Heterogeneous GPU Sharing on Kubernetes
-
Updated
Nov 16, 2026 - Go
Heterogeneous GPU Sharing on Kubernetes
Personal fork for vLLM inference optimization on Iluvatar BI-V150: attention launch policies, profiling, numerical validation and reproducible benchmarks.
ROSClaw platform pack for the Iluvatar TY1200 embodied-AI box: executable ops skill, controlled modeld, OpenAI-compatible providers (Cosmos-Reason2-2B + Qwen3-Embedding-0.6B), digest-pinned compose deployment
Public Rust GPU research prototype: native Metal SIMD matrices, online attention and mixed Apple MPS execution. Measured on M4 Pro; CoreX research. Commercial contact via LinkedIn.
To associate your repository with the iluvatar topic, visit your repo's landing page and select "manage topics."