This is a sample app built on the llama.cpp project.
https://github.com/ggerganov/llama.cpp
As llama.cpp is a Swift package, this app simply depends on the llama.cpp, and utilizes its API.
By default, tinyllama and ELYZA japanese Llama is set as a resource. Please re-link the model weight in the SwiftLlamaApp/Resources/models directory.
You can downloade the model weight from the following URLs.
https://huggingface.co/mmnga/ELYZA-japanese-Llama-2-7b-instruct-gguf
https://huggingface.co/TheBloke/TinyLlama-1.1B-Chat-v0.3-GGUF/tree/main
There is a blog entry for this sample.