SkillsAdd

modelscope.cn

funsloth-train

Generate Unsloth training notebooks and scripts. Use when the user wants to create a training notebook, configure fine-tuning parameters, or set up SFT/DPO/GRPO training.

Installation

$ npx skills add https://modelscope.cn
View on skills.sh ↗ Website ↗

Also in this package

Other skills from modelscope.cn · top by installs.

npx skills add https://modelscope.cn
skill-creator 2K frontend-design 1.9K pptx-generator 1.3K minimax-docx 1.2K find-skills 1.1K minimax-xlsx 1.2K minimax-pdf 1.2K docx 942 self-improvement 932 pdf 914

Browse all from modelscope.cn

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 3,869 B

History

  1. First recorded snapshot · 0 installs

SKILL.md

Unsloth Training Notebook Generator

Generate training notebooks for fine-tuning with Unsloth.

Quick Start

Copy and customize the template notebook:

notebooks/sft_template.ipynb

Or use a training script directly:

python scripts/train_sft.py  # Supervised fine-tuning
python scripts/train_dpo.py  # Direct preference optimization
python scripts/train_grpo.py # Group relative policy optimization

Configuration Modes

Ask the user which mode they prefer:

  1. Sensible defaults - Production-ready notebook with recommended settings
  2. Guide me - Walk through each option with explanations
  3. Leave it empty - Notebook with ipywidgets for runtime configuration

Mode 1: Sensible Defaults

Use these production-ready defaults:

Parameter Default Reasoning
Model unsloth/llama-3.1-8b-unsloth-bnb-4bit Good balance
Max seq length 2048 Covers most use cases
Load in 4-bit True 70% VRAM reduction
LoRA rank 16 Good trade-off
Batch size 2 Works on 8GB+ VRAM
Gradient accumulation 4 Effective batch of 8
Learning rate 2e-4 Unsloth recommended
Epochs 1 Often sufficient

Mode 2: Guide Me

Ask questions in order. See [MODELSELECTION.md](references/MODELSELECTION.md) for model options and [TRAININGMETHODS.md](references/TRAININGMETHODS.md) for technique details.

Key Questions

  1. Model family: Llama, Qwen, Gemma, Phi, Mistral, DeepSeek?
  2. Model size: Based on VRAM (see [HARDWAREGUIDE.md](references/HARDWAREGUIDE.md))
  3. Training technique: SFT, DPO, GRPO, ORPO, KTO?
  4. Quantization: 4-bit (recommended), 8-bit, 16-bit?
  5. LoRA rank: 8, 16, 32, 64?
  6. Sequence length: 512, 1024, 2048, 4096?
  7. Batch size: 1, 2, 4, 8?
  8. Learning rate: 1e-5, 5e-5, 2e-4, 5e-4?
  9. Training duration: 1 epoch, 3 epochs, or specific steps?

Mode 3: ipywidgets

Generate a notebook with interactive configuration widgets. Users select options at runtime.

Notebook Structure

Generate notebooks with these sections:

  1. Title and Overview - What the notebook does
  2. Installation - Install Unsloth
  3. Imports and GPU Check - Verify environment
  4. Configuration - All tunable parameters
  5. Load Model - FastLanguageModel.from_pretrained()
  6. Apply LoRA - FastLanguageModel.getpeftmodel()
  7. Load Dataset - Format-appropriate loading
  8. Training - SFTTrainer/DPOTrainer/GRPOTrainer
  9. Save Model - LoRA adapter + merged model
  10. Test Inference - Quick verification

After Generation

Ask where to run training:

  1. Hugging Face Jobs - Cloud GPUs (funsloth-hfjobs)
  2. RunPod - Flexible GPU rentals (funsloth-runpod)
  3. Local - Your own GPU (funsloth-local)

Context to Pass

notebook_path: "./training_notebook.ipynb"
model_name: "unsloth/llama-3.1-8b-unsloth-bnb-4bit"
dataset_name: "mlabonne/FineTome-100k"
technique: "SFT"
lora_rank: 16
max_seq_length: 2048
batch_size: 2
learning_rate: 2e-4
num_epochs: 1

Bundled Resources

  • [notebooks/sfttemplate.ipynb](notebooks/sfttemplate.ipynb) - Ready-to-use SFT template
  • [scripts/trainsft.py](scripts/trainsft.py) - SFT script template
  • [scripts/traindpo.py](scripts/traindpo.py) - DPO script template
  • [scripts/traingrpo.py](scripts/traingrpo.py) - GRPO script template
  • [references/MODELSELECTION.md](references/MODELSELECTION.md) - Model recommendations
  • [references/HARDWAREGUIDE.md](references/HARDWAREGUIDE.md) - VRAM requirements
  • [references/TRAININGMETHODS.md](references/TRAININGMETHODS.md) - SFT vs DPO vs GRPO