Fix merge conflicts

MigrationWorkingNotes.md

@@ -0,0 +1,95 @@
+# UV Setup
+Deprecated notes for myself
+## Structure
+```
+docker/
+  ├── cpu/
+  │   ├── pyproject.toml     # CPU deps (torch CPU)
+  │   └── requirements.lock  # CPU lockfile
+  ├── gpu/
+  │   ├── pyproject.toml     # GPU deps (torch CUDA)
+  │   └── requirements.lock  # GPU lockfile
+  ├── rocm/
+  │   ├── pyproject.toml     # ROCM deps (torch ROCM)
+  │   └── requirements.lock  # ROCM lockfile
+  └── shared/
+      └── pyproject.toml     # Common deps
+```
+
+## Regenerate Lock Files
+
+### CPU
+```bash
+cd docker/cpu
+uv pip compile pyproject.toml ../shared/pyproject.toml --output-file requirements.lock
+```
+
+### GPU
+```bash
+cd docker/gpu
+uv pip compile pyproject.toml ../shared/pyproject.toml --output-file requirements.lock
+```
+
+### ROCM
+```bash
+cd docker/rocm
+uv pip compile pyproject.toml ../shared/pyproject.toml --output-file requirements.lock
+```
+
+## Local Dev Setup
+
+### CPU
+```bash
+cd docker/cpu
+uv venv
+.venv\Scripts\activate  # Windows
+uv pip sync requirements.lock
+```
+
+### GPU
+```bash
+cd docker/gpu
+uv venv
+.venv\Scripts\activate  # Windows
+uv pip sync requirements.lock --extra-index-url https://download.pytorch.org/whl/cu121 --index-strategy unsafe-best-match
+```
+
+### ROCM
+```bash
+cd docker/rocm
+uv venv
+source .venv/bin/activate
+# not tested on Windows
+#.venv\Scripts\activate  # Windows
+uv pip sync requirements.lock --extra-index-url https://download.pytorch.org/whl/rocm6.2
+```
+
+### Run Server
+```bash
+# From project root with venv active:
+uvicorn api.src.main:app --reload
+```
+
+## Docker
+
+### CPU
+```bash
+cd docker/cpu
+docker compose up
+```
+
+### GPU
+```bash
+cd docker/gpu
+docker compose up
+```
+
+### ROCM
+```bash
+cd docker/rocm
+docker compose up
+```
+
+## Known Issues
+- Module imports: Run server from project root
+- PyTorch CUDA: Always use --extra-index-url and --index-strategy for GPU env

docker/rocm/Dockerfile

@@ -0,0 +1,64 @@
+FROM rocm/dev-ubuntu-22.04:6.3.1
+
+# Install Python and other dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    python3.10 \
+    python3.10-venv \
+    espeak-ng \
+    git \
+    libsndfile1 \
+    curl \
+    && apt-get clean \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install uv
+COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
+
+# Create non-root user
+RUN useradd -m -u 1000 appuser
+
+# Create directories and set ownership
+RUN mkdir -p /app/models && \
+    mkdir -p /app/api/src/voices && \
+    chown -R appuser:appuser /app
+
+USER appuser
+
+# Download and extract models
+WORKDIR /app/models
+RUN curl -L -o model.tar.gz https://github.com/remsky/Kokoro-FastAPI/releases/download/v0.0.1/kokoro-82m-pytorch.tar.gz && \
+    tar xzf model.tar.gz && \
+    rm model.tar.gz
+
+# Download and extract voice models
+WORKDIR /app/api/src/voices
+RUN curl -L -o voices.tar.gz https://github.com/remsky/Kokoro-FastAPI/releases/download/v0.0.1/voice-models.tar.gz && \
+    tar xzf voices.tar.gz && \
+    rm voices.tar.gz
+
+# Switch back to app directory
+WORKDIR /app
+
+# Copy dependency files
+COPY --chown=appuser:appuser pyproject.toml ./pyproject.toml
+
+# Install dependencies
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv venv && \
+    uv sync --extra rocm --no-install-project
+
+# Copy project files
+COPY --chown=appuser:appuser api ./api
+
+# Install project
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv sync --extra rocm
+
+# Set environment variables
+ENV PYTHONUNBUFFERED=1
+ENV PYTHONPATH=/app:/app/models
+ENV PATH="/app/.venv/bin:$PATH"
+ENV UV_LINK_MODE=copy
+
+# Run FastAPI server
+CMD ["uv", "run", "python", "-m", "uvicorn", "api.src.main:app", "--host", "0.0.0.0", "--port", "8880", "--log-level", "debug"]

