hyperloom-remote-mn-qwen3-30b

Two nodes, one 30B MoE, 4 hours. PD-disagg or RayJob, sglang tuned, auto-monitored. You sleep.

Updated 2026-08-13 · Source https://github.com/AMD-AGI/Hyperloom/tree/HEAD/docs/how-to/multi-node/…

What does this skill do?

Run a 4-hour multi-node Hyperloom Qwen3-30B-A3B optimization (Infera PD-disaggregated or RayJob aggregated) with --nodes 2 and sglang MoE tuning on MI325X. Hand this skill to the agent to launch and monitor the run.

Why is it worth installing?

Unlike bare launchers, it weaves PD-disagg and RayJob topologies, tuning, and monitoring into one ready-to-run fabric.

How do you install it?

git clone --depth 1 https://github.com/AMD-AGI/Hyperloom.git Then point your agent at the skill directory.

Where does it come from?

Repository: AMD-AGI/Hyperloom 138 stars on GitHub at the time of listing. The repository ships 1 skills in total. Homepage: https://rocm.docs.amd.com/projects/hyperloom

developmentskillAMD-AGI

hyperloom-qwen3-14b-fp8-12h  ·  Hyperloom