The edgemodelr package includes code from several sources:

================================================================================
llama.cpp and GGML Library
================================================================================
Files: src/llama/*, src/ggml/*
Copyright: (c) 2023-2024 The ggml authors
License: MIT
URL: https://github.com/ggml-org/llama.cpp
     https://github.com/ggerganov/ggml
Primary Author: Georgi Gerganov

The llama.cpp inference engine and GGML tensor library provide the core
functionality for local large language model inference. These are distributed
under the MIT License.

================================================================================
YaRN RoPE Implementation
================================================================================
Files: src/ggml/ggml-cpu/ops.cpp (rope_yarn function and related code)
Copyright: (c) 2023 Jeffrey Quesnelle and Bowen Peng
License: MIT
URL: https://github.com/jquesnelle/yarn

The YaRN (Yet another RoPE extensioN method) scaled rotary position embedding
implementation is used for efficient attention mechanisms.

================================================================================
DRY Sampler
================================================================================
Files: src/llama/llama-sampling.cpp (DRY sampler related functions)
Author: pi6am
License: MIT (via Koboldcpp)
URL: https://github.com/LostRuins/koboldcpp/pull/982

The DRY (Don't Repeat Yourself) sampler implementation ported from Koboldcpp
provides advanced text generation capabilities.

================================================================================
Z-Algorithm Implementation
================================================================================
Files: src/llama/llama-sampling.cpp (Z-algorithm for suffix matching)
Author: Ivan Yurchenko
License: Public Domain
URL: https://github.com/ivanyu/string-algorithms

The Z-algorithm implementation is used for efficient string matching in the
DRY sampler.

================================================================================

All components are compatible with the MIT License under which edgemodelr
is distributed.