AGENT SKILL

masked_sparse_optimizer

Apply masked Adam-style sparse updates so only selected per-example unlabeled weights and their optimizer moments change during SSL hyperparameter optimization.

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

01 前言 Front Matter

name
masked_sparse_optimizer
description
Apply masked Adam-style sparse updates so only selected per-example unlabeled weights and their optimizer moments change during SSL hyperparameter optimization.

02 资源文件 2

03 SKILL.md 正文

Masked Sparse Optimizer

Use this skill when per-example unlabeled weights are stored globally but only a minibatch subset receives hypergradients. It preserves the paper contract that untouched examples should not have their Adam moments decayed or otherwise modified.

Do not use this skill for dense parameter vectors where every coordinate receives a gradient at every step.

Inputs

  • Weight dictionary keyed by unlabeled example id.
  • Gradient dictionary for the selected ids only.
  • Optimizer moment state keyed by id.
  • Adam hyperparameters and projection bounds.

Outputs

  • Updated weights with selected ids changed and unselected ids untouched.
  • Updated moment state only for selected ids.
  • Step diagnostics for audit logs.

Workflow

  1. Copy the current weight and optimizer state dictionaries.
  2. For each selected id, update first and second moments using Adam recurrences.
  3. Apply bias correction, step the selected scalar weight, and project to non-negative values.
  4. Leave all unselected ids and moments byte-for-byte unchanged.

Validation

Run python tests/test_masked_adam.py or validate through validate_skill_tree.py --run-tests.

Limitations

This is a scalar per-example optimizer for recovery and skill reuse. It does not replace framework optimizers for dense neural-network parameters.

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