fix: lock PyTorch CUDA builds for transcription

SKILL.md

@@ -291,7 +291,7 @@ Use whatever vision capability the agent's runtime provides to analyze extracted
 
 - **Which model** to use (depends on the agent's runtime and available vision capabilities).
 - **Batch size** — depends on the model's context window. See `references/frame-extraction-guide.md` for token cost estimates at different resolutions.
-- **Windows setup troubleshooting** — see `references/windows-funasr-uv-setup.md` for Windows-host PowerShell transcription commands and uv/Python wheel issues such as `editdistance` source builds.
+- **Windows setup troubleshooting** — see `references/windows-funasr-uv-setup.md` for Windows-host PowerShell transcription commands, Python wheel issues such as `editdistance` source builds, and PyTorch CUDA-vs-CPU build checks.
 - **How many frames** per analysis pass — use the binary-search approach from Step 6.
 
 **Completion criteria:** the agent has enough visual understanding to answer the user's question or produce the requested edit plan.

references/windows-funasr-uv-setup.md

@@ -36,6 +36,30 @@ Get-Content $Log -Tail 160
 
 From WSL, wrap this in `powershell.exe -NoProfile -ExecutionPolicy Bypass -Command '...'`.
 
+## PyTorch CUDA build pitfall
+
+If `funasr-nano` logs `on cpu` on a Windows NVIDIA machine, check PyTorch:
+
+```powershell
+$TransDir = "$env:USERPROFILE\video-edit-planner-skill\scripts\transcription"
+Set-Location $TransDir
+uv run python -c "import torch; print(torch.__version__); print(torch.version.cuda); print(torch.cuda.is_available()); print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'no cuda')"
+```
+
+If this prints `+cpu`, `None`, and `False`, the environment has CPU-only PyTorch. Do **not** fix this with a one-off `uv pip install --torch-backend=auto` unless the project lock/config is also updated: `uv run` will sync from `uv.lock` and may immediately uninstall the CUDA build and reinstall the CPU build.
+
+This project configures `torch` and `torchaudio` to use the PyTorch `cu130` index on Linux/Windows via `[tool.uv.sources]`. After pulling a version that includes that configuration, rebuild the environment with:
+
+```powershell
+$TransDir = "$env:USERPROFILE\video-edit-planner-skill\scripts\transcription"
+Set-Location $TransDir
+git pull
+uv sync --reinstall
+uv run python -c "import torch; print(torch.__version__); print(torch.version.cuda); print(torch.cuda.is_available()); print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'no cuda')"
+```
+
+Expected on an NVIDIA Windows host: `torch` version contains `+cu130`, `torch.version.cuda` is not `None`, and `torch.cuda.is_available()` is `True`.
+
 ## `editdistance` / Python version pitfall
 
 Observed failure:

scripts/transcription/pyproject.toml

@@ -17,6 +17,19 @@ funasr-nano = "funasr_nano:main"
 funasr-fast = "funasr_fast:main"
 funasr-regular = "funasr_regular:main"
 
+[[tool.uv.index]]
+name = "pytorch-cu130"
+url = "https://download.pytorch.org/whl/cu130"
+explicit = true
+
+[tool.uv.sources]
+torch = [
+    { index = "pytorch-cu130", marker = "sys_platform == 'linux' or sys_platform == 'win32'" },
+]
+torchaudio = [
+    { index = "pytorch-cu130", marker = "sys_platform == 'linux' or sys_platform == 'win32'" },
+]
+
 [build-system]
 requires = ["setuptools>=69"]
 build-backend = "setuptools.build_meta"

scripts/transcription/uv.lock

@@ -1,6 +1,10 @@
 version = 1
 revision = 3
 requires-python = "==3.12.*"
+resolution-markers = [
+    "sys_platform == 'linux' or sys_platform == 'win32'",
+    "sys_platform != 'linux' and sys_platform != 'win32'",
+]
 
 [[package]]
 name = "aliyun-python-sdk-core"
@@ -336,16 +340,20 @@ source = { editable = "." }
 dependencies = [
     { name = "funasr" },
     { name = "soundfile" },
-    { name = "torch" },
+    { name = "torch", version = "2.13.0", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform != 'linux' and sys_platform != 'win32'" },
-    { name = "torchaudio" },
+    { name = "torch", version = "2.13.0+cu130", source = { registry = "https://download.pytorch.org/whl/cu130" }, marker = "sys_platform == 'linux' or sys_platform == 'win32'" },
+    { name = "torchaudio", version = "2.11.0", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform != 'linux' and sys_platform != 'win32'" },
+    { name = "torchaudio", version = "2.11.0+cu130", source = { registry = "https://download.pytorch.org/whl/cu130" }, marker = "sys_platform == 'linux' or sys_platform == 'win32'" },
 ]
 
