fb_greedy_policy
Extract no-planning greedy policies from FB forward embeddings and a projected reward vector.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- fb_greedy_policy
- description
- Extract no-planning greedy policies from FB forward embeddings and a projected reward vector.
02 资源文件 2
- 脚本scripts/greedy_policy.py1.6 KB
- 脚本tests/test_greedy_policy.py206 B
03 SKILL.md 正文
FB Greedy Policy Extraction
Use this skill when z_R has already been estimated and a policy must be derived by direct FB scoring. Do not use it to plan, search, or update embeddings for the new reward.
Inputs
- Forward embeddings
Findexed as state-action rows. - Task vector
z_R. - Number of states and actions, with deterministic row order
state * action_count + action.
Outputs
- Q-value table
F(s,a,z_R)^T z_R. - Greedy action for each state with deterministic lowest-index tie breaking.
Workflow
- Compute dot products for all state-action rows.
- Group scores by state and select the maximum action.
- Record ties and avoid any environment planning or reward-specific optimization.
- Hand the policy to recovery or evaluation modules.
Validation
Run python scripts/greedy_policy.py --self-test. Tests verify known best actions and deterministic tie handling.
Limitations
The script assumes finite action enumeration and precomputed forward embeddings.
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).