intextus/colbertv2.0-onnx

This repository contains the ONNX version of colbert-ir/colbertv2.0 exported specifically for the intextus accelerated C++ late-interaction engine.

Model Details

• Base Model: colbert-ir/colbertv2.0
• Output Embedding Dimension: 128
• Opset Version: 18
• Exported using: PyTorch ONNX Dynamo Exporter

Usage with Intextus

from intextus import LateInteractionEncoder

model = LateInteractionEncoder("intextus/colbertv2.0-onnx")
q = model.encode_queries("What is late interaction?")
d = model.encode_docs("ColBERT uses late interaction with token-level representation.")