Sakura-GalTransl-7B-v3
| Property | Value |
|---|---|
| Parameter Count | 7B |
| License | CC-BY-NC-SA 4.0 |
| Authors | SakuraLLM (sakuraumi and xd2333) |
| Model Hub | Hugging Face |
| Language Pair | Japanese to Simplified Chinese |
What is Sakura-GalTransl-7B-v3?
Sakura-GalTransl-7B-v3 is a specialized language model designed specifically for translating visual novels (Galgame) from Japanese to Simplified Chinese. Built on the Sakura-7B-Qwen2.5-v1.0 architecture and enhanced using GRPO reinforcement, this model represents a significant advancement in machine translation for visual novel content.
Implementation Details
The model requires minimum 6GB VRAM and supports various quantization levels (IQ4_XS, Q4_K, Q5_K, Q6_k) for different memory-performance trade-offs. It's optimized for the GalTransl visual novel translation tool and supports GPT dictionary integration for consistent terminology.
- Maintains formatting elements like inline line breaks, control characters, and ruby annotations
- Balances hardware requirements with translation quality and stability
- Supports batch translation of 7-10 sentences for optimal results
- Recommended temperature setting of 0.6 for v3
Core Capabilities
- Specialized visual novel script translation
- Context-aware pronoun usage
- Terminology consistency through custom glossaries
- High preservation of source formatting and special characters
- Balanced performance on consumer hardware
Frequently Asked Questions
Q: What makes this model unique?
The model's specialization in visual novel translation, with specific optimizations for maintaining script formatting and structure, sets it apart from general-purpose translation models. It's designed to run efficiently on consumer hardware while delivering high-quality translations.
Q: What are the recommended use cases?
The model is specifically designed for translating Japanese visual novels to Simplified Chinese, with particular attention to maintaining script formatting, handling special characters, and ensuring consistent terminology usage through custom glossaries. It's ideal for fan translation projects (non-commercial use only).
