Skip to content

notebook-finetuning

Version 2026.144.1443
Repo superproject

Unsloth fine-tuning notebook collection Provisions Unsloth fine-tuning notebooks into the workspace volume’s finetuning/ subdirectory at deploy time. Includes Unsloth_Setup, Vision_Training, SFT, GRPO, DPO, Reward, RLOO, and QLoRA variants. Data-only — no packages, no services. Consumed by jupyter-ml-notebook for an Unsloth-ready JupyterLab image.

This candy’s plan: — the runnable spec charly check executes against a live deployment. check: steps are idempotent probes; run: steps change state.

Intent Step
check the finetuning subdirectory is provisioned in the workspace
check the Unsloth_Setup producer notebook is present
check the Unsloth_Setup producer notebook is staged in the image data tree
check the notebooks.yaml manifest cataloguing the collection is staged
check a substantial collection of training notebooks is staged, not just the setup notebook
agent-check the notebooks span the documented Unsloth training methods (SFT, GRPO, DPO, Reward, RLOO, QLoRA) as runnable examples
check file=/data/workspace/finetuning
check file=/data/workspace/finetuning/03_SFT_Training_Qwen.ipynb
check file=/data/workspace/finetuning/08_QLoRA_Multi_Adapter_Qwen_Think.ipynb