Skip to content

Dataset · nvidia

Llama-Nemotron VLM Multimodal Dataset

nvidia/Llama-Nemotron-VLM-Dataset-v1

NVIDIA dataset for training and evaluating the Llama-Nemotron VLM multimodal model.

What's inside

An NVIDIA dataset for multimodal tasks: visual question answering, image-text-to-text, and image-to-text. Metadata indicates a size of 1M to 10M examples, with modality noted as text. The card includes a changelog (OCR image and instruction fixes) and instructions for loading examples via Megatron Energon.

What you can do with it

  • Fine-tuning VLMs for visual question answering
  • SFT for image-to-text models
  • OCR subtask verification
Tasks
Visual question answering, Image text → text, Image → text
Size
1M–10M rows
Downloads this month
3.9K
Likes
173

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Llama-Nemotron-VLM-Dataset-v1")