ynklab/Qwen2.5-7B-Sep_0c1t
ynklab/Qwen2.5-7B-Sep_0c1t is a 7.6 billion parameter language model developed by ynklab, fine-tuned from Qwen/Qwen2.5-7B-Instruct. This model is specifically optimized for multilingual chunk-level machine translation, supporting translation between English and nine other languages. It utilizes a fixed-range chunking approach for document-level translation, processing source chunks into target chunks without preceding context. The model is designed for high-quality, length-consistent machine translation of document segments.
Loading preview...
Model Overview
ynklab/Qwen2.5-7B-Sep_0c1t is a 7.6 billion parameter model, fine-tuned from Qwen/Qwen2.5-7B-Instruct, specifically for multilingual chunk-level machine translation. It is the Sep_0c1t variant, meaning it translates individual source chunks into target chunks without relying on prior contextual chunks. This model was developed as part of the Doc2FRC project, focusing on length-consistent document-level translation using fixed-range chunking.
Key Capabilities
- Multilingual Translation: Supports translation between English and German, Spanish, French, Italian, Korean, Dutch, Portuguese, Russian, and Chinese in both directions.
- Chunk-Level Translation: Optimized for translating text segments (chunks) within a document, trained on chunks ranging from 256 to 512 tokens.
- Fixed-Range Chunking: Designed to work with a fixed-range chunking methodology, as detailed in the associated Doc2FRC GitHub repository.
- Full-Parameter Fine-tuning: Achieved through full-parameter supervised fine-tuning on 2,004,084 examples derived from the DocBlocks dataset.
Recommended Use Cases
This model is ideal for developers and researchers requiring high-quality, consistent machine translation of documents by processing them in fixed-size chunks. It is particularly suited for scenarios where document-level context management is handled externally, allowing the model to focus on accurate chunk-to-chunk translation. For detailed implementation, including document chunking and inference scripts, refer to the Doc2FRC repository.