README.md

---
license: cc-by-4.0
task_categories:
- text-retrieval
- question-answering
language:
- en
tags:
- retrieval
- text
- lance
pretty_name: fineweb-edu-lance
size_categories:
- 1B<n<10B
---

<center>
    <img src="https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/wwRnEQydH9qdRtFofIE-A.png" alt="FineWeb-Edu: The finest collection of educational content the web has to offer">
</center>

# FineWeb-Edu (Lance Format)

FineWeb-edu dataset consists of over 1.5 billion rows of educational web pages filtered from the FineWeb dataset.
Each passage ships with cleaned text, metadata, and 384-dim text embeddings for retrieval-heavy workloads.

## Why Lance?

Lance is an open-source format designed for multimodal AI data, offering significant advantages over traditional formats for modern AI workloads.

- **Blazing Fast Random Access**: Optimized for fetching scattered rows, making it ideal for random sampling, real-time ML serving, and interactive applications without performance degradation.
- **Native Multimodal Support**: Store text, embeddings, and other data types together in a single file. Large binary objects are loaded lazily, and vectors are optimized for fast similarity search.
- **Efficient Data Evolution**: Add new columns and backfill data without rewriting the entire dataset. This is perfect for evolving ML features, adding new embeddings, or introducing moderation tags over time.
- **Versatile Querying**: Supports combining vector similarity search, full-text search, and SQL-style filtering in a single query, accelerated by on-disk indexes.

## Quick Start

### Load with `datasets.load_dataset`

```python
import datasets

hf_ds = datasets.load_dataset(
    "lance-format/fineweb-edu",
    split="train",
    streaming=True,
)
# Take first three rows and print titles
for row in hf_ds.take(3):
    print(row["title"])
```

### Load with Lance

Use Lance's native connector when you need ANN search, FTS, or direct access to embeddings while still pointing to the copy hosted on Hugging Face:

```python
import lance

ds = lance.dataset("hf://datasets/lance-format/fineweb-edu/data/train.lance")print(f"Total passages: {ds.count_rows():,}")
```

> **⚠️ HuggingFace Streaming Note**
> 
> **You may hit rate limits on HuggingFace's free tier.** For best performance and to avoid rate limits, pass a token for an account with a
> Pro, Teams or Enterprise subscription (which come with much higher rate limits), or download the dataset locally:
> 
> ```bash
> # Download once
> huggingface-cli download lance-format/fineweb-edu --repo-type dataset --local-dir ./fineweb-edu
> 
> # Then load locally
> ds = lance.dataset("./fineweb-edu")
> ```
> 
> Streaming is recommended only for quick exploration and testing.

### Load with LanceDB

These tables can also be consumed by [LanceDB](https://docs.lancedb.com/), the multimodal lakehouse for AI (built on top of Lance).
LanceDB provides several convenience APIs for search, index creation and data updates on top of the Lance format.

```python
import lancedb

db = lancedb.connect("hf://datasets/lance-format/fineweb-edu/data")
tbl = db.open_table("train")
print(f"LanceDB table opened with {len(tbl)} passages")
```

## Index Creation

> [!WARNING]
> This dataset does not currently come with a pre-built ANN (vector).

To run vector search queries, you should download the dataset locally and build the index yourself.
The following steps show how to do this.

```bash
# Download once
huggingface-cli download lance-format/fineweb-edu --repo-type dataset --local-dir ./fineweb-edu
```

```python
# Load the local dataset in Lance
import lance
ds = lance.dataset("./fineweb-edu")

# Build a vector index as needed
# ds.create_index(...)
```
See the [Lance documentation](https://lance.org/quickstart/vector-search/#build-the-search-index) for the index building API.

## Quick Start

```python
import lance
import pyarrow as pa

lance_ds = lance.dataset("hf://datasets/lance-format/fineweb-edu/data/train.lance")

# Browse titles & language without touching embeddings
rows = lance_ds.scanner(
    columns=["title", "language"],
    limit=5
).to_table().to_pylist()

# Vector similarity from the on-dataset ANN index
ref = lance_ds.take([0], columns=["text_embedding", "title"])
query_vec = pa.array([ref.to_pylist()[0]["text_embedding"]],
                     type=ref.schema.field("text_embedding").type)

results = lance_ds.scanner(
    nearest={
        "column": "text_embedding",
        "q": query_vec[0],
        "k": 5,
        "nprobes": 8,
        "refine_factor": 20,
    },
    columns=["title", "language", "text"],
).to_table().to_pylist()
```

> **Hugging Face Streaming Note**
> - Streaming uses conservative ANN parameters (`nprobes`, `refine_factor`) to stay within HF rate limits.
> - Prefer local copies (`huggingface-cli download lance-format/fineweb-edu --local-dir ./fineweb`) for heavy workloads, then point Lance at `./fineweb`.

