Just released an experimental llama.cpp bindings for KMP (Android and iOS) for now. Check <Here> . T...

# feed

[ayodele]

Just released an experimental llama.cpp bindings for KMP (Android and iOS) for now. Check Here . The example project in the repo contains an example on how to use the API with Compose and SwiftUI. But there's an example using Compose multiplatform here . We hope to add more features and make it more stable. But of course contributions are welcome. Kindly leave a star if you find it interesting 😄.

[Sergey Y.]

This seems like a very interesting project. However, from the description, it’s not entirely clear how llama.cpp is compiled for Android.
I see that for iOS, it requires the Metal framework, so it’s hardware-accelerated. But what about Android? Is it purely CPU-based?

[ayodele]

Yes it's purely CPU based for now. llamacpp have an experimental support using OpenCL. Which we hope to add. But it'll only support some Adreno chips with Q4_0 quantization