decoding

 1"""
 2Composable inference algorithms with LLMs and programmable logic.
 3
 4Check out the [`README`](https://github.com/benlipkin/decoding/blob/main/README.md)
 5and [`TUTORIAL`](https://github.com/benlipkin/decoding/blob/main/TUTORIAL.md)
 6on [GitHub](https://github.com/benlipkin/decoding/) for more information.
 7
 8Core modules:
 9- `decoding.models`: Interface for working with Language Models (LMs), and
10    leveraging them to generate text or score sequences.
11- `decoding.scorers`: Interface for constructing custom scoring functions
12    that can be used to rank and steer generation.
13- `decoding.generators`: Interface for implementing custom generation algorithms
14    that can be used to sample controlled text from LMs.
15
16Supporting modules:
17- `decoding.pmf`: Data structures for probability mass functions and other collections
18    of measures as well as algorithms for calculating information-theoretic quantities.
19- `decoding.samplers`: Methods for sampling from distributions.
20- `decoding.estimators`: Decision rules for deriving point estimates from distributions.
21    Supports a flexible Minimum Bayes Risk (MBR) interface that accepts arbitrary
22    user-defined utility functions.
23- `decoding.metrics`: Metrics that may be useful for constructing scoring functions.
24- `decoding.utils`: Miscellaneous helper functions for the library.
25"""
26
27from decoding import (
28    estimators,
29    generators,
30    metrics,
31    models,
32    pmf,
33    samplers,
34    scorers,
35    utils,
36)
37
38__all__ = [
39    "estimators",
40    "generators",
41    "metrics",
42    "models",
43    "pmf",
44    "samplers",
45    "scorers",
46    "utils",
47]