wake up, ANE pretraining just dropped
I blew up another model on my MacBook Pro.
I'm pretraining a 48.8M param GPT in Rust on Apple Neural Engine. From scratch. Not fine-tuning. Not inference. Not LoRA.
As far as I can find, no one has done this. Adjacent work exists — pMetal (fine-tuning), ane-infer (inference), mlx-rs (GPU).