ockerman0/AnubisLemonade-70B-v1
AnubisLemonade-70B-v1 is a 70 billion parameter language model created by ockerman0, resulting from an experimental NuSLERP merge of TheDrummer's Anubis-70B-v1.1 and sophosympatheia's StrawberryLemonade-70B-v1.2. This model aims to balance the distinct qualities of its base models, offering a versatile foundation for various generative AI tasks. With a context length of 32768 tokens, it is designed for users seeking a blend of capabilities from two established 70B models.
Loading preview...
Model Overview
AnubisLemonade-70B-v1 is a 70 billion parameter language model developed by ockerman0. It is an experimental merge of two distinct pre-trained models: TheDrummer's Anubis-70B-v1.1 and sophosympatheia's StrawberryLemonade-70B-v1.2. The primary goal of this merge was to achieve a balanced integration of the strengths and characteristics of both constituent models.
Merge Details
This model was created using the NuSLERP merge method via mergekit. The configuration weighted Anubis-70B-v1.1 at 65% and StrawberryLemonade-70B-v1.2 at 35%, with the merge performed in bfloat16 precision. This specific weighting was chosen to explore a particular blend of their respective qualities.
Recommended Usage
For optimal performance, users are encouraged to experiment with sampling parameters. The developer suggests starting with a Temperature of 1 and a Min-P of 0.025. Feedback on performance and behavior is actively encouraged to help refine understanding of this merged model's capabilities.
Key Differentiator
Unlike single-source models, AnubisLemonade-70B-v1 offers a unique blend derived from two well-regarded 70B models, potentially providing a broader range of emergent capabilities or a more balanced performance profile across different tasks than either base model individually.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.