Dataset · nvidia
Data for training NVIDIA embedding models
nvidia/embed-nemotron-dataset-v1
Multilingual compilation of datasets for fine-tuning embeddings for retrieval, reranking, similarity, and classification.
What's inside
This is a compilation of high-quality fine-tuning datasets released along with the llama-embed-nemotron-8b model from NVIDIA. The model is designed for retrieval, reranking, semantic similarity, and classification. The dataset is multilingual, with a size of 10 to 100 million examples; specific sources and format are not disclosed in the card.
What you can do with it
- Fine-tuning embedding models for retrieval and reranking
- Training sentence-similarity and text-classification models
- Fine-tuning llama-embed-nemotron-8b or compatible embeddings
- Tasks
- Text retrieval, Text ranking, Sentence similarity, Text classification
- Languages
- Multilingual
- Size
- 10M–100M rows
- Downloads this month
- 921
- Likes
- 113
Models trained on it
- nvidia/llama-embed-nemotron-8bEmbeddings 11.1K
- dmis-lab/UltraBERT-Embed-150MSentence similarity 1.4K
- mradermacher/llama-embed-nemotron-8b-GGUFEmbeddings 470
- PhysShell/UltraBERT-Embed-150MSentence similarity 24
- suryatmodulus/UltraBERT-Embed-150MSentence similarity 24
- TheRealXeko/llama-embed-nemotron-8b-benchEmbeddings 15
How to load it
from datasets import load_dataset
ds = load_dataset("nvidia/embed-nemotron-dataset-v1")