AGENT SKILL

molecule-data-prep

Prepare molecular and protein-ligand data for DGL-LifeSci by validating SMILES/CSV inputs, constructing graphs, selecting featurizers and datasets, and applying supported split/evaluation helpers.

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

01 前言 Front Matter

name
molecule-data-prep
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Prepare molecular and protein-ligand data for DGL-LifeSci by validating SMILES/CSV inputs, constructing graphs, selecting featurizers and datasets, and applying supported split/evaluation helpers.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

Molecule Data Prep

Use this sub-skill when a task needs DGL-LifeSci data ingestion or graph preparation before model training or inference:

  • Convert SMILES or RDKit molecules into DGL graphs with dgllife.utils constructors.
  • Choose atom/bond featurizers and check feature field names and dimensions.
  • Load custom CSV or unlabeled SMILES inputs with MoleculeCSVDataset or UnlabeledSMILES.
  • Select built-in MoleculeNet, Alchemy, or PDBBind-style dataset classes without writing training loops.
  • Split datasets and use support utilities such as Meter and EarlyStopping for prepared data pipelines.
  • Validate tiny fixtures offline with the bundled input validator.

Start Here

  1. Read references/data-formats.md to decide the expected input schema and dataset class.
  2. Read references/api-reference.md for verified constructor signatures, feature fields, graph shapes, and splitter gotchas.
  3. Run scripts/validate_molecule_inputs.py --help and then validate the user's tiny CSV or TXT fixture before building a full pipeline.
  4. If validation fails, use references/troubleshooting.md to diagnose invalid SMILES, missing labels, import errors, cache mismatches, or split failures.

Quick Commands

Validate a labeled CSV and report graph shape summaries:

python scripts/validate_molecule_inputs.py \
  --input molecules.csv --format csv --smiles-column smiles --tasks task1,task2 \
  --require-labels --graph bigraph --max-rows 50

Validate an inference TXT file without labels:

python scripts/validate_molecule_inputs.py \
  --input smiles.txt --format txt --graph complete --max-rows 25

Routing Boundaries

Stay in this sub-skill for data schemas, graph construction, featurizers, splits, dataset classes, and tiny fixture validation.

Route elsewhere when the task asks for:

  • Property model architecture, batching, loss functions, or full train/eval loops: use property-prediction.
  • Pretrained GIN checkpoints, load_pretrained, or model-zoo selection: use model-zoo-pretraining.
  • WLN reaction center/ranking datasets or reaction-specific atom-map handling: use reaction-prediction.
  • Binding-affinity model architecture details after PDBBind-style complex graphs are prepared: use ../binding-affinity/SKILL.md.

Bundled Resources

  • references/api-reference.md: verified DGL-LifeSci signatures, graph constructors, featurizers, splitters, Meter, and EarlyStopping notes.
  • references/data-formats.md: CSV, TXT, SMILES, built-in dataset, and PDBBind-style data expectations.
  • references/troubleshooting.md: import/install, optional dependency, invalid data, feature-shape, split, cache, and helper failure fixes.
  • scripts/validate_molecule_inputs.py: safe offline validator for CSV/TXT SMILES fixtures and optional graph shape summaries.

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