---

name: s-tir-tuning description: "Guides TVM S-TIR scheduling, transformations, dlight rules, meta-schedule tuning, builders, runners, and CPU/GPU tuning decisions." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

S-TIR Scheduling and Tuning

Use this route when a task involves tvm.s_tir.Schedule, schedule primitives, transform passes, dlight/default schedules, meta-schedule, tune_tir, or bounded tuning of TIR generated from Relax or TIRx.

Route

1. Confirm TVM imports and the selected target is available. Use ../install-build/SKILL.md for package/build failures.
2. Decide whether the user needs manual scheduling, a transform pass, dlight, or meta-schedule tuning.
3. For manual schedules and debug traces, read references/schedule-workflows.md.
4. For tune_tir, builder/runner/database/cost-model decisions, or bounded tuning runs, read references/meta-schedule.md.
5. Run scripts/meta_schedule_import_probe.py before a tuning run to verify the API and optional dependencies.
6. Use references/troubleshooting.md for ScheduleError, invalid traces, no-candidate tuning, database, XGBoost, runner/RPC, and backend issues.

API anchors

from tvm.s_tir import Schedule
from tvm.s_tir import meta_schedule as ms

sch = Schedule(mod, seed=None, debug_mask="none", error_render_level="detail", enable_check=True)
ms.tune_tir(mod, target, work_dir, max_trials_global, runner="local", database="json")

tune_tir returns a meta-schedule Database. Use tiny max_trials_global values for smoke checks, a disposable work_dir, and a CPU target unless the user has explicitly prepared a GPU/RPC backend.

Boundaries

• Relax model import and high-level pipelines: ../relax-compile/SKILL.md.
• TIRx tile primitives and native kernel dispatch: ../tirx-kernels/SKILL.md.
• RPC tracker/server setup for remote runners: ../rpc-deployment/SKILL.md.
• Package/toolchain failures: ../install-build/SKILL.md.