DataPilot/Llama3.1-ArrowSE-v0.4

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Jul 24, 2024License:llama3Architecture:Transformer0.0K Warm

DataPilot/Llama3.1-ArrowSE-v0.4 is an 8 billion parameter Llama 3.1-based instruction-tuned language model, specifically enhanced for Japanese language performance. Created using Mergekit, it combines Meta-Llama-3.1-8B-Instruct with elyza/Llama-3-ELYZA-JP-8B and nvidia/Llama3-ChatQA-1.5-8B. This model is optimized for general conversational tasks and question-answering in Japanese, leveraging its merged architecture for improved linguistic nuance.

Loading preview...

Overview

DataPilot/Llama3.1-ArrowSE-v0.4 is an 8 billion parameter instruction-tuned model built upon the Llama 3.1 architecture. Its primary goal is to significantly enhance Japanese language capabilities through a strategic merge of existing models. The model was created using the Mergekit tool, specifically employing the TIES merge method.

Key Capabilities

  • Enhanced Japanese Performance: The model is specifically fine-tuned and merged to improve its understanding and generation of Japanese text.
  • Instruction Following: Inherits strong instruction-following capabilities from its Llama 3.1-8B-Instruct base.
  • Chat and Q&A: Benefits from the inclusion of Llama3-ChatQA-1.5-8B, making it suitable for conversational AI and question-answering tasks.

Merge Details

This model is a merge of three distinct base models, with specific weighting to optimize for Japanese language and conversational abilities:

  • meta-llama/Meta-Llama-3.1-8B-Instruct (weight: 1)
  • elyza/Llama-3-ELYZA-JP-8B (weight: 0.7) - Contributes significantly to Japanese language proficiency.
  • nvidia/Llama3-ChatQA-1.5-8B (weight: 0.15) - Enhances chat and Q&A capabilities.

Good For

  • Applications requiring robust Japanese language generation and comprehension.
  • Building Japanese-speaking chatbots and virtual assistants.
  • General instruction-following tasks where Japanese is the primary language.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p