unsloth-studio
Recipe card from the charly-jupyter plugin (Images — the deployable catalog).
unsloth-studio
Section titled “unsloth-studio”Unsloth Studio web UI for LLM fine-tuning with GPU acceleration.
Box Properties
Section titled “Box Properties”| Property | Value |
|---|---|
| Base | nvidia |
| Candies | agent-forwarding, unsloth-studio, notebook-finetuning, dbus, charly |
| Platforms | linux/amd64 |
| Ports | 8888, 8000 |
| Registry | ghcr.io/opencharly |
Candy Composition
Section titled “Candy Composition”The unsloth-studio candy is a Tier 2 environment-owner meta-layer that:
- Owns the pixi.toml (fine-tuning Python environment)
- Composes two Tier 1 sub-candies via
candy: [llama-cpp, unsloth] - Defines the supervisord service for the Studio web UI
Build order: pixi environment → llama-cpp (binaries) → unsloth (vLLM 0.19 wheel + unsloth pip + torch.compile patch) → supervisord config
Full Candy Stack
Section titled “Full Candy Stack”fedora→nvidia(CUDA base)pixi→python→supervisord(transitive)unsloth-studio— Tier 2 meta-layer (owns pixi.toml, service config)llama-cpp— llama.cpp binaries (Tier 1, viacandy:)unsloth— vLLM 0.19 + unsloth pip install + torch.compile patch (Tier 1, viacandy:)
| Port | Service | Protocol |
|---|---|---|
| 8888 | Unsloth Studio UI | HTTP |
| 8000 | vLLM API server | HTTP |
Volumes
Section titled “Volumes”| Name | Path | Purpose |
|---|---|---|
| models | ~/.cache/huggingface | HuggingFace model cache |
| workspace | /workspace | Training data and outputs |
Quick Start
Section titled “Quick Start”charly box build unsloth-studiocharly config unsloth-studiocharly start unsloth-studio# Open http://localhost:8888Key Candies
Section titled “Key Candies”/charly-jupyter:unsloth-studio— Studio web UI service + pixi.toml (Tier 2)/charly-jupyter:llama-cpp— llama.cpp binaries (Tier 1 sub-candy)/charly-jupyter:unsloth— vLLM 0.19 + unsloth fine-tuning + torch.compile patch (Tier 1 sub-candy)/charly-jupyter:notebook-finetuning— 37 Unsloth fine-tuning notebooks provisioned into workspace volume/charly-distros:nvidia— GPU runtime and CDI device auto-detection (base)/charly-distros:cuda— CUDA toolkit and libraries (via nvidia base)/charly-infrastructure:dbus-layer— session bus for desktop notifications/charly-tools:charly— in-containercharlyCLI toolchain (scripting, service management)/charly-distros:agent-forwarding— SSH/GPG/direnv agent forwarding
Related Boxes
Section titled “Related Boxes”/charly-distros:nvidia— parent (GPU without Studio)/charly-jupyter:jupyter-ml— alternative ML UI with JupyterLab + CRDT MCP (same Tier 1 sub-candies)/charly-languages:python-ml— ML libraries without any UI/charly-jupyter:jupyter— legacy Jupyter with ML (shares port 8888)- CachyOS variant —
cachyos.unsloth-studiois the CachyOS GPU sibling (built on thecachyos.nvidiaGPU base) in theopencharly/distro-cachyossubmodule. See/charly-distros:cachyos.
Verification
Section titled “Verification”After charly start:
charly status unsloth-studio— container runningcharly service status unsloth-studio— all services RUNNINGcurl -s -o /dev/null -w '%{http_code}' http://localhost:8888— Studio HTTP returns 200
When to Use This Skill
Section titled “When to Use This Skill”MUST be invoked when the task involves the unsloth-studio box, LLM fine-tuning via web UI, or Unsloth Studio deployment. Invoke this skill BEFORE reading source code or launching Explore agents.
Related
Section titled “Related”/charly-image:image— image family umbrella (candy:image entries — those carryingbase:/from:— incharly.yml, build/validate/inspect/list)/charly-build:build— the embedded build vocabulary (distros, builders, init-systems)