Edentns/DataVortexS-10.7B-dpo-v1.8

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:10.7BQuant:FP8Ctx Length:4kLicense:cc-by-nc-sa-4.0Architecture:Transformer0.0K Open Weights Warm

Edentns/DataVortexS-10.7B-dpo-v1.8 is a 10.7 billion parameter language model developed by Kwangseok Yang, Jeongwon Choi, Seunghyun Choi, and Hyoseok Choi. Built upon the megastudy/M-SOLAR-10.7B-v1.3 base model, it is instruction-tuned using the Alpaca (Chat) format. This model demonstrates strong performance on Korean language benchmarks, including Ko LM Eval Harness and Ko-LLM-Leaderboard, making it suitable for Korean-centric conversational AI applications.

Loading preview...

DataVortexS-10.7B-dpo-v1.8 Overview

DataVortexS-10.7B-dpo-v1.8 is a 10.7 billion parameter language model developed by a team including Kwangseok Yang and Seunghyun Choi. It is built on the megastudy/M-SOLAR-10.7B-v1.3 base model and has been instruction-tuned following the Alpaca (Chat) format, making it well-suited for conversational tasks.

Key Capabilities & Performance

  • Instruction Following: Utilizes the Alpaca (Chat) instruction format for clear and structured interactions.
  • Korean Language Proficiency: Demonstrates strong performance across various Korean language benchmarks.
    • Ko LM Eval Harness: Achieves an average of 0.789423 on 50-shot evaluations, with notable scores in kobest_sentineg (0.949596) and kobest_boolq (0.923032).
    • Ko-LLM-Leaderboard: Scores an average of 58.15, with specific results like 66.68 on Ko-HellaSwag and 61.04 on Ko-CommonGen V2.
  • Implementation: Provides a chat_template instruction format for easy integration with transformers library, including example Python code for generation.

Good For

  • Developing conversational AI applications in Korean.
  • Tasks requiring strong understanding and generation of Korean text.
  • Researchers and developers looking for a performant Korean LLM in the 10B parameter range.

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