nbeerbower/bophades-mistral-7B
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Mar 30, 2024License:apache-2.0Architecture:Transformer Open Weights Cold

The nbeerbower/bophades-mistral-7B is a 7 billion parameter language model based on the Mistral architecture, created by nbeerbower through a merge of several pre-trained models using the DARE TIES method. This model integrates diverse capabilities from its constituent models, including those focused on mathematical reasoning and general language understanding. It is designed for applications requiring a compact yet versatile model with a 4096-token context length.

Loading preview...

Model Overview

The nbeerbower/bophades-mistral-7B is a 7 billion parameter language model developed by nbeerbower. It was constructed using the DARE TIES merge method, combining the strengths of multiple pre-trained models into a single, cohesive unit. The base model for this merge was yam-peleg/Experiment26-7B.

Key Characteristics

This model is a composite of several specialized 7B models, suggesting a broad range of potential capabilities. The constituent models include:

  • paulml/NeuralOmniWestBeaglake-7B
  • paulml/OmniBeagleSquaredMBX-v3-7B
  • yam-peleg/Experiment21-7B
  • Kukedlc/NeuralMaths-Experiment-7b
  • Gille/StrangeMerges_16-7B-slerp
  • vanillaOVO/correction_1

The inclusion of models like Kukedlc/NeuralMaths-Experiment-7b indicates a potential emphasis on mathematical reasoning or problem-solving, while others likely contribute to general language understanding and generation. The merge configuration utilized specific density and weight parameters for each model, aiming for an optimized blend of their respective strengths.

Potential Use Cases

Given its merged nature and the specific components, bophades-mistral-7B could be suitable for:

  • General text generation and understanding
  • Tasks requiring numerical or mathematical reasoning
  • Applications where a compact 7B model with diverse capabilities is beneficial
  • Experimentation with merged model architectures