Harish102005/Qwen2.5-Coder-7B-manim
Harish102005/Qwen2.5-Coder-7B-manim is a 7.6 billion parameter Qwen2.5-Coder-7B model fine-tuned by Harish102005 using QLoRA. This specialized model excels at generating Manim (Mathematical Animation Engine) Python code from natural language descriptions. It is specifically optimized for creating 2D mathematical animations and visualizations, making it ideal for educational content and rapid prototyping.
Loading preview...
Model Overview
Harish102005/Qwen2.5-Coder-7B-manim is a specialized 7.6 billion parameter language model, fine-tuned from the Qwen2.5-Coder-7B base model. Its primary function is to translate natural language prompts into executable Manim (Mathematical Animation Engine) Python code. This model was fine-tuned using QLoRA with Unsloth on a dataset of 2,407 examples derived from the 3Blue1Brown Manim dataset, focusing on generating code for mathematical animations.
Key Capabilities
- Manim Code Generation: Converts natural language descriptions into Manim Python code for creating animations.
- Specialized Fine-tuning: Optimized for Manim, particularly for 2D mathematical visualizations and educational content.
- Efficient Inference: Utilizes Unsloth for fast loading and inference, supporting 4-bit quantization.
Use Cases
- Educational Content Creation: Generate animations for math tutorials and scientific visualizations.
- Rapid Prototyping: Quickly create visual content and animation sequences in Manim.
- Learning Manim: Assist users in understanding and generating Manim syntax and animation techniques.
- Content Automation: Facilitate batch generation of animations from textual descriptions.
Limitations
- Primarily designed for 2D Manim animations; complex 3D scenes may be challenging.
- Training data is limited to patterns found in the 3Blue1Brown Manim dataset.
- Advanced Manim features like custom shaders or highly complex mobjects are not fully supported.