## Usage Examples

The steps below assume you've created an index on the dataset locally.

### 1. Sample documents

You can project specific columns (excluding the embeddings) and run filter queries on them.

```python
scanner = ds.scanner(
    columns=["title", "language", "text"],
    filter="language = 'en'",
    limit=5,
)
for doc in scanner.to_table().to_pylist():
    print(doc["title"], doc["language"])
    print(doc["text"][:200], "...\n")
```

### 2. Vector search for semantically similar passages

The example below shows a vector search on the `text_embedding` column.

```python
ref_doc = ds.take([123], columns=["text_embedding", "title", "text"]).to_pylist()[0]
emb_type = ds.to_table(columns=["text_embedding"], limit=1).schema.field("text_embedding").type
query = pa.array([ref_doc["text_embedding"]], type=emb_type)

neighbors = ds.scanner(
    nearest={
        "column": "text_embedding",
        "q": query[0],
        "k": 6,
        "nprobes": 8,
        "refine_factor": 20,
    },
    columns=["title", "language", "text"],
).to_table().to_pylist()[1:]
```

### 3. Full-text search with Lance FTS

```python
hits = ds.scanner(
    full_text_query="quantum computing",
    columns=["title", "language", "text"],
    limit=10,
    fast_search=True,
).to_table().to_pylist()
```

## Dataset Evolution

Lance supports flexible schema and data evolution ([docs](https://lance.org/guide/data_evolution/?h=evol)). You can add/drop columns, backfill with SQL or Python, rename fields, or change data types without rewriting the whole dataset. In practice this lets you:
- Introduce fresh metadata (moderation labels, embeddings, quality scores) as new signals become available.
- Add new columns to existing datasets without re-exporting terabytes of video.
- Adjust column names or shrink storage (e.g., cast embeddings to float16) while keeping previous snapshots queryable for reproducibility.

```python
import lance
import pyarrow as pa
import numpy as np

# Assume ds is a local Lance dataset
# ds = lance.dataset("./fineweb_edu_local")

base = pa.table({"id": pa.array([1, 2, 3]), "text": pa.array(["A", "B", "C"])})
dataset = lance.write_dataset(base, "fineweb_evolution", mode="overwrite")

# 1. Add a schema-only column (data to be added later)
dataset.add_columns(pa.field("subject", pa.string()))

# 2. Add a column with data
dataset.add_columns({"quality_bucket": "'unknown'"})

# 3. Generate rich columns via Python batch UDFs
@lance.batch_udf()
def random_embedding(batch):
    vecs = np.random.rand(batch.num_rows, 384).astype("float32")
    return pa.RecordBatch.from_arrays(
        [pa.FixedSizeListArray.from_arrays(vecs.ravel(), 384)],
        names=["text_embedding"],
    )

dataset.add_columns(random_embedding)

# 4. Bring in  annotations with merge
labels = pa.table({"id": pa.array([1, 2, 3]), "label": pa.array(["math", "history", "science"])})
dataset.merge(labels, "id")

# 5. Rename or cast columns as needs change
dataset.alter_columns({"path": "subject", "name": "topic"})
dataset.alter_columns({"path": "text_embedding", "data_type": pa.list_(pa.float16(), 384)})
```
You can iterate on embeddings, quality tags, or moderation fields while keeping earlier dataset versions available for reproducible experiments.

## LanceDB

LanceDB users can follow the following examples to run search queries on the dataset.

### LanceDB Vector Search
```python
import lancedb

db = lancedb.connect("hf://datasets/lance-format/fineweb-edu/data")
tbl = db.open_table("train")

# Get a passage to use as a query
ref_passage = tbl.limit(1).offset(123).select(["text_embedding", "text"]).to_pandas().to_dict('records')[0]
query_embedding = ref_passage["text_embedding"]

results = tbl.search(query_embedding) \
    .limit(5) \
    .to_list()
```

### LanceDB Full-Text Search
```python
import lancedb

db = lancedb.connect("hf://datasets/lance-format/fineweb-edu/data")
tbl = db.open_table("train")

results = tbl.search("quantum computing") \
    .select(["title", "language", "text"]) \
    .limit(10) \
    .to_list()
```

## Citation

You can cite the paper from orginal dataset (https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu) https://arxiv.org/abs/2406.17557 or this dataset:

```
@misc{lozhkov2024fineweb-edu,
    author       = { Lozhkov, Anton and Ben Allal, Loubna and von Werra, Leandro and Wolf, Thomas },  
    title        = { FineWeb-Edu: the Finest Collection of Educational Content }, 
    year         = 2024,  
    url          = { https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu },  
    doi          = { 10.57967/hf/2497 },
    publisher    = { Hugging Face }
}
```