doctomoe2019/strix-dual-exp
Two-rank tensor-parallel LLM serving over USB4v2 between AMD Strix Halo hosts: a custom Linux thunderbolt stream transport, self-healing link, and gufo TP2 integration
GitHub repository with 12 stars and 0 forks.
Language: C
Topics: amdgpu, inference, linux-kernel, llm, rocm, strix-halo, tensor-parallelism, thunderbolt, usb4