AGENT SKILL

sac_update_step

Execute one deterministic reduced Soft Actor-Critic critic actor and target update for recovery evidence.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
sac_update_step
description
Execute one deterministic reduced Soft Actor-Critic critic actor and target update for recovery evidence.

02 资源文件 2

03 SKILL.md 正文

SAC Update Step

Use this skill for bounded recovery experiments that must show the SAC update mechanism without full MuJoCo or neural-network training. It consumes a validated replay batch and scalar parameters, then emits losses, changed parameters, and mechanism flags.

Inputs

  • Replay batch with transition dictionaries.
  • Scalar parameters value, q1, q2, policy, and target_value.
  • Hyperparameters gamma, alpha, lr, and tau.

Outputs

  • loss_before and loss_after.
  • params_before and params_after.
  • Mechanism checks for entropy, replay use, twin Q, actor/critic update, target update, and optimizer change.

Workflow

  1. Compute soft value targets using min(q1, q2) - alpha * log_prob.
  2. Compute Q targets from reward plus discounted target value.
  3. Compute actor pressure from alpha * log_prob - min(q1, q2).
  4. Apply a small scalar gradient-like update to value, Q, and policy parameters.
  5. Apply Polyak averaging to the target value parameter.

Validation

Run python tests/test_update.py or validate the full skill tree with the Distiller validator.

Limitations

This is a declared reduced/proxy update and must not be reported as full paper-scale SAC training.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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).

★ 351