AGENT SKILL

lsc-benchmarks

Routes OGB-LSC workflows for PCQM4M, MAG240M, WikiKG90M, and submission helpers.

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

01 前言 Front Matter

name
lsc-benchmarks
license
MIT
metadata
{ "disco-role": "operating" }
description
Routes OGB-LSC workflows for PCQM4M, MAG240M, WikiKG90M, and submission helpers.
disable-model-invocation
true

02 资源文件 5

03 SKILL.md 正文

LSC Benchmarks

Use this subskill for the OGB-LSC family: PCQM4M, PCQM4Mv2, MAG240M, WikiKG90M, WikiKG90Mv2, submission files, and evaluator smoke checks.

Use this subskill when

  • The task names PCQM4M, PCQM4Mv2, MAG240M, WikiKG90M, WikiKG90Mv2, or OGB-LSC.
  • The task asks for test-dev, test-challenge, submission filenames, or evaluator-specific smoke checks.
  • The task mentions only_smiles, to_pyg_hetero_data, split_test, or checkpointed inference helpers.

First decisions

  1. Read references/workflows.md for the dataset families and the main execution flow.
  2. Read references/api-reference.md for the public class names and evaluator expectations.
  3. Read references/submission.md when the task is about output filenames or submission shapes.
  4. Read references/troubleshooting.md when a dataset is deprecated, a checkpoint is missing, or a shape assertion fails.
  5. If you only need to verify evaluator shapes without downloading a dataset, run scripts/lsc-smoke.py.

Main workflow

  • Decide which LSC dataset family the request targets.
  • Check whether the request is a data-loading task, an evaluator task, or a submission-format task.
  • Use only_smiles=True when you only need the molecular strings for the PCQM4M family.
  • Use to_pyg_hetero_data() when you need a PyG hetero representation for MAG240M.
  • Treat WikiKG90M as deprecated and prefer the v2 workflow for new work.

Common routing choices

  • PCQM4M / PCQM4Mv2 -> molecular regression and submission helpers.
  • MAG240M -> large heterogeneous node classification.
  • WikiKG90M / WikiKG90Mv2 -> knowledge-graph completion and top-10 submission arrays.

Optional backend note

The LSC examples may rely on PyG, DGL, or external frameworks such as DGL-KE/SMORE. Those are optional or external to the core OGB package. If they are missing, keep the workflow to the core dataset and evaluator surface unless those extras are explicitly requested.

What not to do here

  • Do not run the full heavyweight benchmark training loops as the runtime answer.
  • Do not treat PCQM4M or WikiKG90M as preferred starting points for new work.
  • Do not depend on the source checkout for the final skill content.

Related references

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