docker/rocm/docker-compose.yml

@@ -0,0 +1,44 @@
+name: kokoro-tts
+services:
+  kokoro-tts:
+    # image: ghcr.io/remsky/kokoro-fastapi-rocm:v0.1.0
+    build:
+      context: ../..
+      dockerfile: docker/rocm/Dockerfile
+    volumes:
+      - ../../api/src:/app/api/src  # Mount src for development
+      - ../../api/src/voices:/app/api/src/voices  # Mount voices for persistence
+    ports:
+      - "8880:8880"
+    environment:
+      - PYTHONPATH=/app:/app/models
+      - TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1
+      # This suppresses excessive warning logs. Probably not a good idea to suppress, but no other solution found
+      # (see https://github.com/ROCm/MIOpen/issues/2981)
+      - MIOPEN_LOG_LEVEL=3
+    devices:
+      - /dev/kfd:/dev/kfd
+      - /dev/dri:/dev/dri
+    security_opt:
+      - seccomp=unconfined
+    group_add:
+      - video
+    ipc: host
+
+  # Gradio UI service
+  gradio-ui:
+    image: ghcr.io/remsky/kokoro-fastapi-ui:v0.1.0
+    # Uncomment below to build from source instead of using the released image
+    # build:
+      # context: ../../ui
+    ports:
+      - "7860:7860"
+    volumes:
+      - ../../ui/data:/app/ui/data
+      - ../../ui/app.py:/app/app.py  # Mount app.py for hot reload
+    environment:
+      - GRADIO_WATCH=1  # Enable hot reloading
+      - PYTHONUNBUFFERED=1  # Ensure Python output is not buffered
+      - DISABLE_LOCAL_SAVING=false  # Set to 'true' to disable local saving and hide file view
+      - API_HOST=kokoro-tts  # Set TTS service URL
+      - API_PORT=8880  # Set TTS service PORT

docker/rocm/pyproject.toml

@@ -0,0 +1,24 @@
+[project]
+name = "kokoro-fastapi-rocm"
+version = "0.1.0"
+description = "FastAPI TTS Service - ROCM Version"
+readme = "../README.md"
+requires-python = ">=3.10"
+dependencies = [
+    # Core ML/DL for rocm
+    "pytorch-triton-rocm==3.1.0",
+    "torch==2.5.1+rocm6.2",
+    "transformers==4.47.1",
+]
+
+[tool.uv.workspace]
+members = ["../shared"]
+
+[tool.uv.sources]
+torch = { index = "pytorch-rocm" }
+pytorch-triton-rocm = { index = "pytorch-rocm" }
+
+[[tool.uv.index]]
+name = "pytorch-rocm"
+url = "https://download.pytorch.org/whl/rocm6.2"
+explicit = true

