Document AITuner harness behavior

This commit is contained in:
2026-04-27 16:34:19 +08:00
parent a962781b6c
commit bc884f6701

View File

@@ -0,0 +1,83 @@
# AITuner Harness Summary
## What The Harness Adds
The harness turns each LLM proposal from open-ended config search into a bottleneck-directed decision.
1. Workload profile
- Extracts L-C-A features from the trace window:
- L: prompt length percentiles and tail ratio.
- C: prefix/cache reuse estimates from `hash_ids` when available.
- A: request rate, burst ratio, and interarrival variation.
- These features are injected into the prompt as a structured `Harnesses` section.
2. Trial diagnostics
- Reads recent trial result JSON.
- Summarizes feasible probes, all-infeasible probes, pass rates, request rates, latency percentiles, and failed SLO reason counts.
- Classifies the active bottleneck as `ttft_prefill`, `decode_tpot`, `admission_or_queueing`, `launch_or_memory`, or unknown.
3. Knob-family harnesses
- Maps bottlenecks to a small number of plausible knob families.
- Current harness families:
- `tensor-parallel-size`: long-prompt TTFT/prefill bottlenecks.
- `max-num-batched-tokens`: prefill batching or fragmentation, with trust-region guards.
- `max-num-seqs`: cache-heavy or admission-limited workloads.
- `enable-chunked-prefill`: long-tail prompt blocking.
- `gpu-memory-utilization`: memory headroom after topology and batching are stable.
- Each family has `use_when`, `procedure`, `guards`, and `active_now` fields.
4. Proposal discipline
- The prompt requires the LLM to choose at most one primary knob family unless history proves a coupled change is needed.
- It must use adjacent legal topology choices and stay inside topology constraints.
- It receives tested config signatures, so it should not repeat already-tried configs.
- It can return `should_stop=true` when no adjacent harness-guided probe is justified.
5. Baseline-first loop
- LLM-driven `study tune` now evaluates the initial engine config first unless `--skip-baseline` is passed.
- This aligns the loop with evaluate-then-search: the first LLM proposal sees measured bottleneck evidence rather than guessing from static config.
## What Accelerates Convergence
The speedup comes from reducing wasted proposal families, not from changing the benchmark metric.
1. Topology-before-runtime on prefill bottlenecks
- For long-prompt, low-cache-reuse windows, the harness activates the TP harness before speculative runtime knobs.
- Example: qwen27b 0-8k chat reached `TP=2, DP=1` at iter 2 under harness replay, while the original run spent iter 2 on `DP=2` and iter 3 on `DP=4`.
2. Guarded stop after a strong incumbent
- If the newest trial is the incumbent and improves per-GPU throughput by at least `3x` over baseline, the harness requires direct evidence before trying runtime-only tweaks.
- Without that guard, the LLM still proposed weak MBT trials after finding the qwen27b best config.
- With the guard, it emits `should_stop=true`.
3. All-infeasible plateau detection
- When recent all-infeasible trials at the same sampling threshold stop improving pass rate and p95 TTFT, the harness blocks repeating the same primary knob family.
- This prevents continuing a direction such as DP-only scale-out after DP4 and DP8 plateau.
4. Cleaner early-stop handling
- Early-stopped probes no longer leave in-flight requests polluting the next probe.
- Default behavior drains in-flight requests for comparable production runs.
- Engine relaunch after early stop is available as opt-in for faster smoke studies, but it is not the default because it can change warm-state comparability.
## qwen27b 0-8k Evidence
Source: `docs/qwen27b-chat-0-8k-harness-fig18.md`.
Metric: best-so-far feasible `request_rate_per_gpu`.
| Variant | Iter 1 | Iter 2 | Iter 3 | Iter 4 | Iter 5-12 |
| --- | ---: | ---: | ---: | ---: | ---: |
| Before harness | 0.0350 | 0.0617 | 0.0617 | 0.2025 | 0.2025 |
| After harness strict replay | 0.0350 | 0.2025 | 0.2025 stop | 0.2025 | 0.2025 |
Result:
- Before harness reached the best value at iter 4.
- After harness reached the same value at iter 2 and stopped at iter 3.
- Iterations-to-best improved from `4` to `2`, a `2x` convergence speedup on this case.
- The harness also avoided eight post-best infeasible runtime-only probes.
## Current Risks
- The harness is prompt-guided, not a hard verifier for every rule. If future LLM outputs ignore a fired guard, proposal validation should reject the blocked family explicitly.
- Strong-incumbent stopping is deliberately conservative for the qwen27b pattern. Workloads with narrow runtime sweet spots, such as qwen235b thinking prefill-only, may need a weaker stop threshold or a "continue local refinement" exception.
- Full fresh reruns on large models are expensive. Strict replay is useful for measuring proposal-path improvements when the proposed configs already exist in prior measured runs, but publication-quality claims still need fresh no-relaunch runs when time allows.