Weyaxi/Nous-Hermes-2-SUS-Chat-34B-Slerp
Weyaxi/Nous-Hermes-2-SUS-Chat-34B-Slerp is a 34 billion parameter language model created by Weyaxi through a slerp merge of Nous-Hermes-2-Yi-34B and SUS-Chat-34B. This model leverages the strengths of its base components, offering a robust foundation for general conversational AI tasks. It is designed for chat applications, providing a balanced performance across various benchmarks, including reasoning and common sense understanding.
Loading preview...
Nous-Hermes-2-SUS-Chat-34B-Slerp Overview
This model, developed by Weyaxi, is a 34 billion parameter language model created using the mergekit tool with a slerp (spherical linear interpolation) merging method. It combines the Nous-Hermes-2-Yi-34B and SUS-Chat-34B models, aiming to leverage the complementary strengths of both. The base model for this merge was Yi-34B.
Key Capabilities & Features
- Merged Architecture: Benefits from the combined knowledge and capabilities of two distinct 34B models.
- Chat Optimization: Primarily designed for conversational AI, supporting prompt templates like ChatML and Human-Assistant formats.
- Performance: Achieves an average score of 74.06 on the Open LLM Leaderboard, with notable scores in:
- HellaSwag (10-Shot): 84.97
- MMLU (5-Shot): 77.00
- Winogrande (5-shot): 83.58
- Quantized Versions Available: Optimized versions for various hardware are provided by TheBloke, including GPTQ, GGUF, and AWQ.
When to Use This Model
This model is suitable for developers looking for a capable 34B model for general-purpose chat and conversational applications. Its balanced performance across reasoning, common sense, and language understanding benchmarks makes it a versatile choice for tasks requiring robust dialogue generation and comprehension.