On-Device on Warpspace Bloghttp://blog.caveduck.io/tags/on-device/Recent content in On-Device on Warpspace BlogHugo -- 0.154.5en-USTue, 07 Jul 2026 00:00:00 +0000That Free Korean NPU the Government Hands Out — Can You Actually Use It? Notes From Porting an Unsupported Modelhttp://blog.caveduck.io/posts/modernbert-on-rebellions-npu/Tue, 07 Jul 2026 00:00:00 +0000http://blog.caveduck.io/posts/modernbert-on-rebellions-npu/Moving a persona-deviation detector (an mmBERT-based classifier) that we served on GPU in production onto a Korean NPU (Rebellions ATOM+) — the journey from recognizing the problem to writing the adapter, compiling on real hardware, and benchmarking.Neural Network Compilation, Made Simple — How a PyTorch Model Becomes Hardware Languagehttp://blog.caveduck.io/posts/neural-net-compilation/Mon, 06 Jul 2026 00:00:00 +0000http://blog.caveduck.io/posts/neural-net-compilation/What exactly does it mean to ‘compile a model’? From the difference with eager execution to graph capture, operator fusion, kernel lowering, static shapes, and memory planning — we walk through it step by step with diagrams and an interactive demo.