Maternion/qwen-manim-coder-2

Qwen2.5-Coder-14B Fine-tuned for Manim Code Generation

This model is a fine-tuned version of Qwen/Qwen2.5-Coder-14B-Instruct specialized for generating mathematical animation code using ManimCE v0.19.0.

Training Details

• Base Model: Qwen/Qwen2.5-Coder-14B-Instruct
• Training Method: LoRA (Low-Rank Adaptation)
• Dataset: bespokelabs/bespoke-manim (900 high-quality examples)
• Training Date: August 2025
• Trainable Parameters: 68,812,800 (0.46% of base model)

Usage

This model generates Manim animation code from natural language descriptions:

Example prompts:

• "Create a Manim animation showing the quadratic formula"
• "Animate the Pythagorean theorem with a right triangle"
• "Show the limit definition of a derivative"

Model Type

This repository contains LoRA adapter weights that need to be merged with the base model for inference.