ndavidson/ultraphi-slerp
TEXT GENERATIONConcurrency Cost:1Model Size:3BQuant:BF16Ctx Length:2kPublished:Mar 20, 2024License:apache-2.0Architecture:Transformer Open Weights Cold

ndavidson/ultraphi-slerp is a 3 billion parameter language model created by ndavidson, formed by merging rhysjones/phi-2-orange-v2 and liminerity/Phigments12 using the slerp method. This model leverages the strengths of its base components, offering a compact yet capable solution for general text generation tasks within a 2048-token context window. Its merged architecture aims to combine diverse linguistic patterns and knowledge from its constituent models.

Loading preview...

ultraphi-slerp Overview

ultraphi-slerp is a 3 billion parameter language model developed by ndavidson, created through a merge of two distinct models: rhysjones/phi-2-orange-v2 and liminerity/Phigments12. This merge was performed using the slerp (Spherical Linear Interpolation) method via LazyMergekit, a technique designed to combine the weights of multiple models to achieve enhanced or specialized capabilities.

Key Capabilities

  • Merged Intelligence: Combines the knowledge and linguistic patterns from phi-2-orange-v2 and Phigments12.
  • Efficient Size: At 3 billion parameters, it offers a balance between performance and computational efficiency.
  • Standard Context Window: Supports a context length of 2048 tokens, suitable for a variety of text generation tasks.

Good for

  • General Text Generation: Ideal for tasks requiring coherent and contextually relevant text outputs.
  • Experimentation with Merged Models: Provides a practical example of a slerp-merged model for researchers and developers.
  • Resource-Constrained Environments: Its smaller size makes it suitable for deployment where computational resources are a consideration.