gradient_stat_annealing
Update PINN data-loss weights from residual and data-fit gradient statistics using the paper's moving-average annealing rule.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- gradient_stat_annealing
- description
- Update PINN data-loss weights from residual and data-fit gradient statistics using the paper's moving-average annealing rule.
02 资源文件 2
- 脚本scripts/annealing.py1.2 KB
- 脚本tests/test_annealing.py275 B
03 SKILL.md 正文
Gradient Statistic Annealing
Use this skill when a composite PINN loss keeps residual and data-fit losses separate and needs the paper's adaptive lambda update. Do not use it when only aggregate loss gradients are available.
Inputs
- Residual-gradient magnitudes or a residual-gradient vector.
- One or more data-fit gradient vectors.
- Current lambda values.
- Smoothing coefficient
alphaand numericalepsilon.
Outputs
- Updated lambdas.
- Instantaneous lambda estimates.
- Residual max-gradient statistic and data mean-gradient statistics.
- Balance ratios for recovery mechanism checks.
Workflow
- Compute
max(abs(grad L_r)). - Compute
mean(abs(grad L_i))for each data-fit loss. - Compute
lambda_hat_i = max_grad_r / (mean_grad_i + epsilon). - Apply
lambda_i = (1 - alpha) * lambda_i + alpha * lambda_hat_i. - Log all values in the training trace.
Validation
Run python scripts/annealing.py --self-test or the generated skill tests.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).