nayohan/llama3-instrucTrans-enko-8b

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:8kPublished:May 13, 2024License:llama3Architecture:Transformer0.0K Warm

The nayohan/llama3-instrucTrans-enko-8b is an 8 billion parameter instruction-tuned causal language model based on Llama-3-8B-it, developed by nayohan. This model is specifically fine-tuned for English to Korean translation tasks, leveraging custom translation datasets. It demonstrates strong performance in various English-Korean translation benchmarks, making it suitable for applications requiring accurate cross-lingual communication.

Loading preview...

Model Overview

The nayohan/llama3-instrucTrans-enko-8b is an 8 billion parameter model built upon the Llama-3-8B-it architecture, specifically optimized for English to Korean translation. It was fine-tuned using specialized datasets, including nayohan/aihub-en-ko-translation-1.2m and nayohan/translate_corpus_313k, to enhance its performance in this specific domain.

Key Capabilities

  • High-Quality English-Korean Translation: Excels at translating English text into Korean, as demonstrated by its leading performance across multiple benchmarks.
  • Instruction-Following: Designed to follow translation instructions effectively, making it suitable for direct application in translation pipelines.
  • Competitive Benchmarking: Achieves the highest average score (16.13) across Aihub, Flores, iwslt, and ko_news_eval40 English-Korean translation datasets, outperforming other models like EEVE-10.8b-it, KULLM3, and nhndq-nllb in several categories.

Good For

  • Automated Translation Systems: Ideal for integrating into applications requiring robust English to Korean translation capabilities.
  • Research and Development: Useful for researchers exploring fine-tuning techniques for low-resource language pairs or specific translation tasks.
  • Content Localization: Can be employed for localizing English content into Korean with high fidelity.

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