k2

Hint

We use the lowercase k in k2. That is, it is k2, not K2.

k2 is able to seamlessly integrate Finite State Automaton (FSA) and Finite State Transducer (FST) algorithms into autograd-based machine learning toolkits like PyTorch [1]. k2 supports CPU as well as CUDA. It can process a batch of FSTs at the same time.

[1]

Support for TensorFlow will be added in the future.

We have used k2 to compute CTC loss, LF-MMI loss, and to do decoding including lattice rescoring in the field of automatic speech recognition (ASR) [2].

We hope k2 will have many other applications as well.

[2]

You can find ASR recipes in https://github.com/k2-fsa/icefall

Contents:

• Installation
  • From pre-compiled wheels (recommended)
  • Install from source
  • For developers
  • CUDA and cuDNN
  • FAQs
  • Reporting issues
• Core concepts in k2
  • A simple FSA example
  • Attributes
  • Semirings
  • Vectors of FSAs
  • Autograd
  • Dense fsa vector
  • Ragged arrays
  • References
• Python tutorials
  • Ragged tensor tutorials
  • Fsa tutorials
  • Fsa algorithms tutorials
  • _k2
• Python API reference
  • version
  • k2
  • k2.ragged