Anton Marini

To chime in, I think this can be revisited if we decide to tackle internally pooling IOSurfaces, which would allow us to vend unique surfaces per frame (though they would...

for reference : ``` /* These calls let you attach property list types to a IOSurface buffer. These calls are expensive (they essentially must serialize the data into the kernel)...

Hi. Firstly, thanks for this repo. This project is awesome! Forgive me if im incorrect in understanding the ramifications of this, but one thought after a brief look at this...

Thanks @ggerganov - and to be clear, I wasn't trying to point out any flaws, I'm aware this entire endeavor is a work in progress and theres a lot of...

@jvcleave I have a second auvidea board, and I kind of stopped messing with it due to other things. Would you be up for me sending you one of the...

FWIW, wrt to Mac at least, Apple has released CoreML version of Stable Diffusion which seems decent: https://github.com/apple/ml-stable-diffusion Perhaps relevant to fast, local Mac only installs?

Hi Swift bindings would be amazing for a few reasons. Im one of the founders at https://ozu.ai - we're making tools for multi modal creative video understanding which uses inference...

@rovo79 MLX as I understand it wont leverage the ANE, see this closed issue https://github.com/ml-explore/mlx/issues/18 That said, CoreML allows you to expose 'custom layers' which could be implemented in MLX,...

WRT to Llama.cpp startup - It could be initial CoreML model compilation. GGML (the Llama.cpp runtime) can leverage the ANE, and I think (?) via the that tokenizer can run...

Many moons ago I wrote a pure AVFoundation based CMSampleBuffer decoder which only keeps the 30 seconds of memory buffers available - so you never go above that: Im unsure...