ztx-exp-manager

star 5

Manage ML/LLM experiment workflows with ztxexp. Use when the agent needs to build parameter configurations, run experiments, validate run artifacts, analyze results, or troubleshoot failed runs.

ztxtech By ztxtech schedule Updated 3/2/2026

name: ztx-exp-manager description: Manage ML/LLM experiment workflows with ztxexp. Use when the agent needs to build parameter configurations, run experiments, validate run artifacts, analyze results, or troubleshoot failed runs.

ztx-exp-manager Skill(中文)

核心规则

  1. 优先使用 ExperimentPipeline 组织实验,不手写零散调度脚本。
  2. 实验函数契约固定为:exp_fn(ctx: RunContext) -> dict | None
  3. 成功判定必须使用:run.json.status == "succeeded"
  4. 产物分工:
    • 最终指标:return dict -> metrics.json
    • 过程指标:ctx.log_metric(...) -> metrics.jsonl
    • 业务文件:写入 artifacts/

常用流程

  1. 先用 grid/variants/random_search 构建配置。
  2. 再执行 pipeline.run(exp_fn, mode=...)
  3. ResultAnalyzer 聚合、导出和清理。

常用命令

  • pip install -U ztxexp
  • ztxexp init-vibe
  • python -m pytest

排障清单

  1. 先看 run.json.status
  2. 失败时查看 error.log
  3. 曲线缺失时检查 ctx.log_metric(...) 是否调用。

ztx-exp-manager Skill (English)

Core Rules

  1. Prefer ExperimentPipeline for workflow orchestration.
  2. Keep experiment contract: exp_fn(ctx: RunContext) -> dict | None.
  3. Determine success strictly by run.json.status == "succeeded".
  4. Artifact responsibilities:
    • Final metrics: return dict -> metrics.json
    • Step metrics: ctx.log_metric(...) -> metrics.jsonl
    • Business outputs: write into artifacts/

Typical Workflow

  1. Build configs via grid/variants/random_search.
  2. Execute with pipeline.run(exp_fn, mode=...).
  3. Aggregate and clean results using ResultAnalyzer.

Common Commands

  • pip install -U ztxexp
  • ztxexp init-vibe
  • python -m pytest

Troubleshooting

  1. Check run.json.status first.
  2. Read error.log on failure.
  3. Verify ctx.log_metric(...) for missing curves.
Install via CLI
npx skills add https://github.com/ztxtech/ztxexp --skill ztx-exp-manager
Repository Details
star Stars 5
call_split Forks 0
navigation Branch main
article Path SKILL.md
More from Creator