Plain-AyI
← All projects
Training & fine-tuning

Liger Kernel

Speed-ups from LinkedIn that make training large AI models faster and lighter on memory.

Research-backedNew this week

What it does

Liger Kernel is a set of optimized building blocks for training large language models. They replace standard pieces of a model's training code with faster versions written in Triton, a language for programming graphics cards.

Its makers say this speeds up training and cuts memory use, which matters for fine-tuning popular models like Llama, Mistral and Gemma. It plugs into existing PyTorch and Hugging Face training setups.

Who it's for

Machine-learning engineers who train or fine-tune large models on GPUs.

Can I use it without coding?

Developers only. It is a Python library used inside training code. New to this? Start here.

Can I run it?

Cost
FreeNothing to pay or sign up for.
Runs on
Linux, Server
Computer
Works best on a powerful computer or one with a graphics card (GPU)
Privacy
Runs on your own device: your data stays with you
API key
Not needed

Try it

Using Claude Code, Cursor or another AI coding assistant? Paste the prompt and it will walk you through the setup.New to this?

Before you install

  • Only download from the official GitHub page. Copies with similar names exist for many popular projects.
  • License (BSD-2-Clause): Permissive: free to use, change and share, including for business.
  • More on staying safe

Where people are talking about it