RCRobert Collinsinselfenrichment.hashnode.dev·Jun 3 · 8 min readBF16 Edge Cases When Porting CUDA-Oriented TTS Models to MPS and MLX-AudioWhen porting CUDA-oriented TTS models to Apple Silicon, one subtle but important issue is BF16 behavior on Metal. This has shown up in DiaTTS work on MPS, and the same class of issue appears relevant 00
RCRobert Collinsinselfenrichment.hashnode.dev·May 11 · 16 min readRevisiting MetalGraph: What I Learned About MLX, Replay, and Sidecar PerformanceTo set the stage, this post covers follow-up work that made MetalGraph feel less like it is doomed to be only a pure investigation tool and more like something that could be useful in practice. MetalG00
RCRobert Collinsinselfenrichment.hashnode.dev·May 10 · 14 min readThe Honest Limits Reached With MetalGraphMetal Graph is an early-stage reusable explicit graph execution API over Apple Metal. MetalGraph started with a simple question: can I (or rather, chatGPT and I) build something graph-like around Appl00
RCRobert Collinsinselfenrichment.hashnode.dev·Dec 5, 2025 · 4 min readMy First Rust Crate: tinysetqueueThe FIFO Queue That Hates Duplicates This week I published my very first Rust crate, tinysetqueue. A couple of years ago I made a different data structure that was “library-worthy”, but I didn’t know how to use rust (or more specifically, crates.io) ...00
RCRobert Collinsinselfenrichment.hashnode.dev·Sep 15, 2025 · 6 min readMobileLLM-R1-950M meets Apple SiliconFrom Stub to Coherent MLX Model, a story-shaped release note Getting Meta's efficient 950M parameter model running natively on Apple Silicon for fast, local inference I’ve done a tiny bit of contributions to the llama.cpp community, including adding ...00