diffdock
Route DiffDock docking, web UI, training/data preparation, and benchmark evaluation tasks to focused repo-specific guidance.
- License 许可
- MIT
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- diffdock
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Route DiffDock docking, web UI, training/data preparation, and benchmark evaluation tasks to focused repo-specific guidance.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/install-and-runtime.md4.0 KB
- 参考文档references/repo-provenance.md1.6 KB
- 参考文档references/repo-routing-metadata.json325 B
- 参考文档references/troubleshooting.md2.4 KB
03 SKILL.md 正文
DiffDock
Use this repo skill when a user is working with DiffDock: diffusion-based small-molecule docking to protein structures, DiffDock-L inference, the Gradio UI, training/data preparation, or benchmark evaluation.
Start Here
- Read install-and-runtime.md for setup routes, dependency families, backend expectations, and safe preflight checks.
- Use scripts/check_runtime_environment.py to inspect a candidate runtime without launching docking, training, or benchmarks.
- Route to the focused sub-skill below, then use its bundled references and helper scripts before running expensive commands.
- Read troubleshooting.md when failures involve imports, CUDA/Torch/PyG, RDKit/ProDy/ESM/OpenFold, model checkpoints, network downloads, data paths, or GNINA.
- Check repo-provenance.md before deciding whether this skill is current for a checkout.
Routes
- Docking prediction: Use docking-inference for single-complex or batch CSV inference, input validation, command construction, model/config overrides, output files, and confidence-ranked poses.
- Web interface: Use web-ui for the Gradio app, single-complex UI inputs, launch/deploy issues, and downloaded output zip inspection.
- Training and data: Use training-data for score/confidence model training plans, dataset layouts, split files, ESM embeddings, cache checks, and checkpoint compatibility.
- Benchmarks and metrics: Use evaluation-benchmarks for PDBBind, BindingMOAD/DockGen, PoseBusters evaluation, RMSD/confidence metrics, GNINA post-processing, and vendored
spyrmsdchecks.
Repository Model
DiffDock is a script-style research repository rather than a package with console entry points. Future agents should use the bundled command builders to assemble commands, then run those commands only in a user-provided DiffDock runtime checkout or project context that contains the expected DiffDock modules, data, checkpoints, and dependencies.
The generated skill is self-contained for planning, validation, troubleshooting, and command construction. It does not bundle model weights, processed benchmark datasets, GNINA, ESM models, or the full DiffDock source tree.
Safe Defaults
- Prefer PDB protein inputs over sequence folding for lightweight inference smoke tests.
- Use command builders and validators first; they do not import the heavy DiffDock runtime stack.
- Treat full inference, Gradio launch, ESM extraction, training, benchmark evaluation, GNINA, and model downloads as potentially long-running or network/backend dependent.
- Keep user data paths explicit; do not assume the original example or dataset paths exist.
- Preserve
model_parameters.ymlbeside checkpoints when moving trained score or confidence models.
Required References
- install-and-runtime.md covers setup, Docker/conda expectations, optional dependencies, and environment checks.
- troubleshooting.md covers cross-cutting failure modes shared by sub-skills.
- repo-routing-metadata.json is consumed by DisCo's managed repo-skills-router during import.
- repo-provenance.md records the source snapshot and evidence baseline.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.