Contents:

• TorchPfaffian: PyTorch-Based Pfaffian Computation
• Description
  • Statement of need
  • Features
  • Requirements
  • Installation
  • Usage
• Important Links
• Found a bug, have a question or a feature request?
  • License
  • Citation
  • Related software

Tutorials:

• The pfaffian function
  • Input: a single skew-symmetric matrix
  • Input: a batch of matrices
  • Input dtype: real or complex
  • Output: signed value or magnitude
  • Output: differentiable
  • Summary
• Domino tilings and dimer probabilities
  • The Kasteleyn matrix
  • Counting tilings
  • Dimer probabilities from the gradient
  • Many weightings at once
• Performance
  • One thread
  • Eight threads
  • Reading the results
• Limitations
  • Input assumptions
  • Supported dtypes and half precision
  • torch.func transforms
  • Thread count and MKL
  • Derivatives at singular matrices
  • Tiny pivots
  • Devices and platforms

Modules:

• torch_pfaffian package
  • Subpackages
  • Submodules
  • torch_pfaffian.utils module
  • Module contents