AGENT SKILL

molecular-workflows

Plan TorchDrug molecular property prediction, molecular pretraining and finetuning, molecule generation, and retrosynthesis workflows.

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

01 前言 Front Matter

name
molecular-workflows
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Plan TorchDrug molecular property prediction, molecular pretraining and finetuning, molecule generation, and retrosynthesis workflows.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

Molecular Workflows

Use this sub-skill when a user needs to build or debug a TorchDrug workflow for small molecules or chemical reactions.

When To Use

  • Build property prediction pipelines for molecule datasets such as ClinTox, BACE, or custom SMILES/label datasets.
  • Plan molecular self-supervised pretraining with InfoGraph, AttributeMasking, or related pretraining tasks, then finetune with PropertyPrediction.
  • Set up ZINC250k molecule generation with GCPNGeneration or GraphAF-style AutoregressiveGeneration, including PPO finetuning for QED or penalized logP.
  • Set up USPTO50k retrosynthesis with G2Gs: center identification, synthon completion, and the combined Retrosynthesis beam-search pipeline.
  • Interpret molecular metrics, generated SMILES, beam-search candidates, and common chemistry-specific failures.

Route Elsewhere

  • Raw data.Graph, data.Molecule, packed batches, masks, collators, and custom data object construction belong in the graph-data sub-skill.
  • GNN layer internals, custom model/layer design, compiled extensions, and representation output shapes belong in the layers-and-extensions sub-skill.
  • Generic core.Engine checkpointing, save/load, logging, scheduler, distributed settings, and config serialization belong in the training-engine sub-skill.
  • Protein structure, protein function, residue, sequence, and contact workflows belong in the protein-workflows sub-skill.
  • Knowledge graph reasoning and entity/relation ranking workflows belong in the knowledge-graphs sub-skill.

References

  • Property and pretraining covers ClinTox/BACE property prediction, custom molecule datasets, split choices, feature compatibility, GIN, InfoGraph, AttributeMasking, and finetuning.
  • Generation and retrosynthesis covers ZINC250k GCPN/GraphAF setup, PPO objective caveats, USPTO50k reaction/synthon modes, G2Gs training, and beam-search output interpretation.
  • Troubleshooting covers dataset downloads, RDKit and invalid SMILES, feature mismatches, criterion/metric choices, generation validity and resampling, retrosynthesis split seeds, and compute expectations.

Safe Planner

Run python scripts/plan_molecular_workflow.py --workflow property, pretrain, generation, or retrosynthesis from this sub-skill directory to print a no-download, no-training checklist and minimal API skeleton.

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