Sakalti/Magro-7b-v1.1

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Dec 12, 2024License:mitArchitecture:Transformer Open Weights Cold

Magro-7b-v1.1 by Sakalti is a 7 billion parameter language model, merged using the TIES method with HuggingFaceH4/zephyr-7b-alpha as its base. This model integrates the characteristics of Sakalti/magro-7B, offering a specialized blend of capabilities. It is designed for general language tasks, leveraging its merged architecture for enhanced performance within a 4096-token context length.

Loading preview...

Overview

Sakalti/Magro-7b-v1.1 is a 7 billion parameter language model created by Sakalti. It was developed using the TIES merge method from mergekit, combining the strengths of existing pre-trained models.

Merge Details

This model utilizes HuggingFaceH4/zephyr-7b-alpha as its foundational base model. The primary model integrated into this merge is Sakalti/magro-7B. The merging process involved specific configurations, including a weight and density of 1 for Sakalti/magro-7B, with normalization and int8_masking enabled, and a bfloat16 dtype.

Key Characteristics

  • Architecture: Merged model based on Zephyr-7b-alpha.
  • Parameter Count: 7 billion parameters.
  • Merge Method: TIES (Trimmed, Iterative, & Self-consistent).
  • Base Model: HuggingFaceH4/zephyr-7b-alpha.
  • Integrated Model: Sakalti/magro-7B.

Potential Use Cases

This model is suitable for applications requiring a 7B parameter model with characteristics derived from its merged components, offering a balance of performance and efficiency for various natural language processing tasks.