Joerg Hiller
Could 07, 2025 15:38
NVIDIA introduces Nemotron-CC, a trillion-token dataset for giant language fashions, built-in with NeMo Curator. This progressive pipeline optimizes information high quality and amount for superior AI mannequin coaching.
NVIDIA has built-in its Nemotron-CC pipeline into the NeMo Curator, providing a groundbreaking method to curating high-quality datasets for giant language fashions (LLMs). The Nemotron-CC dataset leverages a 6.3-trillion-token English language assortment from Frequent Crawl, aiming to boost the accuracy of LLMs considerably, in response to NVIDIA.
Developments in Information Curation
The Nemotron-CC pipeline addresses the restrictions of conventional information curation strategies, which regularly discard doubtlessly helpful information attributable to heuristic filtering. By using classifier ensembling and artificial information rephrasing, the pipeline generates 2 trillion tokens of high-quality artificial information, recovering as much as 90% of content material misplaced by filtering.
Progressive Pipeline Options
The pipeline’s information curation course of begins with HTML-to-text extraction utilizing instruments like jusText and FastText for language identification. It then applies deduplication to take away redundant information, using NVIDIA RAPIDS libraries for environment friendly processing. The method consists of 28 heuristic filters to make sure information high quality and a PerplexityFilter module for additional refinement.
High quality labeling is achieved via an ensemble of classifiers that assess and categorize paperwork into high quality ranges, facilitating focused artificial information era. This method permits the creation of numerous QA pairs, distilled content material, and arranged data lists from the textual content.
Impression on LLM Coaching
Coaching LLMs with the Nemotron-CC dataset yields important enhancements. As an illustration, a Llama 3.1 mannequin educated on a 1 trillion-token subset of Nemotron-CC achieved a 5.6-point enhance within the MMLU rating in comparison with fashions educated on conventional datasets. Moreover, fashions educated on lengthy horizon tokens, together with Nemotron-CC, noticed a 5-point increase in benchmark scores.
Getting Began with Nemotron-CC
The Nemotron-CC pipeline is on the market for builders aiming to pretrain basis fashions or carry out domain-adaptive pretraining throughout numerous fields. NVIDIA supplies a step-by-step tutorial and APIs for personalization, enabling customers to optimize the pipeline for particular wants. The combination into NeMo Curator permits for seamless growth of each pretraining and fine-tuning datasets.
For extra info, go to the NVIDIA weblog.
Picture supply: Shutterstock
