFile, noUpgrade = false) { const python = pythonDistribution(profile); const nativeWheels = require('./native_wheels.cjs').wheelhouseFor(profile.dependencyTarget); return ['pip', 'compile', requirementsFile, '-o', lockFile, + '--build-constraints', path.join(__dirname, 'build-constraints.txt'), '--python-platform', python.triple, '--python-version', python.version, ...(nativeWheels ? ['--find-links', nativeWheels] : []), ...(profile.binaryOnly ? ['--no-build', '--index-strategy', 'unsafe-best-match'] : []), diff --git a/extensions/Voice/README.md b/extensions/Voice/README.md index f7aa2c0c..65e0066a 100644 --- a/extensions/Voice/README.md +++ b/extensions/Voice/README.md @@ -4,6 +4,12 @@ Voice 使用仓库内的 `vendor/cosyvoice` 与 `vendor/matcha`,Python 依赖 `dependencies/dependencies.yaml` 声明,并生成各平台 requirements、lock 和依赖图。 无需进入 Voice 目录另建 venv,也不要安装上游训练 requirements 覆盖 SSUI 的 Torch。 +macOS 上 Pynini 2.1.7 没有 PyPI wheel,需要 OpenFst 1.8.4 的 GRM 头文件和动态库。 +根目录安装器与 Mac CI 会通过 Homebrew 安装并检查该前置库,并传入编译器/链接器路径; +因此 Mac 开发环境需要先安装 Homebrew。锁文件解析与安装均沿用 +`dependencies/build-constraints.txt` 的隔离构建约束,避免 Whisper 构建时丢失 `pkg_resources`。 +Intel CI 使用验证后的原生 PyTorch wheel 解析完整锁,并将该锁交给后续完整安装与运行检查。 + ## 从根目录运行 先按 SSUI 的标准流程安装对应运行时。Windows AMD 使用项目的 Windows x64 ROCm diff --git a/scripts/check_runtime.py b/scripts/check_runtime.py index 0a79cf83..dbafb35a 100644 --- a/scripts/check_runtime.py +++ b/scripts/check_runtime.py @@ -24,6 +24,13 @@ def main(): assert torch.tensor([1.0, 2.0]).sum().item() == 3 boxes = torch.tensor([[0., 0., 1., 1.], [0., 0., 1., 1.]]) assert torchvision.ops.nms(boxes, torch.tensor([0.9, 0.8]), 0.5).tolist() == [0] + if a.target.startswith('macos-'): + # Exercise the native OpenFst binding after the full lock installation, + # rather than treating a successful sdist metadata resolution as proof. + import pynini + import hyperpyyaml + import whisper + assert (pynini.accep('one') @ pynini.cross('one', '1')).string() == '1' print(f'OK {a.target}: Python {platform.python_version()}, torch {torch.__version__}, torchvision {torchvision.__version__}') From 6b5a62c2dc134c3cc8bc5724539d51b7c0ba32aa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E8=A5=BF=E9=A3=8E=E9=80=8D=E9=81=A5=E6=B8=B8?= Date: Thu, 12 Nov 2026 00:23:47 +0800 Subject: [PATCH 3/5] fix(deps): synchronize verified Intel lock and exercise macOS native inference --- .github/workflows/ci.yml | 7 + dependencies/macosx-x86_64-dependencies.yaml | 1302 ++++++++++++++++++ dependencies/macosx-x86_64.lock | 88 +- extensions/Voice/README.md | 6 +- 4 files changed, 1394 insertions(+), 9 deletions(-) create mode 100644 dependencies/macosx-x86_64-dependencies.yaml diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index fc367506..eff1dd1a 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -331,6 +331,13 @@ jobs: # Keep the full lock installation and native build isolation below. run: uv pip install -r dependencies/${{ steps.runtime.outputs.lock }} --index-strategy unsafe-best-match --build-constraint dependencies/build-constraints.txt + - name: Verify macOS native inference dependencies + if: runner.os == 'macOS' + timeout-minutes: 5 + run: | + runtime_target=$(node -p "require('./dependencies/runtime.cjs').getRuntime().id") + python scripts/check_runtime.py --target "$runtime_target" + - name: Install system dependencies if: ${{ matrix.platform == 'linux' }} timeout-minutes: 30 diff --git a/dependencies/macosx-x86_64-dependencies.yaml b/dependencies/macosx-x86_64-dependencies.yaml new file mode 100644 index 00000000..10970ce4 --- /dev/null +++ b/dependencies/macosx-x86_64-dependencies.yaml @@ -0,0 +1,1302 @@ +# 包依赖关系 +absl-py: + version: "2.5.1" + dependents: + - mediapipe +accelerate: + version: "1.15.0" + isRoot: true + dependents: + - diffusers + - peft +agent-detector: + version: "2.0.0" + dependents: + - fastapi-cloud-cli +aiohappyeyeballs: + version: "2.7.1" + dependents: + - aiohttp +aiohttp: + version: "3.14.5" + isRoot: true +aioshutil: + version: "1.6" + isRoot: true +aiosignal: + version: "1.4.0" + dependents: + - aiohttp +annotated-doc: + version: "0.0.5" + dependents: + - fastapi + - typer +annotated-types: + version: "0.8.0" + dependents: + - pydantic +antlr4-python3-runtime: + version: "4.9.3" + dependents: + - omegaconf +anyio: + version: "4.15.1" + dependents: + - httpx + - jupyter-server + - starlette + - watchfiles +appnope: + version: "1.0.0" + dependents: + - ipykernel +argon2-cffi: + version: "25.1.0" + dependents: + - jupyter-server +argon2-cffi-bindings: + version: "26.1.0" + dependents: + - argon2-cffi +arrow: + version: "1.4.0" + dependents: + - isoduration +asttokens: + version: "3.0.2" + dependents: + - stack-data +async-lru: + version: "2.4.0" + dependents: + - jupyterlab +attrs: + version: "26.1.0" + dependents: + - aiohttp + - cyclopts + - jsonschema + - mediapipe + - referencing +audioread: + version: "3.1.0" + dependents: + - librosa +babel: + version: "2.18.0" + dependents: + - jupyterlab-server +beautifulsoup4: + version: "4.15.0" + dependents: + - nbconvert +blake3: + version: "1.0.11" + isRoot: true +bleach: + version: "6.4.0" + dependents: + - nbconvert +certifi: + version: "2026.10.23" + dependents: + - httpcore + - httpx + - requests + - sentry-sdk +cffi: + version: "2.1.1" + dependents: + - argon2-cffi-bindings + - sounddevice + - soundfile +charset-normalizer: + version: "3.5.2" + dependents: + - requests +click: + version: "8.5.0" + dependents: + - rich-toolkit + - uvicorn +clip-anytorch: + version: "2.6.0" + isRoot: true +cloudpickle: + version: "3.1.2" + dependents: + - joblib +coloredlogs: + version: "15.0.1" + dependents: + - onnxruntime +comm: + version: "0.2.3" + dependents: + - ipykernel +compel: + version: "2.3.1" + isRoot: true +conformer: + version: "0.3.2" + isRoot: true +contourpy: + version: "1.3.3" + dependents: + - matplotlib +controlnet-aux: + version: "0.0.10" + isRoot: true +cycler: + version: "0.12.1" + dependents: + - matplotlib +cyclopts: + version: "5.3.2" + dependents: + - pyvista +debugpy: + version: "1.8.22" + dependents: + - ipykernel +decorator: + version: "5.3.1" + dependents: + - librosa +defusedxml: + version: "0.7.1" + dependents: + - nbconvert +deprecated: + version: "3.0.0" + isRoot: true +detect-installer: + version: "0.2.1" + dependents: + - fastapi-cloud-cli +diffusers: + version: "0.38.0" + isRoot: true + dependents: + - compel +distlib: + version: "0.4.4" + dependents: + - virtualenv +dnspython: + version: "2.9.1" + dependents: + - email-validator +docstring-parser: + version: "0.18.0" + dependents: + - cyclopts +easydict: + version: "1.13" + isRoot: true +einops: + version: "0.8.2" + isRoot: true + dependents: + - conformer + - controlnet-aux + - spandrel +email-validator: + version: "2.3.0" + dependents: + - fastapi + - pydantic +executing: + version: "2.3.0" + dependents: + - stack-data +fastapi: + version: "0.141.1" + isRoot: true +fastapi-cli: + version: "0.0.32" + dependents: + - fastapi +fastapi-cloud-cli: + version: "0.26.0" + dependents: + - fastapi-cli +fastar: + version: "0.12.0" + dependents: + - fastapi + - fastapi-cloud-cli +fastjsonschema: + version: "2.22.2" + dependents: + - nbformat +filelock: + version: "4.2.4" + dependents: + - controlnet-aux + - diffusers + - huggingface-hub + - python-discovery + - torch + - transformers + - virtualenv +flatbuffers: + version: "25.12.19" + dependents: + - mediapipe + - onnxruntime +fonttools: + version: "4.68.0" + dependents: + - matplotlib +fqdn: + version: "1.6.0" + dependents: + - jsonschema +frozenlist: + version: "1.8.0" + dependents: + - aiohttp + - aiosignal +fsspec: + version: "2026.9.0" + dependents: + - huggingface-hub + - torch +ftfy: + version: "6.3.1" + isRoot: true + dependents: + - clip-anytorch +gguf: + version: "0.19.0" + isRoot: true +glcontext: + version: "3.1.0" + dependents: + - moderngl +h11: + version: "0.16.0" + dependents: + - httpcore + - uvicorn +hf-transfer: + version: "0.1.9" + dependents: + - huggingface-hub +hf-xet: + version: "1.7.1" + dependents: + - huggingface-hub +httpcore: + version: "1.0.9" + dependents: + - httpx +httptools: + version: "0.9.0" + dependents: + - uvicorn +httpx: + version: "0.28.1" + dependents: + - diffusers + - fastapi + - fastapi-cloud-cli + - jupyterlab +huggingface-hub: + version: "0.36.2" + isRoot: true + dependents: + - accelerate + - controlnet-aux + - diffusers + - peft + - timm + - tokenizers + - transformers +humanfriendly: + version: "10.0" + dependents: + - coloredlogs +hyperpyyaml: + version: "1.2.2" + isRoot: true +idna: + version: "3.20" + dependents: + - anyio + - email-validator + - httpx + - jsonschema + - requests + - yarl +igraph: + version: "1.0.0" + isRoot: true +imageio: + version: "2.38.1" + isRoot: true + dependents: + - scikit-image +imageio-ffmpeg: + version: "0.7.0" + dependents: + - imageio +importlib-metadata: + version: "10.1.0" + dependents: + - controlnet-aux + - diffusers +importlib-resources: + version: "7.1.0" + dependents: + - wetextprocessing +inflect: + version: "7.6.0" + isRoot: true +ipykernel: + version: "7.4.0" + dependents: + - jupyterlab +ipython: + version: "9.17.1" + dependents: + - ipykernel +ipython-pygments-lexers: + version: "1.1.1" + dependents: + - ipython +isoduration: + version: "20.11.0" + dependents: + - jsonschema +jaraco-functools: + version: "4.6.0" + dependents: + - inflect +jax: + version: "0.4.38" + dependents: + - mediapipe +jaxlib: + version: "0.4.38" + dependents: + - jax + - mediapipe +jedi: + version: "0.20.1" + dependents: + - ipython +jinja2: + version: "3.1.6" + dependents: + - fastapi + - jupyter-server + - jupyterlab + - jupyterlab-server + - nbconvert + - torch +joblib: + version: "1.6.0" + dependents: + - librosa + - scikit-learn +json5: + version: "0.17.3" + dependents: + - jupyterlab-server +jsonpointer: + version: "3.6" + dependents: + - jsonschema +jsonschema: + version: "4.26.0" + dependents: + - jupyter-events + - jupyterlab-server + - nbformat + - rembg +jsonschema-specifications: + version: "2025.9.1" + dependents: + - jsonschema +jupyter-builder: + version: "1.2.4" + dependents: + - jupyterlab + - notebook +jupyter-client: + version: "8.10.0" + dependents: + - ipykernel + - jupyter-server + - nbclient +jupyter-core: + version: "5.9.1" + dependents: + - ipykernel + - jupyter-builder + - jupyter-client + - jupyter-server + - jupyterlab + - nbclient + - nbconvert + - nbformat +jupyter-events: + version: "0.12.1" + dependents: + - jupyter-server +jupyter-lsp: + version: "2.3.1" + dependents: + - jupyterlab +jupyter-server: + version: "2.21.1" + dependents: + - jupyter-lsp + - jupyterlab + - jupyterlab-server + - notebook + - notebook-shim +jupyter-server-terminals: + version: "0.5.4" + dependents: + - jupyter-server +jupyterlab: + version: "4.6.5" + dependents: + - notebook +jupyterlab-pygments: + version: "0.3.0" + dependents: + - nbconvert +jupyterlab-server: + version: "2.28.1" + dependents: + - jupyterlab + - notebook +kiwisolver: + version: "1.5.1" + dependents: + - matplotlib +kornia: + version: "0.8.3" + isRoot: true +kornia-rs: + version: "0.2.0" + dependents: + - kornia +lark: + version: "1.3.1" + dependents: + - rfc3987-syntax +lazy-loader: + version: "0.6" + dependents: + - librosa + - scikit-image +librosa: + version: "0.11.0" + isRoot: true +llvmlite: + version: "0.45.1" + isRoot: true + dependents: + - numba +markdown-it-py: + version: "4.2.0" + dependents: + - rich +markupsafe: + version: "3.0.4" + dependents: + - jinja2 + - nbconvert +matplotlib: + version: "3.11.2" + isRoot: true + dependents: + - mediapipe + - pyvista + - vtk +matplotlib-inline: + version: "0.2.2" + dependents: + - ipykernel + - ipython +mdurl: + version: "0.1.2" + dependents: + - markdown-it-py +mediapipe: + version: "0.10.21" + isRoot: true +mistune: + version: "3.4.0" + dependents: + - nbconvert +ml-dtypes: + version: "0.5.4" + dependents: + - jax + - jaxlib + - onnx +modelscope: + version: "1.26.0" + isRoot: true +moderngl: + version: "5.13.0" + isRoot: true +more-itertools: + version: "11.2.1" + dependents: + - inflect + - jaraco-functools + - openai-whisper +mpmath: + version: "1.3.0" + dependents: + - sympy +msgpack: + version: "1.2.4" + dependents: + - librosa +multidict: + version: "7.2.1" + dependents: + - aiohttp + - yarl +narwhals: + version: "2.28.0" + dependents: + - scikit-learn +nbclient: + version: "0.11.0" + dependents: + - nbconvert +nbconvert: + version: "7.17.2" + dependents: + - jupyter-server +nbformat: + version: "5.11.1" + dependents: + - jupyter-server + - nbclient + - nbconvert +nest-asyncio2: + version: "1.7.4" + dependents: + - ipykernel +networkx: + version: "3.7" + dependents: + - scikit-image + - torch +notebook: + version: "7.6.3" + dependents: + - compel +notebook-shim: + version: "0.2.4" + dependents: + - jupyterlab + - notebook +numba: + version: "0.62.1" + isRoot: true + dependents: + - librosa + - openai-whisper + - pymatting +numpy: + version: "1.26.4" + isRoot: true + dependents: + - accelerate + - contourpy + - controlnet-aux + - diffusers + - gguf + - imageio + - jax + - jaxlib + - librosa + - matplotlib + - mediapipe + - ml-dtypes + - numba + - onnx + - onnxruntime + - openai-whisper + - opencv-contrib-python + - opencv-python-headless + - pandas + - peft + - plyfile + - pymatting + - pymeshfix + - pyvista + - pyvista-validation + - rembg + - scikit-image + - scikit-learn + - scipy + - soundfile + - soxr + - spandrel + - tifffile + - torchsde + - torchvision + - transformers + - trimesh +omegaconf: + version: "2.3.0" + isRoot: true +onnx: + version: "1.22.0" + isRoot: true +onnxruntime: + version: "1.23.2" + isRoot: true +openai-whisper: + version: "20240930" + isRoot: true +opencv-contrib-python: + version: "4.11.0.86" + dependents: + - mediapipe +opencv-python-headless: + version: "4.11.0.86" + dependents: + - controlnet-aux + - rembg +opt-einsum: + version: "3.4.0" + dependents: + - jax +packaging: + version: "26.3" + dependents: + - accelerate + - huggingface-hub + - ipykernel + - jupyter-events + - jupyter-server + - jupyterlab + - jupyterlab-server + - kornia + - lazy-loader + - matplotlib + - nbconvert + - onnxruntime + - peft + - pooch + - scikit-image + - transformers + - virtualenv +pandas: + version: "3.0.6" + isRoot: true +pandocfilters: + version: "1.5.1" + dependents: + - nbconvert +parso: + version: "0.8.7" + dependents: + - jedi +peft: + version: "0.17.1" + isRoot: true +pexpect: + version: "4.9.0" + dependents: + - ipython +picklescan: + version: "1.0.5" + isRoot: true +pillow: + version: "13.0.0" + isRoot: true + dependents: + - controlnet-aux + - diffusers + - imageio + - matplotlib + - pymatting + - pyvista + - rembg + - scikit-image + - torchvision +platformdirs: + version: "4.14.0" + dependents: + - jupyter-core + - pooch + - virtualenv +plyfile: + version: "1.1.3" + isRoot: true +pooch: + version: "1.9.0" + dependents: + - librosa + - pyvista + - rembg +prometheus-client: + version: "0.26.0" + dependents: + - jupyter-server +prompt-toolkit: + version: "3.0.53" + dependents: + - ipython +propcache: + version: "0.5.4" + dependents: + - aiohttp + - yarl +protobuf: + version: "4.25.9" + dependents: + - mediapipe + - onnx + - onnxruntime +psutil: + version: "7.2.2" + dependents: + - accelerate + - imageio + - ipython + - peft +ptyprocess: + version: "0.7.0" + dependents: + - pexpect + - terminado +pure-eval: + version: "0.2.4" + dependents: + - stack-data +pycparser: + version: "3.12" + dependents: + - cffi +pydantic: + version: "2.14.2" + isRoot: true + dependents: + - fastapi + - fastapi-cloud-cli + - pydantic-extra-types + - pydantic-settings +pydantic-core: + version: "2.50.2" + dependents: + - pydantic +pydantic-extra-types: + version: "2.11.1" + dependents: + - fastapi +pydantic-settings: + version: "2.16.2" + isRoot: true + dependents: + - fastapi +pygments: + version: "2.21.0" + dependents: + - ipython + - ipython-pygments-lexers + - nbconvert + - rich + - rich-rst +pymatting: + version: "1.1.16" + dependents: + - rembg +pymeshfix: + version: "0.18.1" + isRoot: true +pynini: + version: "2.1.7" + isRoot: true +pyobjc-core: + version: "12.2.2" + dependents: + - pyobjc-framework-cocoa +pyobjc-framework-cocoa: + version: "12.2.2" + dependents: + - pyvista +pyparsing: + version: "3.3.3" + dependents: + - compel + - matplotlib +python-dateutil: + version: "3.0.0" + dependents: + - arrow + - jupyter-client + - matplotlib + - pandas +python-discovery: + version: "1.6.7" + dependents: + - virtualenv +python-dotenv: + version: "1.2.4" + dependents: + - pydantic-settings + - uvicorn +python-json-logger: + version: "4.2.0" + dependents: + - jupyter-events +python-multipart: + version: "0.0.32" + dependents: + - fastapi +pyvista: + version: "0.49.2" + isRoot: true +pyvista-validation: + version: "0.2.2" + dependents: + - pyvista +pyyaml: + version: "6.0.3" + isRoot: true + dependents: + - accelerate + - gguf + - huggingface-hub + - hyperpyyaml + - jupyter-events + - omegaconf + - peft + - timm + - transformers + - uvicorn +pyzmq: + version: "27.2.0" + dependents: + - ipykernel + - jupyter-client + - jupyter-server +referencing: + version: "0.37.0" + dependents: + - jsonschema + - jsonschema-specifications + - jupyter-events +regex: + version: "2026.9.29" + dependents: + - clip-anytorch + - diffusers + - tiktoken + - transformers +rembg: + version: "2.0.69" + isRoot: true +requests: + version: "2.34.2" + dependents: + - diffusers + - gguf + - huggingface-hub + - jupyterlab-server + - modelscope + - pooch + - tiktoken + - transformers +restrictedpython: + version: "8.5" + isRoot: true +rfc3339-validator: + version: "0.1.4" + dependents: + - jsonschema + - jupyter-events +rfc3986-validator: + version: "0.1.1" + dependents: + - jsonschema + - jupyter-events +rfc3987-syntax: + version: "1.1.0" + dependents: + - jsonschema +rich: + version: "15.0.0" + dependents: + - cyclopts + - rich-rst + - rich-toolkit + - typer +rich-rst: + version: "2.2.0" + dependents: + - cyclopts +rich-toolkit: + version: "0.20.6" + dependents: + - fastapi-cli + - fastapi-cloud-cli +rignore: + version: "0.8.1" + dependents: + - fastapi-cloud-cli +rpds-py: + version: "2026.9.1" + dependents: + - jsonschema + - referencing +ruamel-yaml: + version: "0.18.17" + isRoot: true + dependents: + - hyperpyyaml +ruamel-yaml-clib: + version: "0.2.15" + dependents: + - ruamel-yaml +safetensors: + version: "0.8.0" + isRoot: true + dependents: + - accelerate + - diffusers + - peft + - spandrel + - timm + - transformers +scikit-image: + version: "0.26.0" + isRoot: true + dependents: + - controlnet-aux + - rembg +scikit-learn: + version: "1.9.1" + dependents: + - librosa +scipy: + version: "1.17.1" + isRoot: true + dependents: + - controlnet-aux + - jax + - jaxlib + - librosa + - pymatting + - rembg + - scikit-image + - scikit-learn + - torchsde +scooby: + version: "0.12.1" + dependents: + - pyvista +send2trash: + version: "2.1.0" + dependents: + - jupyter-server +sentencepiece: + version: "0.2.0" + isRoot: true + dependents: + - mediapipe +sentry-sdk: + version: "2.73.0" + dependents: + - fastapi-cloud-cli +setuptools: + version: "80.9.0" + isRoot: true + dependents: + - torch +shellingham: + version: "1.5.4" + dependents: + - typer +six: + version: "1.17.0" + dependents: + - python-dateutil + - rfc3339-validator +sounddevice: + version: "0.5.6" + dependents: + - mediapipe +soundfile: + version: "0.14.0" + isRoot: true + dependents: + - librosa +soupsieve: + version: "3.0.4" + dependents: + - beautifulsoup4 +soxr: + version: "1.1.0" + dependents: + - librosa +spandrel: + version: "0.4.2" + isRoot: true +stack-data: + version: "0.6.3" + dependents: + - ipython +starlette: + version: "1.10.0" + dependents: + - fastapi +sympy: + version: "1.14.0" + dependents: + - onnxruntime + - torch +terminado: + version: "0.18.1" + dependents: + - jupyter-server + - jupyter-server-terminals +texttable: + version: "1.7.2" + dependents: + - igraph +threadpoolctl: + version: "3.8.0" + dependents: + - scikit-learn +tifffile: + version: "2026.3.3" + dependents: + - scikit-image +tiktoken: + version: "0.14.0" + isRoot: true + dependents: + - openai-whisper +timm: + version: "1.0.30" + dependents: + - controlnet-aux +tinycss2: + version: "1.5.1" + dependents: + - bleach +tokenizers: + version: "0.22.2" + isRoot: true + dependents: + - transformers +torch: + version: "2.8.0+ssui.cpu" + isRoot: true + dependents: + - accelerate + - clip-anytorch + - compel + - conformer + - controlnet-aux + - diffusers + - kornia + - openai-whisper + - peft + - spandrel + - timm + - torchaudio + - torchsde + - torchvision +torchaudio: + version: "2.8.0+ssui.cpu" + isRoot: true +torchsde: + version: "0.2.6" + isRoot: true +torchvision: + version: "0.23.0+ssui.cpu" + isRoot: true + dependents: + - clip-anytorch + - controlnet-aux + - spandrel + - timm +tornado: + version: "6.6" + dependents: + - ipykernel + - jupyter-client + - jupyter-server + - jupyterlab + - notebook + - terminado +tqdm: + version: "4.70.1" + dependents: + - clip-anytorch + - gguf + - huggingface-hub + - modelscope + - openai-whisper + - peft + - rembg + - transformers +traitlets: + version: "5.16.1" + dependents: + - ipykernel + - ipython + - jupyter-builder + - jupyter-client + - jupyter-core + - jupyter-events + - jupyter-server + - jupyterlab + - matplotlib-inline + - nbclient + - nbconvert + - nbformat +trampoline: + version: "0.1.2" + dependents: + - torchsde +transformers: + version: "4.57.6" + isRoot: true + dependents: + - compel + - peft + - transformers-stream-generator +transformers-stream-generator: + version: "0.0.4" + isRoot: true +trimesh: + version: "5.2.0" + isRoot: true +typeguard: + version: "4.6.0" + dependents: + - inflect +typer: + version: "0.27.3" + dependents: + - fastapi-cli + - fastapi-cloud-cli +typing-extensions: + version: "4.16.0" + dependents: + - aiohttp + - aiosignal + - anyio + - beautifulsoup4 + - fastapi + - huggingface-hub + - jupyter-client + - librosa + - onnx + - pydantic + - pydantic-core + - pydantic-extra-types + - pydantic-settings + - pyvista + - pyvista-validation + - referencing + - rich-toolkit + - soundfile + - spandrel + - starlette + - torch + - typeguard + - typing-inspection +typing-inspection: + version: "0.4.4" + dependents: + - fastapi + - pydantic + - pydantic-settings +tzdata: + version: "2026.6" + dependents: + - arrow + - python-dateutil +uri-template: + version: "1.3.0" + dependents: + - jsonschema +urllib3: + version: "2.8.0" + dependents: + - modelscope + - requests + - sentry-sdk +uvicorn: + version: "0.54.0" + dependents: + - fastapi + - fastapi-cli + - fastapi-cloud-cli +uvloop: + version: "0.23.0" + dependents: + - uvicorn +virtualenv: + version: "21.15.1" + isRoot: true +vtk: + version: "9.7.1" + dependents: + - pyvista +watchfiles: + version: "1.3.0" + dependents: + - uvicorn +wcwidth: + version: "0.9.4" + dependents: + - ftfy + - prompt-toolkit +webcolors: + version: "25.10.0" + dependents: + - jsonschema +webencodings: + version: "0.6.1" + dependents: + - bleach + - tinycss2 +websocket-client: + version: "1.9.2" + dependents: + - jupyter-server +websockets: + version: "17.3" + isRoot: true + dependents: + - uvicorn +wetextprocessing: + version: "1.0.4" + url: "https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl" + isRoot: true +wrapt: + version: "2.5.2" + dependents: + - deprecated +xatlas: + version: "0.0.11" + isRoot: true +yarl: + version: "1.25.1" + dependents: + - aiohttp +zipp: + version: "4.1.2" + dependents: + - importlib-metadata diff --git a/dependencies/macosx-x86_64.lock b/dependencies/macosx-x86_64.lock index 5cb2aba4..46e629ee 100644 --- a/dependencies/macosx-x86_64.lock +++ b/dependencies/macosx-x86_64.lock @@ -1,5 +1,5 @@ # This file was autogenerated by uv via the following command: -# uv pip compile dependencies/requirements-macosx-x86_64.txt --python-version 3.12.8 --python-platform x86_64-apple-darwin --index-strategy unsafe-best-match --emit-index-url -o dependencies/macosx-x86_64.lock +# uv pip compile dependencies/requirements-macosx-x86_64.txt --python-version 3.12.8 --python-platform x86_64-apple-darwin --index-strategy unsafe-best-match --build-constraints dependencies/build-constraints.txt --emit-index-url -o native-wheels/macosx-x86_64.lock --index-url https://pypi.org/simple --extra-index-url https://download.pytorch.org/whl/cpu @@ -53,6 +53,8 @@ attrs==26.1.0 # jsonschema # mediapipe # referencing +audioread==3.1.0 + # via librosa babel==2.18.0 # via jupyterlab-server beautifulsoup4==4.15.0 @@ -71,6 +73,7 @@ cffi==2.1.1 # via # argon2-cffi-bindings # sounddevice + # soundfile charset-normalizer==3.5.2 # via requests click==8.5.0 @@ -79,12 +82,16 @@ click==8.5.0 # uvicorn clip-anytorch==2.6.0 # via -r dependencies/requirements-macosx-x86_64.txt +cloudpickle==3.1.2 + # via joblib coloredlogs==15.0.1 # via onnxruntime comm==0.2.3 # via ipykernel compel==2.3.1 # via -r dependencies/requirements-macosx-x86_64.txt +conformer==0.3.2 + # via -r dependencies/requirements-macosx-x86_64.txt contourpy==1.3.3 # via matplotlib controlnet-aux==0.0.10 @@ -95,6 +102,8 @@ cyclopts==5.3.2 # via pyvista debugpy==1.8.22 # via ipykernel +decorator==5.3.1 + # via librosa defusedxml==0.7.1 # via nbconvert deprecated==3.0.0 @@ -116,6 +125,7 @@ easydict==1.13 einops==0.8.2 # via # -r dependencies/requirements-macosx-x86_64.txt + # conformer # controlnet-aux # spandrel email-validator==2.3.0 @@ -199,6 +209,8 @@ huggingface-hub==0.36.2 # transformers humanfriendly==10.0 # via coloredlogs +hyperpyyaml==1.2.2 + # via -r dependencies/requirements-macosx-x86_64.txt idna==3.20 # via # anyio @@ -219,6 +231,10 @@ importlib-metadata==10.1.0 # via # controlnet-aux # diffusers +importlib-resources==7.1.0 + # via wetextprocessing +inflect==7.6.0 + # via -r dependencies/requirements-macosx-x86_64.txt ipykernel==7.4.0 # via jupyterlab ipython==9.17.1 @@ -227,6 +243,8 @@ ipython-pygments-lexers==1.1.1 # via ipython isoduration==20.11.0 # via jsonschema +jaraco-functools==4.6.0 + # via inflect jax==0.4.38 # via mediapipe jaxlib==0.4.38 @@ -243,6 +261,10 @@ jinja2==3.1.6 # jupyterlab-server # nbconvert # torch +joblib==1.6.0 + # via + # librosa + # scikit-learn json5==0.17.3 # via jupyterlab-server jsonpointer==3.6 @@ -304,7 +326,11 @@ kornia-rs==0.2.0 lark==1.3.1 # via rfc3987-syntax lazy-loader==0.6 - # via scikit-image + # via + # librosa + # scikit-image +librosa==0.11.0 + # via -r dependencies/requirements-macosx-x86_64.txt llvmlite==0.45.1 # via # -r dependencies/requirements-macosx-x86_64.txt @@ -340,12 +366,21 @@ modelscope==1.26.0 # via -r dependencies/requirements-macosx-x86_64.txt moderngl==5.13.0 # via -r dependencies/requirements-macosx-x86_64.txt +more-itertools==11.2.1 + # via + # inflect + # jaraco-functools + # openai-whisper mpmath==1.3.0 # via sympy +msgpack==1.2.4 + # via librosa multidict==7.2.1 # via # aiohttp # yarl +narwhals==2.28.0 + # via scikit-learn nbclient==0.11.0 # via nbconvert nbconvert==7.17.2 @@ -370,6 +405,8 @@ notebook-shim==0.2.4 numba==0.62.1 # via # -r dependencies/requirements-macosx-x86_64.txt + # librosa + # openai-whisper # pymatting numpy==1.26.4 # via @@ -382,12 +419,14 @@ numpy==1.26.4 # imageio # jax # jaxlib + # librosa # matplotlib # mediapipe # ml-dtypes # numba # onnx # onnxruntime + # openai-whisper # opencv-contrib-python # opencv-python-headless # pandas @@ -399,7 +438,10 @@ numpy==1.26.4 # pyvista-validation # rembg # scikit-image + # scikit-learn # scipy + # soundfile + # soxr # spandrel # tifffile # torchsde @@ -412,6 +454,8 @@ onnx==1.22.0 # via -r dependencies/requirements-macosx-x86_64.txt onnxruntime==1.23.2 # via -r dependencies/requirements-macosx-x86_64.txt +openai-whisper==20240930 + # via -r dependencies/requirements-macosx-x86_64.txt opencv-contrib-python==4.11.0.86 # via mediapipe opencv-python-headless==4.11.0.86 @@ -472,6 +516,7 @@ plyfile==1.1.3 # via -r dependencies/requirements-macosx-x86_64.txt pooch==1.9.0 # via + # librosa # pyvista # rembg prometheus-client==0.26.0 @@ -527,6 +572,8 @@ pymatting==1.1.16 # via rembg pymeshfix==0.18.1 # via -r dependencies/requirements-macosx-x86_64.txt +pynini==2.1.7 + # via -r dependencies/requirements-macosx-x86_64.txt pyobjc-core==12.2.2 # via pyobjc-framework-cocoa pyobjc-framework-cocoa==12.2.2 @@ -561,6 +608,7 @@ pyyaml==6.0.3 # accelerate # gguf # huggingface-hub + # hyperpyyaml # jupyter-events # omegaconf # peft @@ -625,6 +673,12 @@ rpds-py==2026.9.1 # via # jsonschema # referencing +ruamel-yaml==0.18.17 + # via + # -r dependencies/requirements-macosx-x86_64.txt + # hyperpyyaml +ruamel-yaml-clib==0.2.15 + # via ruamel-yaml safetensors==0.8.0 # via # -r dependencies/requirements-macosx-x86_64.txt @@ -639,15 +693,19 @@ scikit-image==0.26.0 # -r dependencies/requirements-macosx-x86_64.txt # controlnet-aux # rembg +scikit-learn==1.9.1 + # via librosa scipy==1.17.1 # via # -r dependencies/requirements-macosx-x86_64.txt # controlnet-aux # jax # jaxlib + # librosa # pymatting # rembg # scikit-image + # scikit-learn # torchsde scooby==0.12.1 # via pyvista @@ -659,12 +717,12 @@ sentencepiece==0.2.0 # mediapipe sentry-sdk==2.73.0 # via fastapi-cloud-cli -shellingham==1.5.4 - # via typer setuptools==80.9.0 # via # -r dependencies/requirements-macosx-x86_64.txt # torch +shellingham==1.5.4 + # via typer six==1.17.0 # via # python-dateutil @@ -672,9 +730,13 @@ six==1.17.0 sounddevice==0.5.6 # via mediapipe soundfile==0.14.0 - # via -r dependencies/requirements-macosx-x86_64.txt + # via + # -r dependencies/requirements-macosx-x86_64.txt + # librosa soupsieve==3.0.4 # via beautifulsoup4 +soxr==1.1.0 + # via librosa spandrel==0.4.2 # via -r dependencies/requirements-macosx-x86_64.txt stack-data==0.6.3 @@ -691,10 +753,14 @@ terminado==0.18.1 # jupyter-server-terminals texttable==1.7.2 # via igraph +threadpoolctl==3.8.0 + # via scikit-learn tifffile==2026.3.3 # via scikit-image tiktoken==0.14.0 - # via -r dependencies/requirements-macosx-x86_64.txt + # via + # -r dependencies/requirements-macosx-x86_64.txt + # openai-whisper timm==1.0.30 # via controlnet-aux tinycss2==1.5.1 @@ -709,9 +775,11 @@ torch==2.8.0+ssui.cpu # accelerate # clip-anytorch # compel + # conformer # controlnet-aux # diffusers # kornia + # openai-whisper # peft # spandrel # timm @@ -743,6 +811,7 @@ tqdm==4.70.1 # gguf # huggingface-hub # modelscope + # openai-whisper # peft # rembg # transformers @@ -772,6 +841,8 @@ transformers-stream-generator==0.0.4 # via -r dependencies/requirements-macosx-x86_64.txt trimesh==5.2.0 # via -r dependencies/requirements-macosx-x86_64.txt +typeguard==4.6.0 + # via inflect typer==0.27.3 # via # fastapi-cli @@ -785,6 +856,7 @@ typing-extensions==4.16.0 # fastapi # huggingface-hub # jupyter-client + # librosa # onnx # pydantic # pydantic-core @@ -794,9 +866,11 @@ typing-extensions==4.16.0 # pyvista-validation # referencing # rich-toolkit + # soundfile # spandrel # starlette # torch + # typeguard # typing-inspection typing-inspection==0.4.4 # via @@ -843,6 +917,8 @@ websockets==17.3 # via # -r dependencies/requirements-macosx-x86_64.txt # uvicorn +wetextprocessing @ https://github.com/sunxfancy/pynini-2.1.6.post1/releases/download/2.1.6.post1/WeTextProcessing-1.0.4-py3-none-any.whl + # via -r dependencies/requirements-macosx-x86_64.txt wrapt==2.5.2 # via deprecated xatlas==0.0.11 diff --git a/extensions/Voice/README.md b/extensions/Voice/README.md index 65e0066a..3c4a516c 100644 --- a/extensions/Voice/README.md +++ b/extensions/Voice/README.md @@ -83,9 +83,9 @@ yarn test_on model.cosyvoice_model_test `SSUI_COSYVOICE_MODEL_DIR` 使用已有权重,`SSUI_COSYVOICE_PROMPT` 复用官方参考音频。 