docker/rocm/requirements.lock

@@ -0,0 +1,225 @@
+# This file was autogenerated by uv via the following command:
+#    uv pip compile pyproject.toml ../shared/pyproject.toml --output-file requirements.lock
+aiofiles==23.2.1
+    # via kokoro-fastapi (../shared/pyproject.toml)
+annotated-types==0.7.0
+    # via pydantic
+anyio==4.8.0
+    # via starlette
+attrs==24.3.0
+    # via
+    #   clldutils
+    #   csvw
+    #   jsonschema
+    #   phonemizer
+    #   referencing
+babel==2.16.0
+    # via csvw
+certifi==2024.12.14
+    # via requests
+cffi==1.17.1
+    # via soundfile
+charset-normalizer==3.4.1
+    # via requests
+click==8.1.8
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   uvicorn
+clldutils==3.21.0
+    # via segments
+colorama==0.4.6
+    # via csvw
+coloredlogs==15.0.1
+    # via onnxruntime
+colorlog==6.9.0
+    # via clldutils
+csvw==3.5.1
+    # via segments
+dlinfo==1.2.1
+    # via phonemizer
+exceptiongroup==1.2.2
+    # via anyio
+fastapi==0.115.6
+    # via kokoro-fastapi (../shared/pyproject.toml)
+filelock==3.16.1
+    # via
+    #   huggingface-hub
+    #   pytorch-triton-rocm
+    #   torch
+    #   transformers
+flatbuffers==24.12.23
+    # via onnxruntime
+fsspec==2024.12.0
+    # via
+    #   huggingface-hub
+    #   torch
+greenlet==3.1.1
+    # via sqlalchemy
+h11==0.14.0
+    # via uvicorn
+huggingface-hub==0.27.1
+    # via
+    #   tokenizers
+    #   transformers
+humanfriendly==10.0
+    # via coloredlogs
+idna==3.10
+    # via
+    #   anyio
+    #   requests
+isodate==0.7.2
+    # via
+    #   csvw
+    #   rdflib
+jinja2==3.1.5
+    # via torch
+joblib==1.4.2
+    # via phonemizer
+jsonschema==4.23.0
+    # via csvw
+jsonschema-specifications==2024.10.1
+    # via jsonschema
+language-tags==1.2.0
+    # via csvw
+loguru==0.7.3
+    # via kokoro-fastapi (../shared/pyproject.toml)
+lxml==5.3.0
+    # via clldutils
+markdown==3.7
+    # via clldutils
+markupsafe==3.0.2
+    # via
+    #   clldutils
+    #   jinja2
+mpmath==1.3.0
+    # via sympy
+munch==4.0.0
+    # via kokoro-fastapi (../shared/pyproject.toml)
+networkx==3.4.2
+    # via torch
+numpy==2.2.1
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   onnxruntime
+    #   scipy
+    #   soundfile
+    #   transformers
+onnxruntime==1.20.1
+    # via kokoro-fastapi (../shared/pyproject.toml)
+packaging==24.2
+    # via
+    #   huggingface-hub
+    #   onnxruntime
+    #   transformers
+phonemizer==3.3.0
+    # via kokoro-fastapi (../shared/pyproject.toml)
+protobuf==5.29.3
+    # via onnxruntime
+pycparser==2.22
+    # via cffi
+pydantic==2.10.4
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   fastapi
+    #   pydantic-settings
+pydantic-core==2.27.2
+    # via pydantic
+pydantic-settings==2.7.0
+    # via kokoro-fastapi (../shared/pyproject.toml)
+pylatexenc==2.10
+    # via clldutils
+pyparsing==3.2.1
+    # via rdflib
+python-dateutil==2.9.0.post0
+    # via
+    #   clldutils
+    #   csvw
+python-dotenv==1.0.1
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   pydantic-settings
+pytorch-triton-rocm==3.1.0
+    # via
+    #   kokoro-fastapi-rocm (pyproject.toml)
+    #   torch
+pyyaml==6.0.2
+    # via
+    #   huggingface-hub
+    #   transformers
+rdflib==7.1.2
+    # via csvw
+referencing==0.35.1
+    # via
+    #   jsonschema
+    #   jsonschema-specifications
+regex==2024.11.6
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   segments
+    #   tiktoken
+    #   transformers
+requests==2.32.3
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   csvw
+    #   huggingface-hub
+    #   tiktoken
+    #   transformers
+rfc3986==1.5.0
+    # via csvw
+rpds-py==0.22.3
+    # via
+    #   jsonschema
+    #   referencing
+safetensors==0.5.2
+    # via transformers
+scipy==1.14.1
+    # via kokoro-fastapi (../shared/pyproject.toml)
+segments==2.2.1
+    # via phonemizer
+six==1.17.0
+    # via python-dateutil
+sniffio==1.3.1
+    # via anyio
+soundfile==0.13.0
+    # via kokoro-fastapi (../shared/pyproject.toml)
+sqlalchemy==2.0.27
+    # via kokoro-fastapi (../shared/pyproject.toml)
+starlette==0.41.3
+    # via fastapi
+sympy==1.13.1
+    # via
+    #   onnxruntime
+    #   torch
+tabulate==0.9.0
+    # via clldutils
+tiktoken==0.8.0
+    # via kokoro-fastapi (../shared/pyproject.toml)
+tokenizers==0.21.0
+    # via transformers
+torch==2.5.1+rocm6.2
+    # via kokoro-fastapi-rocm (pyproject.toml)
+tqdm==4.67.1
+    # via
+    #   kokoro-fastapi (../shared/pyproject.toml)
+    #   huggingface-hub
+    #   transformers
+transformers==4.47.1
+    # via kokoro-fastapi-rocm (pyproject.toml)
+typing-extensions==4.12.2
+    # via
+    #   anyio
+    #   fastapi
+    #   huggingface-hub
+    #   phonemizer
+    #   pydantic
+    #   pydantic-core
+    #   sqlalchemy
+    #   torch
+    #   uvicorn
+uritemplate==4.1.1
+    # via csvw
+urllib3==2.3.0
+    # via requests
+uvicorn==0.34.0
+    # via kokoro-fastapi (../shared/pyproject.toml)

pyproject.toml

@@ -44,6 +44,10 @@ dependencies = [
 gpu = [
     "torch==2.6.0+cu124",
 ]
+rocm = [
+    "torch==2.5.1+rocm6.2",
+    "pytorch-triton-rocm==3.1.0",
+]
 cpu = [
     "torch==2.6.0",
 ]
@@ -61,6 +65,7 @@ conflicts = [
     [
         { extra = "cpu" },
         { extra = "gpu" },
+        { extra = "rocm" },
     ],
 ]
 
@@ -68,6 +73,11 @@ conflicts = [
 torch = [
     { index = "pytorch-cpu", extra = "cpu" },
     { index = "pytorch-cuda", extra = "gpu" },
+    { index = "pytorch-rocm", extra = "rocm" },
+]
+# uv would not install torch for rocm6.2 without this explicit transient dependency
+pytorch-triton-rocm = [
+    { index = "pytorch-rocm", extra = "rocm" },
 ]
 
 [[tool.uv.index]]
@@ -80,6 +90,11 @@ name = "pytorch-cuda"
 url = "https://download.pytorch.org/whl/cu124"
 explicit = true
 
+[[tool.uv.index]]
+name = "pytorch-rocm"
+url = "https://download.pytorch.org/whl/rocm6.2"
+explicit = true
+
 [build-system]
 requires = ["setuptools>=61.0"]
 build-backend = "setuptools.build_meta"