---
language: 
  - en
    
license: "apache-2.0"
---
## Dataset Summary

This repo provides pre-built indexes for MSMARCO-v2.1 Doc Segmented corpus. All indexes are created through Anserini with different encoders (i.e., the lucene-inverted.msmarco-v2.1-doc-segmented.splade-v3.20250707.4039c3.tar.gz is encoded with the SPLADE-v3 model). This dataset is less than complete and will be updated over time to migrate other pre-built MSMARCO-v2.1 Doc Segmented indexes here.

## Load Dataset
An example to load the dataset:
```python
dataset = load_dataset("castorini/prebuilt-indexes-msmarco-v2.1-doc-segmented", data_files="lucene-inverted.msmarco-v2.1-doc-segmented.splade-v3.20250707.4039c3.tar.gz")
```

An example URL for downloading a pre-built index:
```
https://huggingface.co/datasets/castorini/prebuilt-indexes-msmarco-v2.1-doc-segmented/resolve/main/lucene-inverted.msmarco-v2.1-doc-segmented.splade-v3.20250707.4039c3.tar.gz
```