自定义参考转写请直接使用探针的 `--prompt-text`。 -本次已同步除 Intel macOS 外的平台锁文件。Intel macOS 仍需先取得项目要求的已验证 -原生 PyTorch wheelhouse,并设置 `SSUI_NATIVE_WHEELHOUSE` 后运行 -`yarn update-lock:no-upgrade --target macosx-x86_64`;当前 Windows 验证环境没有该产物。 +Intel macOS 锁文件已使用经过源码版本、架构和哈希验证的原生 PyTorch wheelhouse 完整解析。 +后续刷新时,先取得项目要求的已验证 wheelhouse,设置 `SSUI_NATIVE_WHEELHOUSE`,再运行 +`yarn update-lock:no-upgrade --target macosx-x86_64`。 上游:[CosyVoice](https://github.com/FunAudioLLM/CosyVoice)、 [CosyVoice2 权重](https://huggingface.co/FunAudioLLM/CosyVoice2-0.5B)。 From 8ac0ae9b2533086233b1d03e29c25b8e536c0c3a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E8=A5=BF=E9=A3=8E=E9=80=8D=E9=81=A5=E6=B8=B8?= Date: Fri, 13 Nov 2026 13:34:43 +0800 Subject: [PATCH 4/5] fix(deps): build pinned OpenFst GRM on Linux ARM64 from verified mirror --- .github/workflows/ci.yml | 12 ++- .github/workflows/release.yml | 10 +- .github/workflows/windows-arm64-wheels.yml | 8 +- dependencies/install_requirements.cjs | 2 +- dependencies/native_prerequisites.cjs | 29 ++++++ dependencies/prepare_openfst.py | 108 +++++++++++++++++++++ extensions/Voice/README.md | 10 ++ scripts/check_runtime.py | 2 +- tests/openfst_prerequisites_test.py | 54 +++++++++++ 9 files changed, 220 insertions(+), 15 deletions(-) create mode 100644 dependencies/native_prerequisites.cjs create mode 100644 dependencies/prepare_openfst.py create mode 100644 tests/openfst_prerequisites_test.py diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index eff1dd1a..536315ce 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -63,7 +63,9 @@ jobs: - name: Frontend unit tests run: yarn test:frontend - name: Install Chromium - run: yarn test:ui:install --with-deps + run: | + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update + yarn test:ui:install --with-deps - name: Browser interaction tests run: yarn test:ui && yarn test:desktop-ui - name: Upload browser evidence @@ -108,7 +110,7 @@ jobs: - name: Native Linux libraries if: runner.os == 'Linux' run: | - sudo apt-get update + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev libdbus-1-dev xvfb - name: Setup Rust uses: actions-rust-lang/setup-rust-toolchain@166cdcfd11aee3cb47222f9ddb555ce30ddb9659 @@ -202,7 +204,9 @@ jobs: - name: Install Chromium if: github.event_name != 'schedule' timeout-minutes: 10 - run: yarn playwright install --with-deps chromium + run: | + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update + yarn playwright install --with-deps chromium - name: Install cross-browser engines if: github.event_name == 'schedule' @@ -342,7 +346,7 @@ jobs: if: ${{ matrix.platform == 'linux' }} timeout-minutes: 30 run: | - sudo apt-get -o Acquire::Retries=3 update + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update sudo apt-get -o Acquire::Retries=3 install -y --no-install-recommends libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf libglib2.0-dev pkg-config gstreamer1.0-plugins-base gstreamer1.0-plugins-good gstreamer1.0-libav - name: Restore Rust compilation cache diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml index 9b59554f..51843126 100644 --- a/.github/workflows/release.yml +++ b/.github/workflows/release.yml @@ -169,7 +169,7 @@ jobs: - name: Install Linux build and runtime dependencies if: runner.os == 'Linux' run: | - sudo apt-get update + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf libglib2.0-dev pkg-config xvfb libgl1-mesa-dev libx11-dev gstreamer1.0-plugins-base gstreamer1.0-plugins-good gstreamer1.0-libav - name: Download native Intel macOS wheels @@ -185,10 +185,10 @@ jobs: cp "$RUNNER_TEMP/native-wheels/macosx-x86_64.lock" dependencies/macosx-x86_64.lock echo "SSUI_NATIVE_WHEELHOUSE=$RUNNER_TEMP/native-wheels" >> "$GITHUB_ENV" - - name: Prepare macOS OpenFst GRM prerequisites - if: runner.os == 'macOS' + - name: Prepare native OpenFst GRM prerequisites + if: runner.os == 'macOS' || matrix.id == 'linux-arm64' timeout-minutes: 15 - run: node dependencies/macos_prerequisites.cjs --github-env + run: node dependencies/native_prerequisites.cjs --github-env - name: Install requirements timeout-minutes: 30 @@ -366,7 +366,7 @@ jobs: - name: Prepare Linux display if: runner.os == 'Linux' run: | - sudo apt-get update + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update sudo apt-get install -y xvfb libwebkit2gtk-4.1-0 libayatana-appindicator3-1 librsvg2-2 gstreamer1.0-plugins-good gstreamer1.0-plugins-bad gstreamer1.0-libav - name: Smoke test installed app (NSIS) if: runner.os == 'Windows' diff --git a/.github/workflows/windows-arm64-wheels.yml b/.github/workflows/windows-arm64-wheels.yml index 9679d1ae..511c04d2 100644 --- a/.github/workflows/windows-arm64-wheels.yml +++ b/.github/workflows/windows-arm64-wheels.yml @@ -111,7 +111,7 @@ jobs: - name: Install Linux native build dependencies if: runner.os == 'Linux' run: | - sudo apt-get update + sudo apt-get -o APT::Update::Error-Mode=any -o Acquire::Retries=3 update sudo apt-get install -y libgl1-mesa-dev libx11-dev - name: Download verified Intel wheels if: matrix.id == 'macos-x86_64' @@ -124,10 +124,10 @@ jobs: run: | node dependencies/native_wheels.cjs native-wheels cp native-wheels/macosx-x86_64.lock dependencies/macosx-x86_64.lock - - name: Prepare macOS OpenFst GRM prerequisites - if: runner.os == 'macOS' + - name: Prepare native OpenFst GRM prerequisites + if: runner.os == 'macOS' || matrix.id == 'linux-arm64' timeout-minutes: 15 - run: node dependencies/macos_prerequisites.cjs --github-env + run: node dependencies/native_prerequisites.cjs --github-env - name: Install the full native dependency lock timeout-minutes: 25 env: diff --git a/dependencies/install_requirements.cjs b/dependencies/install_requirements.cjs index e07546b6..0fa20969 100644 --- a/dependencies/install_requirements.cjs +++ b/dependencies/install_requirements.cjs @@ -17,7 +17,7 @@ function getRequirementsFile() { } try { - devEnv = require('./macos_prerequisites.cjs').prepare({ env: devEnv }); + devEnv = require('./native_prerequisites.cjs').prepare({ env: devEnv }); const