deps(diart): improve diart dependency tree. rename gpu-cu129 dependency group to cu129

.github/workflows/support-matrix.yml

@@ -1,134 +0,0 @@
-name: Support Matrix
-
-on:
-  pull_request:
-  workflow_dispatch:
-    inputs:
-      timeout_sec:
-        description: "Per-case timeout in seconds"
-        required: true
-        default: "300"
-
-permissions:
-  contents: read
-
-jobs:
-  test-and-support-matrix:
-    name: |
-      ${{ matrix.os }} | py${{ matrix.python-version }} | tests + support matrix
-    runs-on: ${{ matrix.os }}
-    strategy:
-      fail-fast: false
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-        python-version: ["3.11", "3.12", "3.13"]
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-
-      - name: Setup uv
-        uses: astral-sh/setup-uv@v6
-
-      - name: Install dependencies for tests
-        run: uv sync --extra test --python ${{ matrix.python-version }}
-
-      # - name: Run unit tests
-      #   run: uv run pytest tests/ -v
-
-      - name: Run compatibility matrix
-        shell: bash
-        env:
-          MATRIX_PY: ${{ matrix.python-version }}
-          TIMEOUT_SEC: ${{ github.event.inputs.timeout_sec || '300' }}
-        run: |
-          set -euo pipefail
-
-          base_port=8010
-
-          run_case() {
-            local name="$1"
-            shift 1
-
-            echo "[matrix] scenario=${name}"
-
-            uv run wlk \
-              --host 127.0.0.1 \
-              --port "${base_port}" \
-              --warmup-file "" \
-              --model tiny \
-              "$@" &
-            local server_pid=$!
-
-            cleanup() {
-              kill "${server_pid}" >/dev/null 2>&1 || true
-              wait "${server_pid}" >/dev/null 2>&1 || true
-            }
-            trap cleanup RETURN
-
-            local ready=0
-            local checks=$((TIMEOUT_SEC / 2))
-            local i=0
-            while [ "${i}" -lt "${checks}" ]; do
-              if python -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:${base_port}/', timeout=2).read(1)" >/dev/null 2>&1; then
-                ready=1
-                break
-              fi
-              if ! kill -0 "${server_pid}" >/dev/null 2>&1; then
-                break
-              fi
-              sleep 2
-              i=$((i + 1))
-            done
-
-            if [ "${ready}" -ne 1 ]; then
-              echo "[matrix] ${name} failed (startup_not_ready)"
-              return 1
-            fi
-
-            echo "[matrix] ${name} passed"
-            base_port=$((base_port + 1))
-          }
-
-          # FW profile: cpu + diart + sortformer + translation
-          export UV_PROJECT_ENVIRONMENT=".ci-envs/${{ matrix.os }}-py${{ matrix.python-version }}-fw"
-          uv sync --python "${MATRIX_PY}" --no-dev \
-            --extra cpu \
-            --extra diarization-diart \
-            --extra diarization-sortformer \
-            --extra translation
-
-          run_case \
-            "fw-diart-translation" \
-            --backend faster-whisper \
-            --diarization \
-            --diarization-backend diart \
-            --language en \
-            --target-language es
-
-          run_case \
-            "fw-sortformer-translation" \
-            --backend faster-whisper \
-            --diarization \
-            --diarization-backend sortformer \
-            --language en \
-            --target-language es
-
-          unset UV_PROJECT_ENVIRONMENT
-
-          # Voxtral profile: cpu + diart + voxtral-hf + translation
-          export UV_PROJECT_ENVIRONMENT=".ci-envs/${{ matrix.os }}-py${{ matrix.python-version }}-voxtral"
-          uv sync --python "${MATRIX_PY}" --no-dev \
-            --extra cpu \
-            --extra diarization-diart \
-            --extra voxtral-hf \
-            --extra translation
-
-          run_case \
-            "voxtral-diart-translation" \
-            --backend voxtral \
-            --diarization \
-            --diarization-backend diart \
-            --language en \
-            --target-language es
-
-          unset UV_PROJECT_ENVIRONMENT

Dockerfile

@@ -23,7 +23,7 @@ ENV UV_PYTHON_INSTALL_DIR=/python
 RUN uv python install 3.12
 
 # Install dependencies first to leverage caching
-ARG EXTRAS=gpu-cu129
+ARG EXTRAS=cu129
 COPY pyproject.toml uv.lock /app/
 RUN set -eux; \
   set --; \

