amkyawdev/amk-coder-v2

Model Overview

amkyawdev/amk-coder-v2 is a 1.5 billion parameter language model, fine-tuned by amkyawdev from the Qwen/Qwen2.5-Coder-1.5B base model. Its primary distinction lies in its Myanmar-localized coding agent capabilities, making it unique among code generation models. The model was fine-tuned using LoRA (PEFT) with FP16 mixed precision on a custom Myanmar localized coding agent dataset.

Key Capabilities

• Myanmar Language Code Generation: Excels at generating code from instructions provided in the Myanmar language.
• Coding Assistance: Designed to act as a coding agent, providing support for various programming tasks.
• Qwen2.5-Coder Architecture: Benefits from the robust architecture of the Qwen2.5-Coder series, adapted for localized use.
• Tool Access Integration: Utilizes a ChatML structure that supports tool access, as demonstrated in its prompt format for expert Myanmar AI coding agents.

Good for

• Localized Coding Tools: Ideal for developing coding assistants and tools that cater to Myanmar-speaking developers.
• Instruction-Tuned Code Generation: Suitable for tasks requiring code generation based on specific instructions, particularly in a bilingual (Myanmar/English) context.
• Research and Development: Useful for exploring localized LLM applications in programming and natural language processing.

Limitations

• Not intended for production deployment without further fine-tuning or rigorous testing.
• May generate syntactically incorrect code or not adhere to security best practices.
• Myanmar language support, while a focus, may still have limitations compared to English.