 [package.metadata]
 requires-dist = [
     { name = "funasr", specifier = ">=1.3.14" },
     { name = "soundfile", specifier = ">=0.13.1" },
-    { name = "torch", specifier = ">=2.13.0" },
+    { name = "torch", marker = "sys_platform != 'linux' and sys_platform != 'win32'", specifier = ">=2.13.0" },
-    { name = "torchaudio", specifier = ">=2.11.0" },
+    { name = "torch", marker = "sys_platform == 'linux' or sys_platform == 'win32'", specifier = ">=2.13.0", index = "https://download.pytorch.org/whl/cu130" },
+    { name = "torchaudio", marker = "sys_platform != 'linux' and sys_platform != 'win32'", specifier = ">=2.11.0" },
+    { name = "torchaudio", marker = "sys_platform == 'linux' or sys_platform == 'win32'", specifier = ">=2.11.0", index = "https://download.pytorch.org/whl/cu130" },
 ]
 
 [[package]]
@@ -1291,27 +1299,49 @@ wheels = [
 name = "torch"
 version = "2.13.0"
 source = { registry = "https://pypi.org/simple" }
+resolution-markers = [
+    "sys_platform != 'linux' and sys_platform != 'win32'",
+]
 dependencies = [
-    { name = "cuda-bindings", marker = "sys_platform == 'linux'" },
-    { name = "cuda-toolkit", extra = ["cublas", "cudart", "cufft", "cufile", "cupti", "curand", "cusolver", "cusparse", "nvjitlink", "nvrtc", "nvtx"], marker = "sys_platform == 'linux'" },
     { name = "filelock" },
     { name = "fsspec" },
     { name = "jinja2" },
     { name = "networkx" },
-    { name = "nvidia-cudnn-cu13", marker = "sys_platform == 'linux'" },
-    { name = "nvidia-cusparselt-cu13", marker = "sys_platform == 'linux'" },
-    { name = "nvidia-nccl-cu13", marker = "sys_platform == 'linux'" },
-    { name = "nvidia-nvshmem-cu13", marker = "sys_platform == 'linux'" },
     { name = "setuptools" },
     { name = "sympy" },
-    { name = "triton", marker = "sys_platform == 'linux'" },
     { name = "typing-extensions" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/c4/3a/ed0f4d4d1dcde03bced7aac9a28e800abcdc0cbd06b6775044c9fbd877b7/torch-2.13.0-cp312-cp312-macosx_14_0_arm64.whl", hash = "sha256:2fe228aba290d14b9f31b049be550dbd469c3fd3013d7a19705b30454da97027", size = 111213045, upload-time = "2026-07-08T16:05:22.997Z" },
-    { url = "https://files.pythonhosted.org/packages/df/a9/f6a2a4d763ff1df02e9a64c477029db614295bc9367f4131223791ccc243/torch-2.13.0-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:572df8be8ffb4599c88cbd6a0726f1f854f4da65d2e3c09f0e2c2283333cd6d4", size = 427210998, upload-time = "2026-07-08T16:04:37.708Z" },
+]
-    { url = "https://files.pythonhosted.org/packages/f3/82/fea946351658e6534db52d2cc12bc53087cbf87f9440c5f180f367c1950b/torch-2.13.0-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:796633c4cdf0fe2cdced72d8f88f22e73dbcfce83132763162f6d4bff13b820b", size = 526605292, upload-time = "2026-07-08T16:04:22.81Z" },
+
-    { url = "https://files.pythonhosted.org/packages/21/d6/e8f3c6f7e01f626f77259de9860d2a78bc84c40539e28e79b7e98b0bb659/torch-2.13.0-cp312-cp312-win_amd64.whl", hash = "sha256:024c6cc0c1b085f2f91f20a3dc27b0471d021c31ce84b81be3afdc39f791fd9d", size = 122057313, upload-time = "2026-07-08T16:03:53.43Z" },
+[[package]]
+name = "torch"
+version = "2.13.0+cu130"
+source = { registry = "https://download.pytorch.org/whl/cu130" }
+resolution-markers = [
+    "sys_platform == 'linux' or sys_platform == 'win32'",
+]
+dependencies = [
+    { name = "cuda-bindings", marker = "sys_platform != 'win32'" },
+    { name = "cuda-toolkit", extra = ["cublas", "cudart", "cufft", "cufile", "cupti", "curand", "cusolver", "cusparse", "nvjitlink", "nvrtc", "nvtx"], marker = "sys_platform != 'win32'" },
+    { name = "filelock" },
+    { name = "fsspec" },
+    { name = "jinja2" },
+    { name = "networkx" },
+    { name = "nvidia-cudnn-cu13", marker = "sys_platform != 'win32'" },
+    { name = "nvidia-cusparselt-cu13", marker = "sys_platform != 'win32'" },
+    { name = "nvidia-nccl-cu13", marker = "sys_platform != 'win32'" },
+    { name = "nvidia-nvshmem-cu13", marker = "sys_platform != 'win32'" },
+    { name = "setuptools" },
+    { name = "sympy" },
+    { name = "triton", marker = "sys_platform != 'win32'" },
+    { name = "typing-extensions" },
+]
+wheels = [
+    { url = "https://download-r2.pytorch.org/whl/cu130/torch-2.13.0%2Bcu130-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:5ebd552c887e707c8e64927aceb8377ca2e81588c4e7494bcd23cb8ac0aca14d", upload-time = "2026-07-08T20:24:19Z" },
+    { url = "https://download-r2.pytorch.org/whl/cu130/torch-2.13.0%2Bcu130-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:8db7338e6895c3d4bd89a02ff4209507d1f0cf2ffeb3b898538b5a07d1ea8c1e", upload-time = "2026-07-08T20:24:52Z" },
+    { url = "https://download-r2.pytorch.org/whl/cu130/torch-2.13.0%2Bcu130-cp312-cp312-win_amd64.whl", hash = "sha256:2efab1e83604ca628c6d85b9e188c153690980498d1297081a9dad704919303c", upload-time = "2026-07-08T20:26:27Z" },
 ]
 
