README.md

metadata

license: apache-2.0
task_categories:
  - feature-extraction
language:
  - en
tags:
  - code
pretty_name: Blender 5.0 Documentation SFT Dataset
size_categories:
  - 1K<n<10K

Blender 5.0 Documentation SFT Dataset

A comprehensive fine-tuning dataset containing 31,401 Q&A pairs covering the full Blender 5.0 official documentation (user manual + bpy Python API) in chat format. Intended to train LLM for Local use in OpenBlender addon https://pgcrt.github.io/

Dataset Summary

• Total examples: 31,401 Q&A pairs
• Training set: 29,830 examples
• Validation set: 1,571 examples
• Format: Chat format (conversations array with user/assistant roles)
• Source: Official Blender 5.0 documentation (docs.blender.org)

Content

• Full Blender User Manual (2,197 pages)
• Complete bpy Python API Reference (2,063 pages)
• All modules: bpy.context, bpy.data, bpy.ops, bpy.types (1,674+ classes)
• Additional modules: bmesh, mathutils, gpu, freestyle, blf, aud

Usage

from datasets import load_dataset

dataset = load_dataset("json", data_files={
    "train": "train.jsonl",
    "valid": "valid.jsonl"
})

Model Compatibility

Optimized for fine-tuning with instruction-tuned LLMs

Citation

Data sourced from Blender Documentation.

---

license: apache-2.0