nvidia/Llama-Nemotron-VLM-Dataset-v1

Llama-Nemotron-VLM数据集是一个高质量的视觉语言模型（VLM）后训练数据集，支持NVIDIA发布的Llama-3.1-Nemotron-Nano-VL-8B-V1。该模型专注于文档智能，能够理解和查询图像，并总结文档内容。数据集包括合成和重新注释的数据集，用于OCR和VQA任务，并旨在供商业使用。The Llama-Nemotron-VLM Dataset is a compilation of high-quality VLM post-training datasets that support NVIDIA’s release of Llama-3.1-Nemotron-Nano-VL-8B-V1, a vision language model specialized in document intelligence. The model can understand, query, and summarize images, extracting diverse information from complex documents with precision on a single GPU. The dataset includes synthetic and re-annotated datasets for OCR and VQA tasks, and is intended for commercial use.