pyproject.toml

source config python

File Path: pyproject.toml

Purpose: Defines project metadata, dependencies, and build system configuration (PEP 518).

Project Metadata

• Name: arabic-sign-language-recognition
• Python: >=3.12
• Dependencies:
  • Web: fastapi, uvicorn
  • ML: numpy, opencv-python-headless, mediapipe
  • ONNX: onnx, onnxruntime, onnxscript

Optional Dependencies (extras)

• torch-cpu: Installs CPU-only PyTorch.
• torch-cuda: Installs CUDA-enabled PyTorch.

Development Group (dev)

• Data Analysis: pandas, scikit-learn
• Visualization: seaborn, plotly, streamlit
• Utils: tqdm, openpyxl

Tool Configuration ([tool.uv])

Configures uv (the package manager) to handle the complex PyTorch CPU/CUDA split using mutually exclusive index URLs.

[tool.uv.sources]
torch = [
    { index = "pytorch-cpu", extra = "torch-cpu" },
    { index = "pytorch-cu126", marker = "sys_platform != 'darwin'", extra = "torch-cuda" },
]

Related Documentation

Consumed By:

• Dockerfile - Install step