PeterPaker123/Qwen2.5-7B-ViMetaMathQA-Mini
PeterPaker123/Qwen2.5-7B-ViMetaMathQA-Mini is a 7.6 billion parameter Qwen2.5-based language model fine-tuned for mathematical reasoning in Vietnamese. Developed by PeterPaker123, it leverages Flash Attention 2 and BFloat16 precision for efficient training on a 100,000-sample subset of the translated MetaMathQA dataset. This model is specifically optimized to solve mathematical problems and follow instructional prompts in Vietnamese.
Loading preview...
Overview
PeterPaker123/Qwen2.5-7B-ViMetaMathQA-Mini is a specialized 7.6 billion parameter model built upon the Qwen2.5-7B-Instruct architecture. It has been fine-tuned by PeterPaker123 to excel in mathematical reasoning and problem-solving specifically in Vietnamese.
Key Capabilities
- Vietnamese Mathematical Reasoning: Designed to understand and solve mathematical problems presented in Vietnamese.
- Instruction Following: Capable of following Vietnamese instructional prompts related to mathematical logic.
- Efficient Training: Utilized advanced techniques like Flash Attention 2 and BFloat16 precision on NVIDIA H100 hardware for optimized performance.
- Targeted Dataset: Trained on a 100,000-sample subset of the translated MetaMathQA dataset, focusing on high-quality Vietnamese mathematical data.
Intended Use
This model serves as a mathematical assistant for Vietnamese speakers, particularly effective for:
- Solving basic algebra problems.
- Responding to mathematical queries and instructions in Vietnamese.
Limitations
As a preliminary model, it has limitations including a fine-tuning dataset size of 100,000 samples, which may not cover all advanced mathematical nuances. Like other LLMs, it may occasionally exhibit calculation hallucination, requiring users to verify critical results. It is not recommended for mission-critical applications without rigorous human validation.