URajinda/qwen1.5b-myanmar-cpt-final
URajinda/qwen1.5b-myanmar-cpt-final is a 1.5 billion parameter Continual Pre-trained (CPT) model based on URajinda/ShweYon_Qwen2.5-Burmese-1.5B-v1.2.4, featuring a 131072 token context length. This model is specifically optimized for enhancing Burmese (Myanmar) language capabilities, particularly for spoken and formal text patterns. It utilizes LoRA (Low-Rank Adaptation) for training and is specialized for efficient Burmese vocabulary and token processing, making it ideal for Burmese text generation and as a foundation for Burmese instruction tuning.
Loading preview...
Model Overview
URajinda/qwen1.5b-myanmar-cpt-final is a 1.5 billion parameter Continual Pre-trained (CPT) model, building upon the URajinda/ShweYon_Qwen2.5-Burmese-1.5B-v1.2.4 base. It incorporates a modified, custom Burmese-optimized tokenizer from its base model, ensuring efficient token processing for Myanmar script. The model was trained using LoRA (Low-Rank Adaptation) to further refine its understanding and generation of Burmese.
Key Capabilities
- Enhanced Burmese Language Processing: Specifically optimized for both spoken and formal Burmese text patterns.
- Efficient Tokenization: Utilizes a custom tokenizer designed for Burmese vocabulary and script efficiency.
- Foundation for Instruction Tuning: Serves as a strong base for further Supervised Fine-Tuning (SFT) in Burmese.
Good For
- Burmese Text Generation: Excels at producing high-quality text in the Myanmar language.
- Burmese NLP Development: Ideal for developers and researchers working on Burmese natural language processing tasks.
- Language Model Adaptation: Suitable for projects requiring a specialized and efficient Burmese language model.