garage-bAInd/Stable-Platypus2-13B
TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:Aug 5, 2023License:cc-by-nc-sa-4.0Architecture:Transformer0.0K Open Weights Cold
Stable-Platypus2-13B is a 13 billion parameter auto-regressive language model developed by garage-bAInd, merging Platypus2-13B and StableBeluga-13B. Built on the LLaMA 2 transformer architecture, this model is instruction fine-tuned using STEM and logic-based datasets. It is designed for general English language tasks, demonstrating competitive performance across various benchmarks including ARC, HellaSwag, and MMLU.
Loading preview...
Model Overview
Stable-Platypus2-13B is a 13 billion parameter instruction-tuned language model, created by merging garage-bAInd/Platypus2-13B and stabilityai/StableBeluga-13B. It is based on the LLaMA 2 transformer architecture and is designed for English language tasks.
Key Capabilities
- Instruction Following: Fine-tuned with a focus on STEM and logic-based datasets, enhancing its ability to follow complex instructions.
- Merged Architecture: Combines the strengths of two distinct 13B models, Platypus2-13B (trained by Cole Hunter & Ariel Lee) and StableBeluga-13B (trained by StabilityAI).
- Benchmark Performance: Achieves an average score of 54.25 on the Open LLM Leaderboard, with specific scores including 62.71 on ARC (25-shot), 82.29 on HellaSwag (10-shot), and 58.3 on MMLU (5-shot).
Good For
- General English Language Tasks: Suitable for a wide range of applications requiring strong language understanding and generation.
- Research and Development: Provides a solid base for further fine-tuning or experimentation, particularly for tasks benefiting from its STEM and logic-oriented training.
- Comparative Analysis: Useful for researchers interested in merged model architectures and their performance characteristics against individual base models.