Skill Claude CodeCodex
How to add/upload tuned config CSVs under aiter/configs (incl. modelconfigs/) without introducing duplicate shapes, and how to find & resolve duplicate-shape collisions. Use whenever adding a model's tuned config, merging/uploading config CSVs, editing anything under aiter/configs/, or when a run hits "duplicate shape…
Skill Claude CodeCodex
Standard structure for aiter optests under optests/test.py — @benchmark + runperftest candidate loop, a torch reference, a final markdown summary table, a main guard so the module is importable, and faithful reproduction of the real model call (output buffer, layout, shapes). Use whenever writing, rewriting, or…
Skill Claude CodeCodex
Comprehensive reference for authoring FlyDSL GPU kernels on AMD GPUs. Covers the layout algebra, tiled copy/MMA, buffer ops, loop-carried range loops, SharedAllocator (LDS), autotuning, and common patterns. Use when writing, reviewing, or understanding FlyDSL kernel code.
Skill Claude CodeCodex
Modernize FlyDSL kernels: replace raw MLIR dialects (arith, scf, vector, llvm, memref, math), ArithValue, redundant fx. wrapping, fx.Index, bufferops, SmemPtr/SmemAllocator, copyatomcall/mmaatomcall (loop or single atom), and raw rocdl.mfma with the current fx. surface (fx types, Python control flow, makebuffertensor…
Skill Claude CodeCodex
Compile-time optimization guidance for HIP/C++ kernels using opus.hpp. Use when writing or reviewing OPUS kernels, analyzing compile time, reducing template instantiation overhead, or optimizing hipcc build performance.
Skill Claude CodeCodex
Module-level JIT build-wall optimization for opus-based aiter modules. Use when an aiter JIT module's first-call build wall is a user-visible bottleneck or when adding a new module.
Skill Claude CodeCodex
AI code review for aiter PRs. Catches perf regressions, silent correctness bugs, dispatch gate holes, and AI-generated code patterns. Invoke with a PR number; works through fetch → semantic understanding → rule checklist → verdict. Add new rules here as patterns emerge from real reviews.
Instructions file GitHub Copilot
Instructions for ROCm/aiter, covering aiter triton ops — pr review rules, reuse before adding, folder structure and imports, tuned configs: json placement and naming and tuned configs: json contents.
Plugin Claude Code
Tools to deploy and debug Spur, an AI-native job scheduler.
Skill Claude CodeCodex
Use when the user asks to deploy/install a Spur cluster, or roll out a new build to an already-running one, on one or more bare-metal hosts over SSH. Covers every topology the Ansible playbook does — single-node, multi-node, HA (multi-controller Raft), and HA with separate compute nodes — plus optional PostgreSQL…