G

Gemma 2 Llama Swallow 2b It V0.1

Developed by tokyotech-llm
The Gemma-2-Llama-Swallow series is built through continued pre-training of the gemma-2 model, significantly enhancing Japanese language processing capabilities while retaining original English proficiency.
Downloads 61
Release Time : 4/21/2025

Model Overview

A multilingual large language model based on the Gemma 2 model, focusing on enhancing Japanese processing capabilities while maintaining English proficiency. Suitable for text generation tasks.

Model Features

Enhanced Bilingual Capabilities
Significantly improves Japanese language processing while retaining original English proficiency.
Large-scale Pre-training
Continued pre-training using approximately 200 billion tokens of mixed data including Japanese web corpora and Japanese-English Wikipedia entries.
Instruction Fine-tuned Version
Provides an instruction fine-tuned version with supervised fine-tuning (SFT) using specially constructed synthetic data for Japanese.
Multiple Parameter Sizes
Offers three model variants with 2B, 9B, and 27B parameter sizes.

Model Capabilities

Japanese Text Generation
English Text Generation
Multi-turn Dialogue
Instruction Understanding and Execution

Use Cases

Dialogue Systems
Japanese Customer Service Bot
Used for building Japanese customer service dialogue systems
Content Generation
Japanese Article Creation
Generates Japanese articles, reports, and other content
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase