tabular_usfa_recovery
Run a bounded Trip-MDP proxy experiment that validates USFA TD learning and GPI transfer mechanics.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- tabular_usfa_recovery
- description
- Run a bounded Trip-MDP proxy experiment that validates USFA TD learning and GPI transfer mechanics.
02 资源文件 2
- 脚本scripts/trip_usfa_recovery.py10.9 KB
- 脚本tests/test_trip_usfa.py980 B
03 SKILL.md 正文
Tabular USFA Recovery
Use this skill when full USFA navigation training is unavailable but soft-mode recovery permits a declared mechanism-faithful proxy. The experiment must still execute real TD updates, GPI candidate search, and metric computation.
Inputs
- Training task weights and candidate encodings.
- Test preference grid.
- Output directory for recovery artifacts.
- Paths to the generated USFA helper skills.
Outputs
recovery_result.jsonwith metric and mechanism checks.logs/training_trace.jsonwith loss and parameter changes.logs/generated_data_item.jsondescribing the Trip-MDP proxy item.logs/generated_skill_invocations.jsonshowing helper skill usage.
Workflow
- Build the paper's Trip-MDP-style feature table.
- Initialize a tabular USFA table with trainable entries for sampled encodings.
- Run vector TD updates using the linear-reward successor-feature helper.
- Evaluate interpolated tasks with the GPI helper.
- Save metrics and mechanism checks from the executable script.
Validation
Run:
python tests/test_trip_usfa.py
Limitations
The script is a reduced proxy, not a full 3D navigation reproduction. It is valid only when the run is in soft recovery mode and full runtime blockers are recorded.
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).