Skip to content

Compatibility

fast-vollib is designed as a drop-in replacement for py_vollib and py_vollib_vectorized.

What is preserved

Aspect Notes
Argument order Preserved exactly for all pricing, IV, and Greek entry points
flag values "c" / "p" strings work as before
return_as="dataframe" Default return path remains pandas.DataFrame
return_as="series" Returns pandas.Series, same as upstream
NumPy array output return_as="numpy" returns numpy.ndarray

New kwargs (additive only)

fast-vollib adds two optional keyword-only arguments to every function. They are silently consumed when not provided, so existing call sites continue to work without changes.

Kwarg Default Description
backend "auto" Override the numeric backend for this call
return_native False Return the backend's native tensor type

Runtime monkey-patch

fast-vollib provides two separate patch helpers:

  • patch_py_vollib() for the scalar py_vollib namespace
  • patch_py_vollib_vectorized() for the py_vollib_vectorized namespace

Use the helper that matches the package already imported by your application.

patch_py_vollib()

patch_py_vollib() replaces implementations inside the scalar py_vollib namespace at runtime. This is useful for codebases that import from py_vollib directly and cannot be changed.

# At program startup, before any py_vollib imports are used:
import fast_vollib
fast_vollib.patch_py_vollib()

# Now these transparently use fast-vollib under the hood:
from py_vollib.black_scholes import black_scholes
from py_vollib.black_scholes.implied_volatility import implied_volatility
from py_vollib.black_scholes.greeks.numerical import delta

Namespaces patched:

  • py_vollib.black — black, IV, greeks
  • py_vollib.black_scholes — black_scholes, IV, greeks
  • py_vollib.black_scholes_merton — black_scholes_merton, IV, greeks

Requires py_vollib

patch_py_vollib() raises ImportError if py_vollib is not installed.

patch_py_vollib_vectorized()

patch_py_vollib_vectorized() rewires the py_vollib_vectorized top-level, models, implied_volatility, greeks, and api modules to the fast-vollib implementations.

import fast_vollib
fast_vollib.patch_py_vollib_vectorized()

from py_vollib_vectorized import vectorized_black_scholes

Requires py_vollib_vectorized

patch_py_vollib_vectorized() raises ImportError if py_vollib_vectorized is not installed.

Known differences

Behaviour py_vollib_vectorized fast-vollib
Below-intrinsic IV Returns NaN silently Controlled by on_error= ("warn" / "raise" / "ignore")
Backend NumPy only NumPy, PyTorch, JAX
GPU support No Yes (PyTorch + JAX)
q in Black-76 Treated as zero Correctly set to r (Black-76 forward pricing)