Table of contents

unsloth-studio — Unsloth managed UI

A managed web UI for Unsloth fine-tuning — config, launch, and monitor runs from the browser, no Python required. Built on top of the official unsloth/unsloth:stable image.

Image tag

docker.io/manvarharsh/unsloth-studio:cuda12

What’s in this image

  • Base: unsloth/unsloth:stable (upstream Unsloth image)
  • Unsloth Studio web UI
  • All Unsloth dependencies
  • JupyterHub and Jupyter Notebook (port 8888)
  • OpenSSH server

Default ports

PortService
22SSH
8000JupyterHub / Unsloth Studio
8888Jupyter Notebook

Use cases

  • Point-and-click LoRA / QLoRA fine-tunes
  • Browser-driven dataset prep and run management
  • Quick iteration without writing training scripts
  • Pair with the Unsloth notebook workflow on Podstack

Environment variables

VariableDescription
ENABLE_SSHEnable SSH server
ENABLE_UNSLOTH_STUDIOStart the Studio UI (defaults to true)
PODSTACK_API_URLBackend URL for token validation
SSH_PUBLIC_KEYPublic key for SSH

Persistence

Mount at /data. The Studio writes runs, configs, and outputs under /data/.

See also