context-extension 0.1.1

Zero-Training Context Extension for Transformer Encoders via Nonlinear Absolute Positional Embeddings Interpolation

Zero-Training Context Extension for Transformer Encoders via Nonlinear Absolute Positional Embeddings Interpolation

Official implementation of "Zero-Training Context Extension for Transformer Encoders via Nonlinear Absolute Positional Embeddings Interpolation". Paper preprint is coming soon.

This implementation currently supports only models compatible with Sentence Transformers library.

Models

Models are available at HuggingFace:

Model	Context length	Language
idanylenko/e5-large-v2-ctx1024	1024	English

Installation

To install the package, use pip:

pip install "context-extension>=0.1.1"

Usage

After installing the package you may use extend-context-spline (recommended) or extend-context-linear scripts for embeddings interpolation.

Spline Interpolation

Use this for smooth, nonlinear interpolation to support arbitrary context lengths:

extend-context-spline \
  --model_name_or_path="intfloat/e5-large-v2" \
  --max_seq_length=1024 \
  --embeddings_attr_name="embeddings.position_embeddings" \
  --offset=0 \
  --output_dir="intfloat/e5-large-v2-ctx1024-spline"

Linear Interpolation

Use this to double the model's positional embedding range using linear averaging between consecutive embeddings:

extend-context-linear \
  --model_name_or_path="intfloat/e5-large-v2" \
  --embeddings_attr_name="embeddings.position_embeddings" \
  --offset=0 \
  --output_dir="intfloat/e5-large-v2-ctx1024-linear"

Both commands modify the positional embeddings of a model and save the updated model to the specified directory. You can then upload the resulting model to Hugging Face or use it locally for inference.

For models like RoBERTa that use special tokens in the first few positions, remember to set appropriate --offset argument.