garage-bAInd/Platypus2-7B
Platypus2-7B is a 7 billion parameter instruction fine-tuned autoregressive language model developed by Cole Hunter & Ariel Lee, based on the LLaMA2 transformer architecture. It is specifically trained on a STEM and logic-based dataset, Open-Platypus, making it optimized for tasks requiring reasoning and logical understanding. The model supports a 4096 token context length and is designed for English language applications.
Loading preview...
Platypus2-7B: A LLaMA2-based Instruction-Tuned Model
Platypus2-7B is a 7 billion parameter instruction fine-tuned language model built upon the LLaMA2 transformer architecture. Developed by Cole Hunter and Ariel Lee, this model is distinguished by its training on the garage-bAInd/Open-Platypus dataset, which is heavily focused on STEM and logic-based content. This specialized training aims to enhance the model's performance in tasks requiring analytical and reasoning capabilities.
Key Capabilities & Features
- Architecture: Based on the robust LLaMA2-7B transformer.
- Specialized Training: Instruction fine-tuned using LoRA on a STEM and logic-centric dataset,
Open-Platypus. - Context Length: Supports a 4096 token context window.
- Language: Primarily designed for English language applications.
- Performance: Achieves an average score of 45.69 on the Open LLM Leaderboard, with specific scores including 55.2 on ARC (25-shot) and 49.83 on MMLU (5-shot).
Good For
- Applications requiring strong logical reasoning and STEM-related problem-solving.
- Developers looking for a LLaMA2-based model with enhanced instruction-following capabilities in technical domains.
- Research and development in areas benefiting from models trained on specialized, high-quality datasets.
For more in-depth information, including training details and evaluation results, refer to the Platypus paper and project webpage.