arcee-ai/patent-evol-merge

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kPublished:Apr 17, 2024Architecture:Transformer Featherless Exclusive Cold

The arcee-ai/patent-evol-merge is a 7 billion parameter language model created by arcee-ai, merged from a Llama-2-7B base with Patent-Instruct-7b, Orca-2-7b, and Barcenas-Orca-2-7b using the task arithmetic method. This model is designed to combine the strengths of its constituent models, potentially enhancing performance across various instruction-following and specialized tasks, particularly those related to patent information given the inclusion of Patent-Instruct-7b. It processes a context length of 4096 tokens.

Loading preview...

Model Overview

The arcee-ai/patent-evol-merge is a 7 billion parameter language model developed by arcee-ai. It was constructed using the task arithmetic merge method, leveraging the mergekit tool to combine several pre-trained models. The base model for this merge is a Llama-2-7B-fp16 variant.

Merge Composition

This model integrates capabilities from three distinct models into its architecture:

  • /root/evol_merge_storage/input_models/Patent-Instruct-7b_60368649
  • /root/evol_merge_storage/input_models/Orca-2-7b_2312263870
  • /root/evol_merge_storage/input_models/Barcenas-Orca-2-7b_1478912867

The merging process involved applying specific weights to different layer ranges (0-8, 8-16, 16-24, 24-32) of each contributing model, indicating a fine-grained approach to combining their learned representations. The Patent-Instruct-7b component suggests a specialization in patent-related understanding or generation.

Key Characteristics

  • Architecture: Merged model based on Llama-2-7B.
  • Parameter Count: 7 billion parameters.
  • Context Length: Supports a context window of 4096 tokens.
  • Merge Method: Utilizes the task arithmetic method for combining models, allowing for a weighted integration of different model strengths.

Potential Use Cases

Given its composition, this model is likely well-suited for:

  • Instruction Following: Benefiting from the Orca-2-7b and Barcenas-Orca-2-7b components.
  • Patent-related Tasks: Potentially excelling in tasks requiring understanding or generation of patent information due to the inclusion of Patent-Instruct-7b.
  • General Text Generation: Leveraging the strong base capabilities of Llama-2-7B.