Skip to main content

Data upload

See the API Reference for details and additional options.

Creating an embedding dataset​

The following minimal example allows you to interact with your embeddings dataset in Atlas.

from nomic import atlas
import numpy as np

num_embeddings = 10000
embeddings = np.random.rand(num_embeddings, 512)

dataset = atlas.map_data(embeddings=embeddings)
print(dataset)

This dataset will contain 10,000 random embeddings. You can interact with it in Nomic Atlas as organized by the nomic-project-v1 model by navigating to the browser link.

Creating a text dataset​

from nomic import atlas
import pandas

news_articles = pandas.read_csv('https://raw.githubusercontent.com/nomic-ai/maps/main/data/ag_news_25k.csv')

dataset = atlas.map_data(data=news_articles, indexed_field='text')
print(dataset)

This dataset will contain 25,000 news articles embedded with the default Nomic Embed Text model.

Creating an image dataset​

from nomic import atlas
from datasets import load_dataset

dataset = load_dataset('cifar10', split="train")

images = dataset["img"]
data = [{"label": label} for label in dataset["label"]]
dataset = atlas.map_data(blobs=images, data=data)

This dataset will map the CIFAR10 dataset with the default Nomic Embed Vision model.