README.md

@@ -75,7 +75,7 @@ Go to `chrome-extension` for instructions.
 | Feature | `uv sync` | `pip install -e` |
 |-----------|-------------|-------------|
 | **CPU PyTorch stack** | `uv sync --extra cpu` | `pip install -e ".[cpu]"` |
-| **CUDA 12.9 PyTorch stack** | `uv sync --extra gpu-cu129` | `pip install -e ".[gpu-cu129]"` |
+| **CUDA 12.9 PyTorch stack** | `uv sync --extra cu129` | `pip install -e ".[cu129]"` |
 | **Translation** | `uv sync --extra translation` | `pip install -e ".[translation]"` |
 | **Sentence tokenizer** | `uv sync --extra sentence_tokenizer` | `pip install -e ".[sentence_tokenizer]"` |
 | **Voxtral (HF backend)** | `uv sync --extra voxtral-hf` | `pip install -e ".[voxtral-hf]"` |
@@ -86,10 +86,10 @@ Supported GPU profiles:
 
 ```bash
 # Profile A: Sortformer diarization
-uv sync --extra gpu-cu129 --extra diarization-sortformer
+uv sync --extra cu129 --extra diarization-sortformer
 
 # Profile B: Voxtral HF + translation
-uv sync --extra gpu-cu129 --extra voxtral-hf --extra translation
+uv sync --extra cu129 --extra voxtral-hf --extra translation
 ```
 
 `voxtral-hf` and `diarization-sortformer` are intentionally incompatible extras and must be installed in separate environments.
@@ -322,8 +322,8 @@ docker compose up --build wlk-cpu
 #### Customization
 
 - `--build-arg` Options:
-  - `EXTRAS="gpu-cu129,diarization-sortformer"` - GPU Sortformer profile extras.
+  - `EXTRAS="cu129,diarization-sortformer"` - GPU Sortformer profile extras.
-  - `EXTRAS="gpu-cu129,voxtral-hf,translation"` - GPU Voxtral profile extras.
+  - `EXTRAS="cu129,voxtral-hf,translation"` - GPU Voxtral profile extras.
   - `EXTRAS="cpu,diarization-diart,translation"` - CPU profile extras.
   - Hugging Face cache + token are configured in `compose.yml` using a named volume and `HF_TKN_FILE` (default: `./token`).
 

compose.yml

@@ -4,7 +4,7 @@ services:
       context: .
       dockerfile: Dockerfile
       args:
-        EXTRAS: ${GPU_SORTFORMER_EXTRAS:-gpu-cu129,diarization-sortformer}
+        EXTRAS: ${GPU_SORTFORMER_EXTRAS:-cu129,diarization-sortformer}
     image: wlk:gpu-sortformer
     gpus: all
     ports:
@@ -21,7 +21,7 @@ services:
       context: .
       dockerfile: Dockerfile
       args:
-        EXTRAS: ${GPU_VOXTRAL_EXTRAS:-gpu-cu129,voxtral-hf,translation}
+        EXTRAS: ${GPU_VOXTRAL_EXTRAS:-cu129,voxtral-hf,translation}
     image: wlk:gpu-voxtral
     gpus: all
     ports:

pyproject.toml

@@ -39,21 +39,31 @@ sentence_tokenizer = ["mosestokenizer", "wtpsplit"]
 voxtral-hf = [
     "transformers>=5.2.0; python_version >= '3.10'",
     "mistral-common[audio]",
+    "accelerate>=0.12",
 ]
 cpu = ["torch>=2.0.0", "torchaudio>=2.0.0"]
-gpu-cu129 = [
+cu129 = [
     "torch>=2.0.0",
     "torchaudio>=2.0.0",
     'triton>=2.0.0; platform_machine == "x86_64" and (sys_platform == "linux" or sys_platform == "linux2")',
 ]
 diarization-sortformer = ["nemo-toolkit[asr]>2.4; python_version >= '3.10'"]
-diarization-diart = ["diart"]
+diarization-diart = [
+    "diart",
+    "torch<2.9.0",
+    "torchaudio<2.9.0",
+    "torchvision<0.24.0",
+]
 
 [tool.uv]
 conflicts = [
     [
         { extra = "cpu" },
-        { extra = "gpu-cu129" },
+        { extra = "cu129" },
+    ],
+    [
+        { extra = "diarization-diart" },
+        { extra = "cu129" },
     ],
     [
         { extra = "voxtral-hf" },
@@ -64,11 +74,16 @@ conflicts = [
 [tool.uv.sources]
 torch = [
     { index = "pytorch-cpu", extra = "cpu", marker = "platform_system != 'Darwin'" },
-    { index = "pytorch-cu129", extra = "gpu-cu129", marker = "platform_system == 'Linux' and platform_machine == 'x86_64'" },
+    { index = "pytorch-cpu", extra = "diarization-diart", marker = "platform_system != 'Darwin'" },
+    { index = "pytorch-cu129", extra = "cu129", marker = "platform_system == 'Linux' and platform_machine == 'x86_64'" },
 ]
 torchaudio = [
     { index = "pytorch-cpu", extra = "cpu", marker = "platform_system != 'Darwin'" },
-    { index = "pytorch-cu129", extra = "gpu-cu129", marker = "platform_system == 'Linux' and platform_machine == 'x86_64'" },
+    { index = "pytorch-cpu", extra = "diarization-diart", marker = "platform_system != 'Darwin'" },
+    { index = "pytorch-cu129", extra = "cu129", marker = "platform_system == 'Linux' and platform_machine == 'x86_64'" },
+]
+torchvision = [
+    { index = "pytorch-cpu", extra = "diarization-diart", marker = "platform_system != 'Darwin'" },
 ]
 
 [[tool.uv.index]]