Local AI Engineering with Ollama

Fine-Tuning in the Browser with Unsloth Studio

Everything so far ran from a Python script. Unsloth also ships a browser interface called Unsloth Studio, where you load a model, point it at a dataset, and run the fine-tune from a UI instead of code. It runs the same training underneath, so the concepts you learned still apply; you are just driving them through a screen rather than a script.

Install Studio

Studio installs from its own script. We install a specific version by cloning the repository, checking out the release tag, and installing from that local copy:

# Clone the repo
git clone \
    https://github.com/unslothai/unsloth.git

# Change directory
cd unsloth

# Checkout the release
git checkout v0.1.44-beta

# Install
./install.sh --local

The --local flag installs Studio from the checked-out source instead of pulling the latest from PyPI, which is how you pin it to a known version. The script installs into its own environment and may ask for permission to install system build tools. Let it finish before moving on.

Start the Server

Studio runs as a local web server. Start it and tell it which network interface to listen on:

# Start the server
unsloth studio -H 0.0.0.0 -p 8888

-H 0.0.0.0 exposes the server on all network interfaces, which you need when Studio runs on a remote machine and you want to reach it from your laptop. -p 8888