Transformer Lab is an open source tool that lets you interact, train, and evaluate LLMs on your own machine or in the cloud (currently Mac or Linux, Windows soon) through a simple GUI.
Try it out here:
https://github.com/transformerlab/transformerlab-app
One of our goals is to make it as easy as possible to get started experimenting with LLMs regardless of what platform you have (even if you don’t have a GPU). You can do everything through the UI but it is built on an API that can be extended using plugins. When you start the app it will try to figure out what kind of system you have and install the right set of initial plugins to get you going optimally.
Some things you can do locally:
- Download and interact with LLMs
- Fine tune and evaluate your own models
- Export models to formats that will allow you to run in smaller memory or without a GPU (e.g. export to GGUF or MLX)
It comes with a library of models and datasets to get going easily, but you can also download any model shared on Hugging Face and build your own training data.
Why we built this: we have been really interested in research and experimentation with LLMs but what we found was that the tooling for going deeper is lacking — Transformer Lab was built to address that problem.
We are a small, unfunded team and we’re excited to build a lot more on top of this platform. So any feedback on how to simplify getting started or how to make this more valuable for experimenting locally would be really appreciated.
