elyza/ELYZA-japanese-Llama-2-7b

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Aug 28, 2023License:llama2Architecture:Transformer0.1K Open Weights Cold

ELYZA-japanese-Llama-2-7b is a 6.27 billion parameter Llama 2-based causal language model developed by elyza, specifically pre-trained to enhance its Japanese language capabilities. This model extends the foundational Llama 2 architecture with improved proficiency in Japanese, making it suitable for applications requiring strong Japanese language understanding and generation. It offers both base and instruct versions, with a context length of 4096 tokens.

Loading preview...

ELYZA-japanese-Llama-2-7b: Enhanced Japanese Llama 2 Model

ELYZA-japanese-Llama-2-7b is a 6.27 billion parameter language model built upon the Llama 2 architecture, developed by elyza. Its primary distinction lies in its additional pre-training specifically for Japanese language capabilities, aiming to extend Llama 2's proficiency in this domain. The model is available in several variants, including base and instruct-tuned versions, with some variants featuring an expanded vocabulary size.

Key Capabilities

  • Strong Japanese Language Performance: Optimized for tasks requiring robust understanding and generation of Japanese text.
  • Llama 2 Foundation: Benefits from the underlying architecture and general capabilities of the Llama 2 model family.
  • Multiple Variants: Offers both a base model and an instruction-tuned model (-instruct) for conversational and instruction-following applications.
  • Context Length: Supports a context window of 4096 tokens.

Good for

  • Developers and researchers focusing on Japanese NLP tasks.
  • Applications requiring a Llama 2-based model with enhanced Japanese proficiency.
  • Building Japanese-speaking assistants or chatbots using the instruct-tuned versions.