skills.rest
thinking-machines-lab
thinking-machines-labOFFICIAL
Explore all skills by this author →
WebsiteGitHub

logging

Capture training metrics, logtrees, and traces into metrics.jsonl and logtree.json files.

4.0k|507|Updated Jul 14, 2025
Search Tags:#debugging#logging#metrics#profiling#tracing#training#logtree
One-click install
npx skills add https://github.com/thinking-machines-lab/tinker-cookbook --skill logging-thinking-machines-lab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: logging
Source: https://github.com/thinking-machines-lab/tinker-cookbook/tree/main/.claude/skills/logging
Command: npx skills add https://github.com/thinking-machines-lab/tinker-cookbook --skill logging-thinking-machines-lab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Structured logging, metrics collection, and trace data are essential for understanding and debugging training runs. This skill outlines how to generate, organize, and interpret logs, logtree reports, and profiling traces to accelerate debugging and performance tuning.

Core Features & Use Cases

  • Logging outputs and metrics for every training run (metrics.jsonl, config.json, and log files)
  • Structured rollouts with logtree and trace data to diagnose training behavior
  • Integration with tooling (e.g., W&B) for centralized monitoring and reproducibility

Quick Start

Initialize the logging utilities in your training script and run a small example to generate and inspect metrics, logs, and traces.

Frequently Asked Questions about logging

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I track and debug ML training runs with structured logs?▼

Yes, W&B integration is supported for centralized monitoring and reproducibility. You can capture training outputs and metrics across iterations, writing them to metrics.jsonl and logtree files while syncing data to W&B for analysis.

What is a logtree and how does it help diagnose training behavior?▼

You need a logging API and trace utilities to write metrics.jsonl, logtree.json, and trace events to log paths. These tools capture and organize training outputs, requiring no additional dependencies to start logging metrics.

Can I use this logging approach for both reinforcement learning and supervised learning workflows?▼

Start by initializing logging utilities in your training script to generate metrics.jsonl, config.json, and log files. Run a small example to inspect metrics, logs, and traces, ensuring your logging API correctly captures training outputs.

What's the best way to organize metrics and profiling traces for performance tuning?▼

The best way to organize metrics and traces is writing them to structured formats like metrics.jsonl and logtree.json. This captures training outputs systematically, enabling efficient debugging and performance tuning across iterations.

Related Skills

Explore More Software Engineering Skills→
VilnaCRM-OrgVilnaCRM-Org
5

observability-instrumentation

Emit EMF-based metrics from domain events via event subscribers.

Official
Advanced
niltonfredericoniltonfrederico
2

har-analysis

Analyze HAR files for performance metrics, API patterns, and security findings.

Community
Advanced
Leo1349Leo1349

run-experiment

Deploy machine learning experiments to local or remote GPU servers via SSH.

Community
Advanced
graphsignalgraphsignal
241

graphsignal-profiler

Profile production inference workloads via sidecar process without modifying source code.

Official
Advanced