requirementsFile = getRequirementsFile(); const pythonRuntime = JSON.parse(execFileSync(executable('python', devEnv), [ '-c', 'import json, platform, sys; print(json.dumps({"os": sys.platform, "arch": platform.machine()}))' diff --git a/dependencies/native_prerequisites.cjs b/dependencies/native_prerequisites.cjs new file mode 100644 index 00000000..eb771c6c --- /dev/null +++ b/dependencies/native_prerequisites.cjs @@ -0,0 +1,29 @@ +// Shared root/CI entry point; keep native prerequisites outside Python's +// isolated package build environments and preserve their locked versions. +const fs = require('node:fs'); +const path = require('node:path'); +const { execFileSync } = require('node:child_process'); +const { resolveConfig, executable } = require('./dev_config.cjs'); + +function prepare({ env = process.env, platform = process.platform, arch = process.arch } = {}) { + if (platform === 'darwin') return require('./macos_prerequisites.cjs').prepare({ env }); + if (platform !== 'linux' || arch !== 'arm64') return { ...env }; + const prefix = path.resolve(__dirname, '../.build/native/openfst-1.8.4'); + const result = execFileSync(executable('python', env), [ + path.join(__dirname, 'prepare_openfst.py'), '--prefix', prefix, + ], { encoding: 'utf8', env, stdio: ['ignore', 'pipe', 'inherit'] }); + return { ...env, ...JSON.parse(result) }; +} + +if (require.main === module) { + const env = prepare({ env: resolveConfig() }); + if (process.argv.includes('--github-env')) { + for (const name of ['CPLUS_INCLUDE_PATH', 'LIBRARY_PATH', 'LD_LIBRARY_PATH', 'LDFLAGS']) { + if (env[name] !== undefined) { + if (/[\r\n]/.test(env[name])) throw new Error(`Invalid newline in ${name}`); + fs.appendFileSync(process.env.GITHUB_ENV, `${name}=${env[name]}\n`); + } + } + } +} +module.exports = { prepare }; diff --git a/dependencies/prepare_openfst.py b/dependencies/prepare_openfst.py new file mode 100644 index 00000000..60a13e0b --- /dev/null +++ b/dependencies/prepare_openfst.py @@ -0,0 +1,108 @@ +"""Build the pinned OpenFst GRM SDK needed by Pynini on Linux ARM64.""" +import argparse +import hashlib +import json +import os +from pathlib import Path +import platform +import shutil +import subprocess +import sys +import tarfile +import tempfile +import urllib.request +from urllib.parse import urlparse + +VERSION = "1.8.4" +# MacPorts' public distfile is byte-identical to the upstream release. Keep +# the original SHA-256; upstream's download endpoint currently returns 403. +SOURCE_URL = f"https://distfiles-origin.macports.org/openfst/openfst-{VERSION}.tar.gz" +SOURCE_SHA256 = "a8ebbb6f3d92d07e671500587472518cfc87cb79b9a654a5a8abb2d0eb298016" +CONFIGURE = ["--enable-grm", "--enable-shared", "--disable-static", "--disable-bin"] +REQUIRED = ["include/fst/util.h", "include/fst/extensions/far/far.h", + "lib/libfst.so", "lib/libfstscript.so", "lib/libfstfar.so", + "lib/libfstfarscript.so", "lib/libfstpdtscript.so", "lib/libfstmpdtscript.so"] + + +def verify_archive(archive): + digest = hashlib.sha256(Path(archive).read_bytes()).hexdigest() + if digest != SOURCE_SHA256: + raise RuntimeError(f"OpenFst {VERSION} source SHA-256 mismatch: {digest}") + + +def fetch_archive(url, dest): + if urlparse(url).scheme != "https": + raise ValueError("OpenFst source must use HTTPS") + # No insecure TLS context, alternate origin, proxy or retry on refusal. + class HttpsRedirect(urllib.request.HTTPRedirectHandler): + def redirect_request(self, req, fp, code, msg, headers, newurl): + if urlparse(newurl).scheme != "https": + raise RuntimeError("Refusing a non-HTTPS source redirect") + return super().redirect_request(req, fp, code, msg, headers, newurl) + opener = urllib.request.build_opener(HttpsRedirect()) + with opener.open(url, timeout=120) as response: + if urlparse(response.geturl()).scheme != "https": + raise RuntimeError("Refusing a non-HTTPS source redirect") + with Path(dest).open("wb") as output: + shutil.copyfileobj(response, output) + verify_archive(dest) + + +def build(prefix, archive=None, source_url=SOURCE_URL, jobs=2): + prefix = Path(prefix).resolve() + recipe = dict(version=VERSION, sha256=SOURCE_SHA256, configure=CONFIGURE, + machine=platform.machine(), cxxflags="-O2 -g0 -std=c++17") + stamp = prefix / "ssui-recipe.json" + if (stamp.is_file() and json.loads(stamp.read_text()) == recipe + and all((prefix / name).is_file() for name in REQUIRED)): + return + for command in ("c++", "make"): + if not shutil.which(command): + raise RuntimeError(f"OpenFst requires {command}; install the native C++ build tools") + prefix.parent.mkdir(parents=True, exist_ok=True) + with tempfile.TemporaryDirectory(prefix="openfst-build-", dir=prefix.parent) as work: + work = Path(work) + source_archive = Path(archive).resolve() if archive else work / "source.tar.gz" + if archive: + verify_archive(source_archive) + else: + fetch_archive(source_url, source_archive) + with tarfile.open(source_archive) as source: + source.extractall(work, filter="data") + source_dir = work / f"openfst-{VERSION}" + env = dict(os.environ) + env["CXXFLAGS"] = recipe["cxxflags"] + for args in (["./configure", f"--prefix={prefix}", *CONFIGURE], + ["make", f"-j{jobs}"], ["make", "install"]): + subprocess.run(args, cwd=source_dir, env=env, check=True, + stdout=sys.stderr, stderr=sys.stderr) + missing = [name for name in REQUIRED if not (prefix / name).is_file()] + if missing: + raise RuntimeError(f"OpenFst GRM SDK incomplete: {missing}") + stamp.write_text(json.dumps(recipe, indent=2) + "\n") + + +def main(): + p = argparse.ArgumentParser(description=__doc__) + p.add_argument("--prefix", type=Path, required=True) + p.add_argument("--archive", type=Path, default=os.environ.get("SSUI_OPENFST_ARCHIVE")) + p.add_argument("--jobs", type=int, default=2) + args = p.parse_args() + if sys.platform != "linux" or platform.machine() not in ("aarch64", "arm64"): + p.error("This source prerequisite is for Linux ARM64; other targets retain their native packages") + if args.jobs < 1: + p.error("--jobs must be positive") + build(args.prefix, args.archive, jobs=args.jobs) + prefix = args.prefix.resolve() + def prepend(name, value, separator): + return value + (separator + os.environ[name] if os.environ.get(name) else "") + print(json.dumps({ + "CPLUS_INCLUDE_PATH": prepend("CPLUS_INCLUDE_PATH", str(prefix / "include"), ":"), + "LIBRARY_PATH": prepend("LIBRARY_PATH", str(prefix / "lib"), ":"), + "LD_LIBRARY_PATH": prepend("LD_LIBRARY_PATH", str(prefix / "lib"), ":"), + "LDFLAGS": prepend("LDFLAGS", f"-L{prefix / 'lib'} -Wl,-rpath,{prefix / 'lib'}", " "), + })) + + +if __name__ == "__main__": + main() diff --git a/extensions/Voice/README.md b/extensions/Voice/README.md index 3c4a516c..b5a17d04 100644 --- a/extensions/Voice/README.md +++ b/extensions/Voice/README.md @@ -10,6 +10,16 @@ macOS 上 Pynini 2.1.7 没有 PyPI wheel,需要 OpenFst 1.8.4 的 GRM 头文 `dependencies/build-constraints.txt` 的隔离构建约束,避免 Whisper 构建时丢失 `pkg_resources`。 