sequential_apt_loop
Run a bounded sequential APT proxy loop that updates proposals after corrected atomic training.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- sequential_apt_loop
- description
- Run a bounded sequential APT proxy loop that updates proposals after corrected atomic training.
02 资源文件 2
- 脚本scripts/sequential_proxy.py2.7 KB
- 脚本tests/test_sequential_proxy.py531 B
03 SKILL.md 正文
Sequential APT Loop
Use this skill when a recovery needs to demonstrate the sequential part of Automatic Posterior Transformation under a bounded simulator. It should not be represented as full paper reproduction unless the original benchmark and neural density estimator stack are available.
Inputs
- Observed scalar datum, simulator noise, round count, and simulations per round.
- Prior and initial proposal parameters.
- Atomic training helper implementing proposal-corrected loss.
Outputs
- Per-round proposal and posterior estimate logs.
- Final posterior mean estimate and absolute error against the analytic Gaussian posterior.
- Training trace with parameter movement and loss values.
Workflow
- Begin with the prior as the initial proposal.
- Generate deterministic low-discrepancy simulator samples around each proposal.
- Train the atomic score model on accumulated atom sets.
- Estimate the posterior mean at the observation from corrected atom probabilities.
- Update the next proposal toward that posterior estimate and repeat.
Validation
Run python scripts/sequential_proxy.py --self-test. The test checks that the final proposal moves toward the analytic posterior and that an optimizer step was executed.
Limitations
The script uses a one-dimensional Gaussian simulator as a mechanism-faithful proxy, not the paper's full two-moons or SLCP benchmarks.
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).