Skip to content
Get packs

SwanLab Experiment Tracking

Lets the agent set up SwanLab to log, compare and visualise ML training runs locally or self-hosted.

Source Download skill
Repository
Orchestra-Research/AI-Research-SKILLs
Stars
14,195
Licence
MIT
Files
3 files

What it does

When installed, the agent knows how to instrument machine learning training code with SwanLab, an open-source experiment tracker. It can add code that records metrics, hyperparameter configs, images, audio, text and custom charts for each run, and can set up a local or self-hosted dashboard so results are viewed in a browser rather than only in a managed cloud service. It also covers hooking SwanLab into frameworks such as PyTorch, HuggingFace Transformers, PyTorch Lightning and Fastai.

Great for

  • ML engineers who want to track and compare training runs without relying on a managed SaaS
  • Researchers who need local or self-hosted logging for private or offline experiments
  • Developers adding experiment logging to an existing PyTorch or Transformers training script

Skip it if

  • Teams already standardised on another tracker (such as Weights & Biases or MLflow) who only need help with that tool, since this skill covers SwanLab only
  • Non-technical users with no training code to instrument, because the skill is about writing and running Python logging code

Try asking

  • “Add SwanLab logging to my PyTorch training loop so train and val loss are recorded each epoch”
  • “Set up local-only SwanLab tracking in ./swanlog and show me how to open the dashboard with swanlab watch”
  • “Configure my HuggingFace Trainer to report to SwanLab with a run name”
  • “Log sample images and a loss curve chart to my SwanLab run”
Step-by-step guide: how to use SwanLab Experiment TrackingInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 3 original files, the MIT licence and an attribution note.

    swanlab.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip swanlab.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip swanlab.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Use mode="local" with a logdir when you do not want data sent to any server.
  • Store the SWANLAB_API_KEY in an environment variable rather than pasting it into code or chat.
  • Review the generated training code before running it, since the agent will execute scripts and install packages in your environment.

What's inside

  • references/integrations.md
  • references/visualization.md
  • SKILL.md

Redistributed unmodified from Orchestra-Research/AI-Research-SKILLs under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.