coral-pi-rest-server VS llama.cpp

That said, you can always look at something like https://coral.ai/products/accelerator/ to help with the performance you need.

Should only be ~$60 https://coral.ai/products/accelerator/

The Coral USB accelerator might be of interest if you want to do some light ML with a low power budget.

However, a pi doesn't have the strength to run something like Llama.cpp, of course, so I've been considering using something like the Coral USB Accelerator (https://coral.ai/products/accelerator). As I've been learning more about it, it seems to be very geared towards TensorFlow Lite models. But whisper.cpp and Llama.cpp use ggml models.

Recently found this thing, though I haven't found a use case for me.

if you can compile stuff, then looking at llama.cpp (what ollama uses) is also interesting: https://github.com/ggerganov/llama.cpp

the server is here: https://github.com/ggerganov/llama.cpp/tree/master/examples/...

And you can search for any GGUF on huggingface

Yes, Metal seems to allow a maximum of 1/2 of the RAM for one process, and 3/4 of the RAM allocated to the GPU overall. There’s a kernel hack to fix it, but that comes with the usual system integrity caveats. https://github.com/ggerganov/llama.cpp/discussions/2182

For anyone interested in exploring this, llama.cpp has an example implementation here:

https://github.com/ggerganov/llama.cpp/tree/master/examples/...

Getting started with LLMs can be intimidating. In this tutorial we will show you how to fine-tune a large language model using LoRA, facilitated by tools like llama.cpp and KitOps.

well https://github.com/ggerganov/llama.cpp/issues/6849