labtasker 1.0.0

A task queue system for lab experiments

Make your ML experiment wrapper scripts smarter with...

Install • Tutorial / Demo • Documentation • FAQs • Releases

🌟 Labtasker makes ML experiment wrapper scripts smarter with task prioritization, failure handling, halfway resume and more: just change 1 line of code.

If you like our project, please give us a star ⭐ on GitHub for latest update.

✨ When and Where to Use

TLDR: Replace for loops in your experiment wrapper script with labtasker to enable features like experiment parallelization, dynamic task prioritization, failure handling, halfway resume, and more.

🐳 For detailed examples and concepts, check out the documentation.

🧪️ A Quick Demo

This demo shows how to easily submit task arguments and run jobs in parallel.

It also features an event listener to monitor task execution in real-time and automate workflows, such as sending emails on task failure.

For more detailed steps, please refer to the content in the Tutorial / Demo.

⚡️ Features

• ⚙️ Easy configuration and setup.
• 🧩 Versatile and minimalistic design.
• 🔄 Supports both CLI and Python API for task scheduling.
• 🔌 Customizable plugin system.

🔮 Supercharge Your ML Experiments with Labtasker

• ⚡️ Effortless Parallelization: Distribute tasks across multiple GPU workers with just a few lines of code.
• 🛡️ Intelligent Failure Management: Automatically capture exceptions, retry failed tasks, and maintain detailed error logs.
• 🔄 Seamless Recovery: Resume failed experiments with a single command - no more scavenging through logs or directories.
• 🎯 Real-time Prioritization: Changed your mind about experiment settings? Instantly cancel, add, or reschedule tasks without disrupting existing ones.
• 🤖 Workflow Automation: Set up smart event triggers for email notifications or task workflow based on FSM transition events.
• 📊 Streamlined Logging: All stdout/stderr automatically organized in .labtasker/logs - zero configuration required.
• 🧩 Extensible Plugin System: Create custom command combinations or leverage community plugins to extend functionality.
• 🦾 AI Agent Skills: First-class skill definitions for Claude Code and OpenCode — let your AI assistant decompose and manage experiment scripts automatically.

🛠️ Installation

[!NOTE] You need a running Labtasker server to use the client tools. Simply use the installed Python CLI labtasker-server serve or use docker-compose to deploy the server. See deployment instructions.

1. Install via PyPI

# Install with optional bundled plugins
pip install 'labtasker[plugins]'

2. Install the Latest Version from GitHub

pip install git+https://github.com/luocfprime/labtasker.git

🚀 Quick Start

Use the following command to launch a labtasker server in the background:

labtasker-server serve &

Use the following command to quickly setup a labtasker queue for your project:

labtasker init

Then, use labtasker submit to submit tasks and use labtasker loop to run tasks across any number of workers.

[!TIP] Use AI to help decompose your experiment scripts. Install the Labtasker skill for your agent:

Claude Code — copy docs/docs/guide/skill.md (Claude Code section) to ~/.claude/skills/labtasker/SKILL.md

📚 Documentation

For detailed information on demo, tutorial, deployment, usage, please refer to the documentation.

🔒 License

See LICENSE for details.