Nemotron-Pretraining-Code-v3

Nemotron-Pretraining-Code-v3 Dataset Description: The Nemotron-Pretraining-Code-v3 dataset is part of the Nemotron Pretraining Data collection of pretraining datasets. Designed for the NVIDIA Nemotron 3 family of LLMs, this dataset is intended to improve the coding capabilities of LLMs. The Nemotron-Pretraining-Code-v3 dataset contains the metadata corresponding to the raw source-code update to our Nemotron-Pretraining-Code-v2 and Nemotron-Pretraining-Code-v1 source-code corpora. It adds 146 M new files (173 B tokens) from GitHub (with a cut-off date of September 30, 2025). **This dataset contains only incrementally new files and is meant to be used jointly with the v2 and v1 corpora.** The dataset contains a single split Nemotron-Code-Metadata with the following columns: **commit_id:** The first seven characters of the Git commit hash **rel_path:** The file path of the content relative to the repository root directory **language:** The detected programming language of the content This dataset is ready for commercial use. Dataset Owner(s): NVIDIA Corporation Dataset Creation Date: 05/18/2026 Version: Nemotron-Pretraining-Code-v3 Previous Version(s): Nemotron-Pretraining-Code-v2 Nemotron-Pretraining-Code-v1 Relationship to Previous Version(s): This is an **extension** of the previously released datasets. They are meant to be used together. License/Terms of Use: The **Nemotron-Pretraining-Code-v3** dataset is licensed under the Creative Commons Attribution 4.0 International License (CC-BY-4.0). Intended Usage: The Nemotron-Pretaining-Code-v3 Dataset is intended to be used by the community to continue to improve open models. Dataset Characterization Data Collection Method: Automated Labeling Method: Automated Dataset Format Modality: Text Format: Parquet Dataset Quantification Record Count: 146.3M samples Measurement of Total Data Storage: 8.2 GB Reference(s): If you use our dataset in your research, please cite our NVIDIA Nemotron 3 Ultra tech report. Ethical Consider