← Back to Skills Marketplace
ai-gaoqian

Mlflow Experiment Tracker

by ai-gaoqian · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
40
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install mlflow-experiment-tracker-ai
Description
MLflow 实验追踪智能助手。自动分析实验运行结果、对比超参数配置、 检测过拟合风险、推荐最优模型,为机器学习团队提供端到端的实验管理能力。
README (SKILL.md)

MLflow Experiment Tracker

概述

面向 MLflow 用户的实验分析技能,将原始运行日志转化为可操作的模型优化决策。

核心能力

1. 实验运行分析

  • 自动解析 MLflow Tracking Server 数据
  • 可视化训练曲线(loss/accuracy 趋势)
  • 检测训练异常(震荡、发散、平台期)
  • 识别最佳 checkpoint

2. 超参数对比

  • 多实验横向对比矩阵
  • 超参数重要性排序(基于 SHAP/fANOVA)
  • 推荐下一轮搜索空间
  • 可视化平行坐标图(Parallel Coordinates)

3. 过拟合检测

  • 训练/验证集指标差距分析
  • Early Stopping 最佳时机推荐
  • 学习率调度策略评估
  • 正则化强度适宜性检查

4. 模型选优与注册

  • 多指标加权评分排名
  • 推荐注册到 MLflow Model Registry 的候选模型
  • 生成模型卡(Model Card)文档
  • 版本兼容性检查

5. 实验管理增强

  • 批量重命名和标签管理
  • 实验归档和清理建议
  • 资源消耗分析(GPU 时、内存峰值)
  • 实验复现检查清单

使用方式

分析实验运行: \x3Cexperiment_id>
对比超参数: \x3Cexperiment_ids>
推荐最优模型: \x3Cexperiment_id> \x3Cmetric_name>
检测过拟合: \x3Crun_id>

输出格式

  • 实验分析仪表板(Markdown + 图表)
  • 超参数对比矩阵表
  • 模型推荐报告(含排名理由和部署建议)
  • 过拟合风险热力图

数据底座

基于 MLflow 2.x 官方文档、Optuna/Hyperopt 超参优化最佳实践、Google ML Crash Course、Full Stack Deep Learning 课程内容,覆盖 100+ 常用 ML 指标和 50+ 调参策略。

定价

¥0.50 / 次分析

Usage Guidance
Install only if you are comfortable having the skill help inspect MLflow experiments, runs, and metrics. When using it, give explicit experiment or run IDs and confirm the workspace/project context so the agent does not analyze the wrong MLflow target.
Capability Assessment
Purpose & Capability
The supplied evidence points to MLflow experiment/run analysis behavior, which is coherent with an MLflow-oriented skill and does not show unrelated, destructive, or deceptive capability.
Instruction Scope
The advisory scanner noted broad invocation phrases for experiment, run, and metric identifiers; that can cause accidental use on the wrong target, but no artifact-backed hidden behavior or automatic mutation was shown.
Install Mechanism
No concerning install mechanism, package execution, obfuscated setup, or undeclared dependency behavior was supplied or found during workspace inspection.
Credentials
No evidence shows broad local indexing, credential scraping, background workers, or environment access beyond what would be expected for interacting with user-specified MLflow resources.
Persistence & Privilege
No persistence, privilege escalation, startup hooks, or long-running background behavior was evidenced.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install mlflow-experiment-tracker-ai
  3. After installation, invoke the skill by name or use /mlflow-experiment-tracker-ai
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
MLflow Experiment Tracker 1.0.0 — 首个发布版本! - 支持 MLflow 实验日志的自动解析与训练过程可视化 - 实现多实验超参数对比、重要性排序及搜索空间推荐 - 内置过拟合检测、Early Stopping 建议及正则化评估 - 提供模型选优、注册与模型卡生成、版本兼容性检查 - 增强实验管理能力,包括批量标签、归档和资源分析 - 输出支持仪表板、对比矩阵、模型推荐报告等丰富形式
Metadata
Slug mlflow-experiment-tracker-ai
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Mlflow Experiment Tracker?

MLflow 实验追踪智能助手。自动分析实验运行结果、对比超参数配置、 检测过拟合风险、推荐最优模型,为机器学习团队提供端到端的实验管理能力。 It is an AI Agent Skill for Claude Code / OpenClaw, with 40 downloads so far.

How do I install Mlflow Experiment Tracker?

Run "/install mlflow-experiment-tracker-ai" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Mlflow Experiment Tracker free?

Yes, Mlflow Experiment Tracker is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Mlflow Experiment Tracker support?

Mlflow Experiment Tracker is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Mlflow Experiment Tracker?

It is built and maintained by ai-gaoqian (@ai-gaoqian); the current version is v1.0.0.

💬 Comments