metadata
license: bsd-3-clause
configs:
- config_name: no-vectors
data_files: no-vectors/*.parquet
default: true
- config_name: aws-titan-embed-text-v2
data_files: aws/titan-embed-text-v2/*.parquet
- config_name: cohere-embed-multilingual-v3
data_files: cohere/embed-multilingual-v3/*.parquet
- config_name: openai-text-embedding-3-small
data_files: openai/text-embedding-3-small/*.parquet
- config_name: openai-text-embedding-3-large
data_files: openai/text-embedding-3-large/*.parquet
- config_name: snowflake-arctic-embed
data_files: ollama/snowflake-arctic/*.parquet
size_categories:
- 100K<n<1M
Loading dataset without vector embeddings
You can load the raw dataset without vectors, like this:
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", split="train", streaming=True)
Loading dataset with vector embeddings
You can also load the dataset with vectors, like this:
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "openai-text-embedding-3-small", split="train", streaming=True)
# dataset = load_dataset("weaviate/wiki-sample", "snowflake-arctic-embed", split="train", streaming=True)
for item in dataset:
print(item["text"])
print(item["title"])
print(item["url"])
print(item["wiki_id"])
print(item["vector"])
print()
Supported Datasets
Data only - no vectors
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "no-vectors", split="train", streaming=True)
You can also skip the config name, as "no-vectors is the default dataset:
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", split="train", streaming=True)
AWS
aws-titan-embed-text-v2 - 1024d vectors - generated with AWS Bedrock
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "aws-titan-embed-text-v2", split="train", streaming=True)
Weaviate collection configuration:
from weaviate.classes.config import Configure
client.collections.create(
name="Wiki",
vectorizer_config=[
Configure.NamedVectors.text2vec_aws(
name="main_vector",
model="amazon.titan-embed-text-v2:0",
region="us-east-1", # make sure to use the correct region for you
source_properties=['title', 'text'], # which properties should be used to generate a vector
)
],
)
Cohere
embed-multilingual-v3 - 768d vectors - generated with Ollama
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "cohere-embed-multilingual-v3", split="train", streaming=True)
Weaviate collection configuration:
from weaviate.classes.config import Configure
client.collections.create(
name="Wiki",
vectorizer_config=[
Configure.NamedVectors.text2vec_cohere(
name="main_vector",
model="embed-multilingual-v3.0",
source_properties=['title', 'text'], # which properties should be used to generate a vector
)
],
)
OpenAI
text-embedding-3-small - 1536d vectors - generated with OpenAI
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "openai-text-embedding-3-small", split="train", streaming=True)
Weaviate collection configuration:
from weaviate.classes.config import Configure
client.collections.create(
name="Wiki",
vectorizer_config=[
Configure.NamedVectors.text2vec_openai(
name="main_vector",
model="text-embedding-3-small",
source_properties=['title', 'text'], # which properties should be used to generate a vector
)
],
)
text-embedding-3-large - 3072d vectors - generated with OpenAI
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "openai-text-embedding-3-large", split="train", streaming=True)
Weaviate collection configuration:
from weaviate.classes.config import Configure
client.collections.create(
name="Wiki",
vectorizer_config=[
Configure.NamedVectors.text2vec_openai(
name="main_vector",
model="text-embedding-3-large",
source_properties=['title', 'text'], # which properties should be used to generate a vector
)
],
)
Snowflake
snowflake-arctic-embed - 1024d vectors - generated with Ollama
from datasets import load_dataset
dataset = load_dataset("weaviate/wiki-sample", "snowflake-arctic-embed", split="train", streaming=True)
Weaviate collection configuration:
from weaviate.classes.config import Configure
client.collections.create(
name="Wiki",
vectorizer_config=[
Configure.NamedVectors.text2vec_ollama(
name="main_vector",
model="snowflake-arctic-embed",
api_endpoint="http://host.docker.internal:11434", # If using Docker
source_properties=["title", "text"],
),
],
)