p3_materialization
Materialize PromptSource/P3-style template collections into prompted dataset records with template identity, metadata propagation, coverage counts, and rendering diagnostics.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- p3_materialization
- description
- Materialize PromptSource/P3-style template collections into prompted dataset records with template identity, metadata propagation, coverage counts, and rendering diagnostics.
02 资源文件 2
- 脚本scripts/materialize_p3.py3.5 KB
- 脚本tests/test_p3_materialization.py2.3 KB
03 SKILL.md 正文
P3 Prompted Dataset Materialization
Use this skill when converting a small PromptSource-style template collection and dataset examples into prompted training or evaluation records without reading the original PromptSource repository.
Inputs
dataset_name: dataset identifier.subset_name: optional subset/configuration identifier.templates: JSON list of template records withid,name,template,answer_choices, and metadata fields.examples: JSON list of dataset examples.
Outputs
A JSON object with records, summary, metadata_reports, and errors.
Workflow
- Validate each template's metadata with the metadata-quality helper.
- Render each template over each example using the template-rendering helper.
- Emit only non-skipped, successful prompted records.
- Preserve dataset identity, template id/name, input, target, and metadata in every record.
- Summarize template count, example count, produced records, skipped rows, errors, and represented metrics.
Validation
Run:
python tests/test_p3_materialization.py
python scripts/materialize_p3.py --dataset-name synthetic_snli --examples-json '[{"premise":"A dog runs","hypothesis":"An animal moves","label":0}]' --templates-json '[{"id":"t1","name":"nli","template":"{{premise}} Question: {{hypothesis}}? ||| {{answer_choices[label]}}","answer_choices":["Yes","No","Maybe"],"metrics":["Accuracy"],"valid_outputs_stated":true,"reference":"paper example"}]'
Limitations
This skill materializes supplied examples. It does not download benchmark datasets, launch the PromptSource UI, or reproduce large-scale P3 training.
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).