 [[package]]
@@ -1331,11 +1361,24 @@ wheels = [
 name = "torchaudio"
 version = "2.11.0"
 source = { registry = "https://pypi.org/simple" }
+resolution-markers = [
+    "sys_platform != 'linux' and sys_platform != 'win32'",
+]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/f1/b1/77658817acacd01a72b714440c62f419efc4d90170e704e8e7a2c0918988/torchaudio-2.11.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:a1cf1acc883bee9cb906a933572fed6a8a933f86ef34e9ea7d803f72317e8c1b", size = 684226, upload-time = "2026-03-23T18:13:40.023Z" },
-    { url = "https://files.pythonhosted.org/packages/78/28/c7adc053039f286c2aca0038b766cbe3294e66fec6b29a820e95128f9ede/torchaudio-2.11.0-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:bc653defca1c16154398517a1adc98d0fb7f1dd08e58ced217558d213c2c6e29", size = 1626670, upload-time = "2026-03-23T18:13:42.162Z" },
+]
-    { url = "https://files.pythonhosted.org/packages/88/d8/d6d0f896e064aa67377484efef4911cdcc07bce2929474e1417cc0af18c2/torchaudio-2.11.0-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:6503c0bdb29daf2e6281bb70ea2dfe2c3553b782b619eb5d73bdadd8a3f7cecf", size = 1771992, upload-time = "2026-03-23T18:13:33.188Z" },
+
-    { url = "https://files.pythonhosted.org/packages/23/a8/941277ecc39f7a0a169d554302a1f1afd87c1d94a8aec828891916cea59a/torchaudio-2.11.0-cp312-cp312-win_amd64.whl", hash = "sha256:478110f981e5d40a8d82221732c57a56c85a1d5895fb8fe646e86ee15eded3bd", size = 328663, upload-time = "2026-03-23T18:13:19.218Z" },
+[[package]]
+name = "torchaudio"
+version = "2.11.0+cu130"
+source = { registry = "https://download.pytorch.org/whl/cu130" }
+resolution-markers = [
+    "sys_platform == 'linux' or sys_platform == 'win32'",
+]
+wheels = [
+    { url = "https://download-r2.pytorch.org/whl/cu130/torchaudio-2.11.0%2Bcu130-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:7171f810887e7cd1a4763974d5a1f2e1466692404315bb70705e0f49fb3a28e0", upload-time = "2026-03-23T15:50:26Z" },
+    { url = "https://download-r2.pytorch.org/whl/cu130/torchaudio-2.11.0%2Bcu130-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:3fba988f4301fe13547fe5e99c76d9ae36a27e19ded82eeffed9d2456e12edef", upload-time = "2026-03-23T15:50:26Z" },
+    { url = "https://download-r2.pytorch.org/whl/cu130/torchaudio-2.11.0%2Bcu130-cp312-cp312-win_amd64.whl", hash = "sha256:f74949f9ace1e4a6cf9468bdb3211b9cfa0af6ea348125471ac71c8621d6c77d", upload-time = "2026-03-23T15:50:26Z" },
 ]
 
 [[package]]