Show HN: Flyte devbox now runs GPU workloads on Apple Silicon

Category: ai-ml

Tags: apple-silicon, gpu, mlops, flyte, mps, mlx

Score: 7.7/10 (Innovation: 8, Technical: 8, Documentation: 8, Utility: 7)

Flyte Devbox now allows ML workloads to run natively on Apple Silicon GPUs by bridging containerized Flyte tasks to host-side processes using Metal (MPS) and MLX. It solves the gap of no GPU access in Linux VMs on Macs, enabling local training and serving with the same workflow as on a cluster. The design cleverly runs code outside the container while maintaining the Flyte interface and security with HMAC tokens.

Target audience: ML engineers, data scientists, backend developers

Repository: https://www.union.ai/blog-post/your-code-on-the-apple-silicon-gpu-apple-silicon-flyte-devbox

View on Hacker News