Hugging Face Overview
Model Overview
MoaData/Myrrh_solar_10.7b_3.0 is a 10.7 billion parameter language model developed by Taeeon Park and Gihong Lee at MoAData. This model stands out due to its specialized training methodology and dataset, focusing on the medical domain.
Key Capabilities
- Medical Domain Specialization: The model has undergone DPO (Direct Preference Optimization) training on a proprietary medical dataset, which was custom-built using resources from AI-hub. This focused training aims to enhance its performance and relevance in medical contexts.
- Causal Language Modeling: As a causal language model, it is designed to predict the next token in a sequence, making it suitable for text generation and understanding tasks within its specialized domain.
Training Details
The model was trained using the DPO (Direct Preference Optimization) method, which is known for aligning models with human preferences more effectively. The training dataset is a dpo medical dataset, specifically created by MoAData utilizing resources from AI-hub, indicating a strong emphasis on medical data quality and relevance.
Good For
- Applications requiring specialized knowledge in the medical domain.
- Tasks involving the generation or analysis of medical text.
- Developers looking for a model fine-tuned on a custom medical corpus for improved domain-specific performance.