msa-templates
Prepare and validate Chai MSA and template inputs from aligned parquet, A3M, ColabFold server, m8, and staged ColabFold outputs.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- msa-templates
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Prepare and validate Chai MSA and template inputs from aligned parquet, A3M, ColabFold server, m8, and staged ColabFold outputs.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/api-reference.md7.2 KB
- 参考文档references/msa-template-workflows.md9.0 KB
- 参考文档references/troubleshooting.md8.4 KB
- 脚本scripts/stage_colabfold_outputs.py14.1 KB
- 脚本scripts/validate_aligned_pqt.py9.1 KB
03 SKILL.md 正文
MSA Templates
Use this sub-skill when a Chai task involves evolutionary alignments, .aligned.pqt files, A3M conversion, ColabFold MSA/template servers, template hit .m8 files, or reusing staged ColabFold outputs.
Route
- For local MSAs, read
references/msa-template-workflows.mdand validate each.aligned.pqtwithscripts/validate_aligned_pqt.pybefore folding. - For A3M inputs, use
chai-lab a3m-to-pqtfor one query sequence per directory, then validate the generated filename and query row. - For server MSAs/templates, prefer explicit
--use-msa-serverand--use-templates-serveronly when network calls to the ColabFold service are acceptable. - For local templates, prepare a BLAST m8-style table and use
--template-hits-path; useCHAI_TEMPLATE_CIF_FOLDERwhen template CIFs are custom or pre-cached. - For existing ColabFold output trees, run
scripts/stage_colabfold_outputs.pyto create Chai-ready FASTA,msas/*.aligned.pqt, andall_template_hits.m8files.
Primary References
references/msa-template-workflows.mdcovers local MSA preparation, A3M conversion, server-side generation, template inputs, and ColabFold staging.references/api-reference.mdlists CLI/API entry points,.aligned.pqtschema details, m8 columns, and mutual-exclusion rules.references/troubleshooting.mdmaps common MSA/template symptoms to concrete validation and recovery steps.
Bundled Scripts
scripts/validate_aligned_pqt.py --helpchecks required parquet columns, strict source values, query-row placement, aligned lengths, and Chai filename hashes without running inference.scripts/stage_colabfold_outputs.py --helpadapts ColabFold output folders into a Chai-ready local input tree without calling any network service.
Boundaries
- Use
../cli-inference/SKILL.mdfor fullchai-lab foldexecution, GPU/runtime settings, ranking outputs, andrun_inferenceresult handling. - Use
../input-data-formats/SKILL.mdfor FASTA entity syntax, entity names, and chain naming before matching local m8 query IDs. - Use
../restraints-glycans/SKILL.mdfor contact, pocket, covalent, or glycan restraint files.
04 同仓库更多 Skills
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.
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.
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.
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.
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.
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).