arcee-ai/Patent-Base-InternLM2-7B-Ties

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Apr 5, 2024Architecture:Transformer Cold

The arcee-ai/Patent-Base-InternLM2-7B-Ties is a 7 billion parameter language model created by arcee-ai, merged using the TIES method. It combines NousResearch/Llama-2-7b-hf, chargoddard/internlm2-7b-llama, and arcee-ai/Patent-Base-7b. This model is specifically designed to leverage the strengths of its constituent models, particularly for tasks related to patent data and general language understanding.

Loading preview...

Model Overview

The arcee-ai/Patent-Base-InternLM2-7B-Ties is a 7 billion parameter language model developed by arcee-ai. It was created using the TIES merge method, which combines multiple pre-trained language models to leverage their individual strengths. The base model for this merge was NousResearch/Llama-2-7b-hf.

Key Capabilities

This model integrates knowledge from three distinct sources:

  • NousResearch/Llama-2-7b-hf: Provides a strong general language understanding foundation.
  • chargoddard/internlm2-7b-llama: Contributes to broader linguistic capabilities.
  • arcee-ai/Patent-Base-7b: Likely enhances performance on tasks involving patent-specific terminology and structures, given its name.

The TIES merge method, as described in the original paper, allows for the selective combination of model parameters, aiming to create a more robust and specialized model.

Good For

  • Applications requiring a blend of general language understanding and specialized knowledge, particularly in domains that might benefit from patent-related data.
  • Research into model merging techniques, specifically the TIES method.
  • Use cases where combining the strengths of Llama 2, InternLM2, and a patent-specific base model is advantageous.