Intel CI 使用验证后的原生 PyTorch wheel 解析完整锁,并将该锁交给后续完整安装与运行检查。 +Linux ARM64 同样需要源码编译 Pynini。根目录安装器与 ARM64 验证流程通过 +`dependencies/prepare_openfst.py` 准备 OpenFst 1.8.4 GRM,固定源码 SHA-256,SDK 安装到 +`.build/native/openfst-1.8.4`,并为完整依赖安装提供头文件、链接和运行时库路径。 +源码使用 [MacPorts 官方 distfile 镜像](https://distfiles-origin.macports.org/openfst/openfst-1.8.4.tar.gz), +与上游原归档逐字节一致,保留原 SHA-256 +`a8ebbb6f3d92d07e671500587472518cfc87cb79b9a654a5a8abb2d0eb298016`。 +可用 `SSUI_OPENFST_ARCHIVE` 指向获准取得的本地源码归档,仍必须通过相同 SHA-256 校验。 +源码下载拒绝访问或校验不符会直接失败,不回退到 HTTP、其他来源或不校验的归档。 +Ubuntu CI 的 apt 更新在任一软件源失败时停止,不能以旧索引继续安装和掩盖网络故障。 + ## 从根目录运行 先按 SSUI 的标准流程安装对应运行时。Windows AMD 使用项目的 Windows x64 ROCm diff --git a/scripts/check_runtime.py b/scripts/check_runtime.py index dbafb35a..f6ebeceb 100644 --- a/scripts/check_runtime.py +++ b/scripts/check_runtime.py @@ -24,7 +24,7 @@ def main(): assert torch.tensor([1.0, 2.0]).sum().item() == 3 boxes = torch.tensor([[0., 0., 1., 1.], [0., 0., 1., 1.]]) assert torchvision.ops.nms(boxes, torch.tensor([0.9, 0.8]), 0.5).tolist() == [0] - if a.target.startswith('macos-'): + if a.target.startswith('macos-') or a.target == 'linux-arm64': # Exercise the native OpenFst binding after the full lock installation, # rather than treating a successful sdist metadata resolution as proof. import pynini diff --git a/tests/openfst_prerequisites_test.py b/tests/openfst_prerequisites_test.py new file mode 100644 index 00000000..2b20b7f3 --- /dev/null +++ b/tests/openfst_prerequisites_test.py @@ -0,0 +1,54 @@ +"""Guard the source SDK's integrity boundary without downloading/building it.""" +import importlib.util +from pathlib import Path +import tempfile +import unittest +import urllib.error +from unittest.mock import Mock, patch + +ROOT = Path(__file__).resolve().parents[1] +spec = importlib.util.spec_from_file_location("openfst_prerequisites", ROOT / "dependencies/prepare_openfst.py") +openfst = importlib.util.module_from_spec(spec) +spec.loader.exec_module(openfst) + + +class TestOpenFstSourceIntegrity(unittest.TestCase): + def test_checksum_failure_stops_before_extracting_or_building(self): + # This boundary test needs neither native tools nor filesystem access. + with patch.object(tempfile, "TemporaryDirectory") as work, \ + patch.object(Path, "read_bytes", return_value=b"not the pinned source"), \ + patch.object(Path, "is_file", return_value=False), \ + patch.object(Path, "mkdir"), \ + patch.object(openfst.platform, "machine", return_value="aarch64"), \ + patch.object(openfst.shutil, "which", return_value="/usr/bin/c++"), \ + patch.object(openfst.tarfile, "open") as extract, \ + patch.object(openfst.subprocess, "run") as run: + work.return_value.__enter__.return_value = str(ROOT / "unused-build-context") + with self.assertRaisesRegex(RuntimeError, "SHA-256 mismatch"): + openfst.build(ROOT / "unused-sdk", archive=ROOT / "unused-source.tar.gz") + extract.assert_not_called() + run.assert_not_called() + + def test_plain_http_is_rejected_without_network_access(self): + with patch.object(openfst.urllib.request, "build_opener") as network: + with self.assertRaisesRegex(ValueError, "HTTPS"): + openfst.fetch_archive("http://example.com/source.tar.gz", Path("unused")) + network.assert_not_called() + + def test_access_refusal_propagates_without_retry_or_fallback(self): + network = Mock() + network.open.side_effect = urllib.error.HTTPError(openfst.SOURCE_URL, 403, "Forbidden", None, None) + with patch.object(openfst.urllib.request, "build_opener", return_value=network): + with self.assertRaises(urllib.error.HTTPError): + openfst.fetch_archive(openfst.SOURCE_URL, Path("unused")) + network.open.assert_called_once_with(openfst.SOURCE_URL, timeout=120) + + def test_insecure_redirect_is_rejected_before_following(self): + network = Mock() + network.open.side_effect = urllib.error.HTTPError(openfst.SOURCE_URL, 403, "Forbidden", None, None) + with patch.object(openfst.urllib.request, "build_opener", return_value=network) as create: + with self.assertRaises(urllib.error.HTTPError): + openfst.fetch_archive(openfst.SOURCE_URL, Path("unused")) + handler = create.call_args.args[0] + with self.assertRaisesRegex(RuntimeError, "non-HTTPS"): + handler.redirect_request(None, None, 302, "Redirect", {}, "http://example.com/source.tar.gz") From 96911f679a42cef949462f7a4265797a9e5ac6c2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E8=A5=BF=E9=A3=8E=E9=80=8D=E9=81=A5=E6=B8=B8?= Date: Fri, 13 Nov 2026 13:35:51 +0800 Subject: [PATCH 5/5] fix(deps): retain OpenFst script libraries required by Pynini --- dependencies/prepare_openfst.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/dependencies/prepare_openfst.py b/dependencies/prepare_openfst.py index 60a13e0b..afb0dea7 100644 --- a/dependencies/prepare_openfst.py +++ b/dependencies/prepare_openfst.py @@ -18,7 +18,9 @@ # the original SHA-256; upstream's download endpoint currently returns 403. SOURCE_URL = f"https://distfiles-origin.macports.org/openfst/openfst-{VERSION}.tar.gz" SOURCE_SHA256 = "a8ebbb6f3d92d07e671500587472518cfc87cb79b9a654a5a8abb2d0eb298016" -CONFIGURE = ["--enable-grm", "--enable-shared", "--disable-static", "--disable-bin"] +# OpenFst also gates its script libraries behind --enable-bin. Pynini links +# libfstscript and the GRM script libraries, so these must remain enabled. +CONFIGURE = ["--enable-grm", "--enable-shared", "--disable-static", "--enable-bin"] REQUIRED = ["include/fst/util.h", "include/fst/extensions/far/far.h", "lib/libfst.so", "lib/libfstscript.so", "lib/libfstfar.so", "lib/libfstfarscript.so", "lib/libfstpdtscript.so", "lib/libfstmpdtscript.so"]