data-preparation
Routes BasicTS dataset layout, raw conversion, and fixture validation workflows for forecasting, classification, imputation, and BLAST data.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- data-preparation
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Routes BasicTS dataset layout, raw conversion, and fixture validation workflows for forecasting, classification, imputation, and BLAST data.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/data-formats.md4.4 KB
- 参考文档references/troubleshooting.md3.2 KB
- 脚本scripts/make_tiny_forecasting_dataset.py3.1 KB
- 脚本scripts/validate_basicts_dataset.py7.4 KB
03 SKILL.md 正文
data-preparation
Use this sub-skill when the task is about dataset files, raw-data conversion, or validating that a BasicTS dataset folder matches the expected layout.
Typical triggers
- "Validate a BasicTS dataset folder"
- "How do I create a tiny forecasting fixture?"
- "What files does BasicTS expect for UEA classification?"
- "Why is the BLAST memmap layout failing?"
- "How do the data-preparation scripts work?"
What this sub-skill covers
- forecasting dataset layout used by
BasicTSForecastingDataset - imputation dataset layout used by
BasicTSImputationDataset - UEA classification layout used by
UEADataset - BLAST memmap shard layout used by
BLAST - safe validation of folder/file layouts
- safe creation of tiny forecasting fixtures for smoke use
Read these bundled references first
references/data-formats.mdfor the on-disk schemas.references/troubleshooting.mdfor missing-file and shape problems.scripts/validate_basicts_dataset.pyfor read-only validation.scripts/make_tiny_forecasting_dataset.pyfor a tiny fixture generator.
Route here when the user asks for
- raw dataset conversion
- split-file naming conventions
- timestamp array requirements
- classification label file layout
- BLAST memmap shard or shape validation
- a tiny dataset folder for CPU smoke tests
Route elsewhere when the user asks for
- launcher commands, checkpoint paths, or run execution →
training-evaluation - model forward contracts or output shapes →
model-development - callbacks, metrics, scalers, or taskflow hooks →
pipeline-extension
Working guidance
- Identify the task family first: forecasting, classification, imputation, or BLAST.
- Validate the folder layout before trying to run a trainer against it.
- Check whether timestamps are required before claiming a fixture is usable.
- Use the tiny fixture helper when you need a self-contained smoke dataset.
When to read the helper scripts
- Run
scripts/validate_basicts_dataset.pyto check an existing dataset folder. - Run
scripts/make_tiny_forecasting_dataset.pyto create a minimal forecasting fixture with timestamps andmeta.json.
Why this sub-skill exists
BasicTS data workflows are mostly about file layout and split conventions. Keeping those rules separate from training and model contracts makes it easier for future agents to validate datasets without opening the original repository scripts.
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).