Overview

sethuiyer/OpenDolphinHermes_Llama2_7B is a 7 billion parameter language model built upon the Llama 2 architecture. It was created using a mergekit SLERP operation, combining the strengths of two distinct models: cognitivecomputations/dolphin-llama2-7b and Tensoic/Llama-2-openhermes.

Key Capabilities & Performance

The primary goal of this merge was to develop a 7B model that performs comparably to the base Llama 2 13B model. Benchmarking on the OpenLLM Leaderboard shows competitive results:

• Average Score: 54.24
• AI2 Reasoning Challenge (ARC): 55.03
• HellaSwag: 78.74
• MMLU: 52.25
• TruthfulQA: 46.10
• Winogrande: 73.16
• GSM8k: 20.17

These scores indicate its proficiency in reasoning, common sense, factual knowledge, and question answering, positioning it as a capable model for various NLP tasks.

Usage

The model utilizes a ChatML prompt template, making it suitable for conversational applications. It can be easily integrated into projects using the Hugging Face transformers library, with example Python code provided for text generation.