DireDreadlord/GemCod-Jade-270M
DireDreadlord/GemCod-Jade-270M is a 270 million parameter Gemma3-based causal language model developed by DireDreadlord. It is specifically fine-tuned for accurate and quick code snippet and long-form code generation across major programming languages. This model also offers rudimentary Q/A and subject matter expert capabilities on code-related subjects, designed to run efficiently on laptop-grade GPUs.
Loading preview...
GemCod-Jade-270M Overview
GemCod-Jade-270M is a compact, 270 million parameter code generation model built on the Gemma3 architecture. Developed by DireDreadlord, it is an upgrade from the previous Aquamarine model, offering enhanced long-form code generation and improved explainability of generated snippets while maintaining similar inference times and space requirements.
Key Capabilities
- Code Generation: Specializes in generating both code snippets and long-form code from natural language prompts.
- Multi-language Support: Capable of generating code in all major programming languages.
- Code-related Q&A: Provides basic question-answering and subject matter expertise on coding topics.
- Efficiency: Its small size allows for comfortable execution on laptop-grade GPUs.
Training Details
The model was fine-tuned using Supervised Fine-Tuning (SFT) on the magicoder-glaive-code-instruct dataset, which contains approximately 350,000 rows templated with a custom .jinja chat format. Training was conducted for 3,000 steps on an RTX 3050 (4GB VRAM).
Intended Use
This model is primarily intended for code snippet and long-form code generation from natural language inputs. It is suitable for developers looking for a lightweight, efficient model for coding assistance and code-related inquiries.