sar_reliable_entropy_filter
Use this skill when implementing SAR reliable entropy minimization, entropy-margin filtering, or proxy checks for selected samples.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- sar_reliable_entropy_filter
- description
- Use this skill when implementing SAR reliable entropy minimization, entropy-margin filtering, or proxy checks for selected samples.
02 资源文件 2
03 SKILL.md 正文
Reliable Entropy Filter
When To Use
Use this skill for SAR stable test-time adaptation work that needs: Compute softmax entropy and select reliable samples below the SAR entropy margin. Do not use it as a generic ImageNet-C runner or as permission to read the original SAR repository during recovery.
Inputs
- logits, class count, optional margin
Outputs
- per-sample entropy values, selected indices, mean reliable entropy
Workflow
- Read the caller-provided stream, logits, parameter metadata, or entropy trace.
- Apply the SAR-specific invariant documented in this skill: Entropy is a practical proxy for excluding samples with large noisy gradients during online adaptation.
- Produce structured numeric evidence for recovery logs.
- Keep proxy/reduced results explicitly labeled when the full ImageNet-C runtime is unavailable.
Validation
Run python scripts/reliable_entropy_filter.py when the script has a CLI, then run python -m pytest tests or the Distiller skill validator with --run-tests.
Limitations
This skill captures a reusable SAR mechanism. It does not contain pretrained weights, ImageNet-C data, or original repository code, and it should be combined with other SAR skills for end-to-end recovery.
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).