AGENT SKILL

model-extension

Inspect and extend Graphormer fairseq models, tasks, criterions, and architectures.

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

01 前言 Front Matter

name
model-extension
license
MIT
metadata
{ "disco-role": "operating" }
description
Inspect and extend Graphormer fairseq models, tasks, criterions, and architectures.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

Graphormer model-extension

Use this sub-skill when a future agent must inspect, register, or extend Graphormer fairseq user-dir components: models, model architectures, tasks, criterions, Graphormer encoder internals, Graphormer3D interfaces, or GraphMLP-style custom models.

Route elsewhere

  • Training command schedules, dataset-specific fairseq-train recipes, and GPU/runtime choices belong to the fairseq-training sub-skill.
  • Dataset sources, split contracts, custom dataset modules, and collation/preprocessing details belong to the datasets-and-customization sub-skill.
  • Pretrained checkpoint selection, fine-tuning output-layer choices, and evaluation loops belong to the pretrained-and-evaluation sub-skill.
  • Distributional Graphormer (DiG) diffusion internals belong to the distributional-graphormer sub-skill.

Runtime resources

  1. Read references/model-api-reference.md for verified registry names, API signatures, architecture defaults, import behavior, model outputs, and criterion/task compatibility notes.
  2. Read references/extension-recipes.md for self-contained recipes to add a model, architecture, task, or criterion in a Graphormer fairseq user-dir.
  3. Read references/troubleshooting.md when registries are missing, imports fail, shapes do not match criterions, or Graphormer3D tensors are rejected.
  4. Run scripts/summarize_graphormer_registries.py to safely import a user-dir and summarize fairseq registry entries without training, downloading data, or loading checkpoints.

Safe registry check

python scripts/summarize_graphormer_registries.py --user-dir <graphormer-package-dir> --format text
python scripts/summarize_graphormer_registries.py --user-dir <graphormer-package-dir> --format json --require-complete

<graphormer-package-dir> is the directory that contains Graphormer's models/, tasks/, and criterions/ packages. The script only imports Python modules and reads fairseq registries.

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