Installation¶

Requirements¶

Python 3.10 or later (3.15 is supported at the release-candidate stage)
NumPy ≥ 1.26, SciPy ≥ 1.13 (pulled in automatically)

Python 3.15: support currently covers the numpy and CPU jax backends. Dependency markers omit PyTorch, numba, RAPIDS, and pyarrow on this interpreter; requesting an extra does not make those backends available. The cuda extra still requests JAX CUDA packages on Linux, but that GPU configuration has not been validated on 3.15. The locked pandas version builds from source and needs a C compiler. The typecheck extra uses a beartype 0.23 release candidate.

pip¶

pip install fast-vollib

uv¶

uv add fast-vollib

TestPyPI development snapshots¶

Development builds are published to TestPyPI with VCS-derived versions such as 0.2.1.devN after the v0.2.0 release.

pip¶

pip install --pre \
  --index-url https://test.pypi.org/simple/ \
  --extra-index-url https://pypi.org/simple/ \
  fast-vollib

uv¶

uv pip install --pre \
  --index https://test.pypi.org/simple/ \
  --extra-index-url https://pypi.org/simple/ \
  fast-vollib

Use the normal PyPI channel if you want stable tagged releases only.

Optional extras¶

fast-vollib ships optional extras for GPU and alternate numeric backends.

PyTorch backend¶

pip install "fast-vollib[torch]"
# or
uv add "fast-vollib[torch]"

JAX backend¶

pip install "fast-vollib[jax]"
# or
uv add "fast-vollib[jax]"

Numba backend¶

pip install "fast-vollib[numba]"
# or
uv add "fast-vollib[numba]"

Runtime type checking (optional)¶

Opt-in shape-aware checking of the public API via jaxtyping + beartype. Annotations are already present in the library (PEP 563 strings, zero cost by default); installing this extra only enables the install_import_hook used by fast_vollib._typing.enable_runtime_checks().

pip install "fast-vollib[typecheck]"
# or
uv add "fast-vollib[typecheck]"

See API Reference → Runtime type checking for usage.

Multiple backends¶

pip install "fast-vollib[torch,jax,numba]"
# or
uv add "fast-vollib[torch,jax,numba]"

GPU (Linux only — CUDA 13)¶

Installs PyTorch from the CUDA 13.0 wheel index plus JAX with CUDA 13 support:

uv sync --extra cuda

Note: The cuda extra is Linux-only and requires a CUDA 13.x-capable GPU and driver.

Development install¶

Clone the repository and install with all optional groups:

git clone https://github.com/raeidsaqur/fast-vollib.git
cd fast-vollib

# CPU-only (default)
uv sync --all-groups

# Both backends on CPU/MPS (macOS or Linux without CUDA)
uv sync --all-groups --extra torch --extra jax

# GPU (Linux)
uv sync --all-groups --extra cuda

Optional dependency groups¶

Group	What it installs
`docs`	MkDocs + Material theme for building the documentation site
`bench`	pytest-benchmark and RAPIDS packages for benchmarking

uv sync --group docs   # docs only
uv sync --group bench  # benchmarks only

Verifying the install¶

import fast_vollib
print(fast_vollib.__version__)   # e.g. "0.1.0"
print(fast_vollib.get_backend()) # "numpy", "torch", or "jax"