We present a novel approach to data preparation for developing multilingual Indic large language model. Our meticulous data acquisition spans open-source and proprietary sources, including Common Crawl, Indic books, news articles, and Wikipedia, ensuring a diverse and rich linguistic representation. For each Indic language, we design a custom preprocessing pipeline to effectively eliminate redundant and low-quality text content. Additionally, we perform deduplication on Common Crawl data to address the redundancy present in 70% of the crawled web pages. This study focuses on developing high-quality data, optimizing tokenization for our multilingual dataset for Indic large language models with 3B and 7B parameters, engineered for superior performance in Indic languages. We introduce a novel multilingual tokenizer training strategy, demonstrating our custom-trained Indic tokenizer outperforms the state-of-the-art OpenAI Tiktoken tokenizer, achieving a superior token-to-word ratio for Indic languages.
@article{arxiv.2407.12481,
title = {Krutrim LLM: A Novel Tokenization Strategy for Multilingual Indic Languages with Petabyte-Scale Data Processing},
author = {Rahul Kumar and Shubham Kakde and Divyansh Rajput and Daud Ibrahim and Rishabh Nahata and Pidathala Sowjanya and Deepak Kumarr and Gautam Bhargava and Chandra Khatri},
journal= {arXiv preprint arXiv:2407.12481},
year = {2025}
}