Skip to content

Installation

KeyDNN is a lightweight deep learning framework implemented from scratch in Python with a practical CPU/CUDA execution stack.

Platform support (v2.1.0): Windows 10/11 x64 is the supported platform (CPU + CUDA).
Linux/macOS may build for CPU-only use, but are not officially supported or CI-validated yet.

Install from PyPI

pip install keydnn

Install from source (development)

git clone https://github.com/keywind127/keydnn_v2.git
cd keydnn_v2
pip install -e .

Verify installation

python - <<EOF
import keydnn
print("KeyDNN imported:", keydnn.__name__)
print("Tensor:", keydnn.Tensor)
print("CUDA available:", keydnn.cuda_available())
EOF

If this succeeds, your environment is ready for the Quickstart.

Notes

  • KeyDNN’s public APIs are re-exported at the top level of keydnn.
  • Internal packages follow a layered PIAD architecture and are not part of the public contract.
  • CUDA usage requires additional setup; see Getting Started → CUDA.