Accelerating Multilingual Language Model for Excessively Tokenized Languages
This addresses a specific inefficiency for users of multilingual LLMs in non-Roman alphabetic languages, representing an incremental improvement.
The paper tackles the problem of inefficient text generation in non-Roman alphabetic languages due to token fragmentation in multilingual LLMs, and shows that a targeted fine-tuning framework increases generation speed by 1.7 times while maintaining performance.
Recent advancements in large language models (LLMs) have remarkably enhanced performances on a variety of tasks in multiple languages. However, tokenizers in LLMs trained primarily on English-centric corpora often overly fragment a text into character or Unicode-level tokens in non-Roman alphabetic languages, leading to inefficient text generation. We introduce a simple yet effective framework to accelerate text generation in such languages. Our approach involves employing a new language model head with a vocabulary set tailored to a specific target language for a pre-trained LLM. This is followed by fine-tuning the new head while incorporating a verification step to ensure the model's performance is preserved. We show that this targeted fine-tuning, while freezing other model parameters, effectively reduces token fragmentation for the target language. Our extensive experiments demonstrate that the proposed framework increases the generation speed by a factor of 1.7 while maintaining the performance of pre-trained multilingual models on target monolingual tasks.