These models failed zero-shot transfer to an unseen GPU. Trained only on H100 and H200 measurements and evaluated once on B200, the primary Ridge model had a mean-power WAPE of 30.02% and the secondary GBDT model 43.33%; for p95 power the WAPE was 23.71% (Ridge) and 43.04% (GBDT). All 240 B200 runs were under-predicted, for both models and for both mean and p95 power. Both models are released only as experimental reproduction baselines for that negative result. They are not general-purpose GPU power predictors and should not be used to budget power for any GPU.
🇰🇷 한국어 카드: MODEL_CARD.ko.md
| File | Role | Estimator | Outputs |
|---|---|---|---|
model.joblib |
Primary (fixed before evaluation) | StandardScaler (numeric) + OneHotEncoder(handle_unknown="ignore") (categorical) + Ridge(alpha=1.0) |
mean power (W), p95 power (W) |
model_secondary.joblib |
Secondary (pre-registered diagnostic, never substituted) | numeric passthrough + OneHotEncoder + HistGradientBoostingRegressor (max_iter 200, learning_rate 0.05, min_samples_leaf 8, l2_regularization 1.0, random_state 2026) |
mean power (W), p95 power (W) |
Each file is a Python dict {"name", "models": {"mean_power": Pipeline, "p95_power": Pipeline}, "rule", "n_rows"}.
There was no hyperparameter search.
feature_schema.json)Numeric: requested_input_tokens, requested_output_tokens, max_concurrency, target_rate_rps, burstiness,
power_limit_w. Categorical: workload_family (balanced, decode_heavy, prefill_heavy), loop (closed, open).
Preprocessing (implemented in predict.py, identical to the training code):
source_gpu_type is mapped to power_limit_w = H100 400 W,
H200 600 W, B200 1000 W. Other GPU types are rejected.target_rate_rps and burstiness are set to 0; the loop one-hot distinguishes them.504 valid runs: H100 186 + H200 318. Serving stack: vLLM 0.28.0 serving Qwen3-8B in bf16 on a single GPU, three synthetic workload shapes, closed-loop and open-loop load. B200 data was never used for training or model selection. The measurement data itself is not released.
Labels are NVML GPU board power over the request window (first request start to last request end): mean power (energy-counter difference divided by window length) and p95 power (95th percentile of 0.2 s samples; NVML power is itself roughly a 1 s moving average, so this is not an instantaneous peak). This is GPU board power only, not server, node, or rack power.
| Model | Mean power WAPE | p95 power WAPE | Signed bias (mean / p95) |
|---|---|---|---|
| Ridge (primary) | 30.02% | 23.71% | −30.02% / −23.71% |
| GBDT (secondary) | 43.33% | 43.04% | −43.33% / −43.04% |
| Ridge, by workload | Mean power WAPE | p95 power WAPE |
|---|---|---|
| balanced | 29.96% | 23.96% |
| decode_heavy | 32.27% | 24.53% |
| prefill_heavy | 27.87% | 22.71% |
EVAL_B200.json. Per-run values are not released.model.joblib, model_secondary.joblib). See LICENSE and NOTICE.model.joblib and model_secondary.joblib are joblib/pickle files. Unpickling can execute arbitrary code. Load
them only if you trust where you got them, and check SHA256SUMS first. predict.py refuses to load a file whose
SHA-256 differs from the pinned value.
pip install joblib "scikit-learn==1.9.0" pandas # the only scikit-learn version these files were verified with
sha256sum -c SHA256SUMS
python predict.py example_input.json # both models
python predict.py example_input.json --model primary
example_input.json contains synthetic, hand-written rows (not measured runs). Output is a JSON list with
mean_power_w and p95_power_w per row and model. The files were loaded and reproduced in the environment recorded
in RELEASE_CHECK.md; that is not a statement about the training environment.
ThakiCloud, PowerProof P2 (2026). If you use these baselines, please state that they failed zero-shot transfer to B200.