Skip to content

Dataset · nvidia

Nemotron VLM dataset v2

nvidia/Nemotron-VLM-Dataset-v2

Nearly triple the high-quality samples of v1 for training VLM on image and video QA.

What's inside

The card continues Llama Nemotron VLM Dataset V1 with 3 million samples; v2 contains nearly three times as many high-quality samples. The main focus is three directions, though they are not listed in the available snippet. Metadata specifies visual-question-answering, image-text-to-text, and video-text-to-text tasks; modality is text. Language and format are not specified in the card.

What you can do with it

  • Fine-tuning VLM for visual question answering
  • SFT for multimodal model image-text-to-text
  • Evaluating or fine-tuning on video-text-to-text
Tasks
Visual question answering, Image text → text, Video text → text
Size
1M–10M rows
Downloads this month
11.6K
Likes
99

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-VLM-Dataset-v2")